John Merkel - 2007 - Imperial Roman Production Of Lead And Silver In The Northern Part Of Upper Moesia Mt - Kosmaj Area

Abstract: Archaeometallurgical investigations of Imperial Roman lead and silver production near

the village of Stojnik, near Mt. Kosmaj (some 30 km south of Belgrade), represented a part of the

programme of excavations from 1983–1988 by a joint Serbian-American team. Metallurgical evi-

dence of production near the military fort (castellum) and surrounding civilian settlement was in-

vestigated. Technical studies included excavated samples of smelting slag and lead metal. The ear-

liest samples were dated on numismatic and ceramic evidence to the mid-2nd century. Production

intensified around the civilian settlement through to the 3rd and 4th centuries. Iron-rich slag com-

positions for lead smelting were consistent. Silver recovery by cupellation was thorough.

Keywords: Stojnik, Kosmaj, Imperial Roman mining, archaeometallurgy, lead, silver, smelting,

cupellation.

In1 Roman Mines in Europe, Davies (1935: 213–215) described the Impe-

rial Roman lead and silver mining between the modern villages of Stojnik,

Guberevac and Babe near Mt. Kosmaj, south of Belgrade (Singidunum). His

survey of the mining region described numerous opencast trenches, adits and

galleries which followed veins of “argentiferous cerussite with some galena,

and subsidiary blend, pyrites, chalcopyrite, and cinnabar” (loc. cit.). Several

Roman lead ingots, also known as “lead pigs”, with datable inscriptions from

the mining region were also reported; one to the reign of the emperor Marcus

Aurelius and another to Diocletian. Other archaeological and historical evi-

dence for the Roman province of Upper Moesia indicated significant levels of

IMPERIAL ROMAN PRODUCTION

OF LEAD AND SILVER

IN THE NORTHERN PART OF UPPER MOESIA

(MT. KOSMAJ AREA)

John F. Merkel

Institute of Archaeology, University College London

Glasnik Srpskog arheolo{kog dru{tva Journal of the Serbian Archaeological Society

23 (2007) 39–78.

1 I would like to thank M. R. Werner and V. Kondi} as well as P. Craddock, I. Freestone and H.

Mizota for their contributions to the archaeometallurgical research. At the Institute of Archaeology,

University College London, S. Groom and K. Reeves aided with XRF and EPMA. The manuscript

was reviewed in early stages by M. Hassall (Reader in Roman Archaeology) and P. Craddock (British

Museum). I thank them for their many worthwhile comments and suggestions. I also thank B.

Craddock for her illustrations in figs. 3–4. Funding for fieldwork was provided principally by the

Smithsonian Institution and National Endowment for the Humanities (M. Werner, Principal Inves-

tigator) with smaller grants from SUNY-Albany and the Archaeological Institute, Belgrade.

production from the lead and silver deposits from the 2nd to 4th centuries.

While Davies (loc. cit.) accepted an earlier estimate from 1892 which ’reck-

oned’ some 1,000,000 tons of total slag near the modern villages, an average

assay of the slag, reported as 55% lead and 0.0037% silver, was considered to

have been ’generous’. One of the earlier references cited by Davies (1935:

214–215) was The Geology and Mineral Resources of the Serb-Croat-Slovene

State by Wray (1921). This earlier report following World War I by a geologist

to the Commission Internationale de Ravitaillement contains additional techni-

cal detail, not included by Davies, concerning lead and silver production from

ancient slag in Serbia. Remarking on the extensive quantity of ancient slag,

Wray first suggested comparison of the Roman mining and smelting remains at

Kosmaj to Laurion – the famous silver production centre of the ancient Athe-

nians in Greece.

In ca. 77 A.D., Pliny (HN, XXXIII 31; XXXIV 47) described several pro-

cess related steps for extractive metallurgy of lead and silver, including smelt-

ing and refining (cf. Agricola 1950: 392; Healy 1978: 157, 180; idem 1999:

299–301, 320–326), but the accounts seem rather confused due to the complex-

ity of the ores and his unfamiliarity with the processes. The translation of tech-

nical terms is critical to achieve a metallurgical understanding of the various

steps and products. Writing in the 1st century, Pliny (loc. cit.) attributed the fin-

est silver to the deposits of Spain, before silver production from other prov-

inces had much impact on the Roman economy. Lead smelting produces slag

as a waste product. Slag can be readily identified and it is representative of an-

cient metal production (Bachmann 1982a: 2–7). Essentially, silver production

from lead ores can be usefully simplified to a three-step process: ore benefici-

ation, primary smelting to produce lead metal followed by cupellation of the

lead metal to recover silver (Tylecote 1962: 73–82). The expected samples

from Imperial Roman lead and silver production would include smelting slag,

litharge, lead metal and silver. For example, litharge (PbO) is the representa-

tive waste material that serves as evidence for cupellation as a production step.

Other remains found in an archaeological excavation should include fragments

of lead ore, refractory-ceramic furnaces and crucibles as well as possible wash-

ing/beneficiation installations to concentrate the ore. However, not all of the

metallurgical remains are readily identified during archaeological fieldwork.

Representative sampling of archaeometallurgical remains (Orton 2000: 182)

and technical studies of the expected metallurgical samples are often necessary

to identify, document and assess each production step. The metallurgical data

from the analytical laboratory need to be integrated better into the archaeolog-

ical investigation of Imperial Roman lead and silver production in the region.

Within the mining region, the presence of a Roman town was reported

near Stojnik by Davies (1935: 214–215), but the existence of a major Roman

fort next to the town was not acknowledged (Tomovi} 1995: 204). Wray (1921:

66) also did not mention a Roman fort near Stojnik. As a result, a connection

GSAD/JSAS 23 (2007) Research Papers and Treatises

40

of the Imperial Roman military with lead and silver production was not estab-

lished in these early accounts. The excavation by Vuli} in 1911–1913 of the fort

represents a critical aspect of the investigation. Most of the subsequent re-

search in the region has focused upon aspects of the ancient history, epigraphy

and military/settlement archaeology; not upon the technical analyses of the

Roman “industrial” remains and metallurgical evidence. The Imperial Roman

military connection with the mining and smelting represents a distinctive fea-

ture for Upper Moesia (Mücsy 1974: 195, 216; Du{ani} 1977a: 52–93; idem

1977b: 237–247). As a result of continued research on the Roman occupation

and development of the interior of Upper Moesia, the mining region around

Mt. Kosmaj is now understood to have operated within a larger Imperial Ro-

man administrative district including additional mines, settlements and military

forts further away at Avala and Rudnik (idem 1977a: 93). The name of the dis-

trict was probably metalla Tricorn(i)ensia and the administrative organisation

could have covered at least three local mining regions producing lead and sil-

ver (Du{ani} 1995a: 221). The mining district had been opened under Marcus

Aurelius and administered through Rome as a territorium metalli (Mücsy 1974:

195). Some 5000 ancient mining shafts have been counted in the district, but

evidence for the actual number is more obscure today (Tomovi} 1995: 208).

The economic development of the interior of Upper Moesia for the ex-

ploitation of lead and silver is related to military occupation south of the Ro-

man frontier line of legionary forts (limes) along the Danube (Du{ani} 1977a:

93; idem 1977b: 244). The legionary fort (castrum) at Singidunum had been es-

tablished in 86 A.D. by the Legio IV Flavia Felix. The lead and silver mining

district was first protected (ca. 167–169 A.D.) by auxiliary mounted troops, I

Ulpia Pannoniorum equitata, moved from Upper Pannonia to a new, secondary

fort built near the modern village of Stojnik (Du{ani} 1977b: 238). Reflecting

the importance of lead and silver production within the district, after 169 A.D.,

a mounted cohort, II Aurelia nova, was formed to replace the other auxiliary

unit at the military fort near Stojnik (Mócsy 1974: 216; Du{ani} 1977a: 79;

Wilkes 1981: 519). The significance of mounted cohorts within the mining dis-

trict has been assessed by Du{ani} (1977b). Other Roman auxiliary forts with

mounted cohorts were present within the mining district at Rudnik and

Tricornium, closer to Avala (ibid.: 247). There were also other dispersed Ro-

man settlements (fig. 1) within the mining district (Du{ani} 1977a: 91). The

mounted cohorts provided security on the road to Singidunum and south, es-

corts for metal transport and patrols of the mining district. Military and politi-

cal factors necessitated greater security within the mining district during the

wars with the Marcomanni and other tribes along the Danubian Frontier. The

auxiliary fort (castellum), near Stojnik, and the associated civilian town (vicus –

for a settlement around an auxiliary fort) represent a production centre and

also possibly the administrative centre for lead and silver production within the

mining district (Du{ani} 1977a: 78; idem 1995a: 221; Tomovi} 1995: 206).

41

J. F. Merkel Imperial Roman Production of Lead and Silver

Within the perimeter walls of the fort, the construction of the hospital building

can be dated to 179 A.D. (Mócsy 1974: 195), late in the reign of Marcus

Aurelius. In the southwest area of the settlement, there are also foundations of

a Roman bath (fig. 2). This major building, discovered in 1958, some 50 metres

south of the castellum gate, was later adapted to a major pagan cult building,

and then finally altered for use as an early Christian church and dated to the

middle of the 4th century (Werner 1990: 221; Tomovi} 1995: 206; Du{ani}

1995a: 221). It is significant to note that large blocks of tapped lead smelting

slag can now be observed to have been incorporated into the structural founda-

tions as building material. Into the later 3rd and 4th century, silver production

and prosperity were thought to have declined (Mócsy 1974: 225). Administra-

tion of the mining district came under some civilian control in the later 3rd cen-

tury. According to Mócsy (1974: 224–225), the northern part of the district came

later under an administration centre at Aureus Mons located east of Singi-


42

Fig. 1. The Imperial Roman auxiliary fort (castellum) is located between the modern villages

of Stojnik, Guberevac and Babe. Several major Roman cemeteries provide further evidence

for the wealth of the mining district. The extent of slag on hilltops and eroded on the slopes

was not mapped in detail, so earlier published estimates of total slag weights near Stojnik

have not been revised. Furthermore, large quantities of Roman slag have been removed for

modern reprocessing.

43


Fig. 2. The auxiliary fort (castellum) and associated civilian town (vicus) are illustrated in

this redrawn plan which also shows locations of excavation trenches (Trial Trenches 2 and 3)

from 1983. The foundations of a Roman bath are located under the backfilled mosaic pave-

ment building (plan provided by M. Werner).

dunum. However, Du{ani} (1995a: 219–225) presents evidence for greater ad-

ministrative continuity under direct ownership of the Emperors from Marcus

Aurelius until after Constantine. Prosperity and wealth seems to have been con-

centrated foremost around the fort and settlements near Stojnik within the min-

ing district. Eventual decline in production and prosperity cannot necessarily be

linked to changes in administration. The greater wealth exhibited in the Mt.

Kosmaj area (Mócsy 1974: 216; Du{ani} 1977a: 78) is clear from the nearby ne-

cropolis at Guberevac with inscriptions, tomb features and precious metal grave

goods, such as silver/gold jewellery. Many “foreigners” are named in inscriptions.

According to Mócsy (1974: 217), Dalmations, Thracians, south Moesians and

others from western provinces can be identified. How could these archaeological

aspects relate to the technical, metallurgical evidence for Imperial Roman lead

and silver production?

The estimated total of 1,000,000 tons of Roman lead and silver slag has

been cited to emphasize the huge scale of production centred between the

modern villages of Stojnik, Guberevac and Babe within the mining district

(Davies 1935: 213; Du{ani} 1977a: 78; Tomovi} 1995: 205). Including Avala

and Rudnik within the district, the total estimate would be even larger, but

there are no published total slag estimates for Roman exploitation of these

other mines. The Roman slag from near Mt. Kosmaj has been reworked at var-

ious times in the 20th century to recover lead and silver, so some of the esti-

mated total has been removed. However, the full extent of modern reworking

is uncertain. There are annual production figures for metals, but usually not

enough to gain more than an impression of large scale modern processing

throughout the district to recover lead and silver. Various episodes of modern

reworking the Roman slag, principally starting on a large scale in 1907, contin-

uing irregularly and ending in 1956 (Tomovi} 1995: 205). Production is re-

corded as variable. For example, high production levels for lead production in

Serbia were reached in 1908–1910 and attributed by Wray (1921: 9) to rework-

ing “lead-slags in the neighbourhood of Ripanj” some 3 km from Avala. Silver

production in Serbia correspondingly increased at this time, too. At Babe,

Wray (1921: 66) reported some 3000 metric tons of lead were obtained in 1910

and 1911 from reworking ancient slag. The ancient slag from Babe was also re-

ported to have averaged 6–7% metallic lead (loc. cit.), so the calculated quan-

tity of slag removed in just two years from near Babe was 40–50,000 tons. Ac-

cording to Davies (1935: 213), earlier 20th century industrial buildings and

’smelter’ are located nearby on the north of the Pruten valley (fig. 3). Other

mines, smelting sites and settlements around Avala and Rudnik in the Imperial

Roman administrative district were outside the limits of the current research.

The industrial buildings in the Pruten valley are now ruins, but modern slag is

present in large quantities which need to be assessed.

With such “enormous” estimates of Imperial Roman lead and silver slag,

centred near Stojnik, this mining district would have been on a scale compara-


44

ble to only a few other investigated sites, perhaps second only to Roman Rio

Tinto in south-west Spain with an original estimate by G. Douglas of some

15,000,000 tons of silver smelting slag (Rothenberg and Blanco-Freijeiro 1981:

177). Early estimates of total slag weights at Rio Tinto have been revised to

lower absolute levels with further investigation, correcting assumptions of slag

dump depth with modern drill cores. Other Roman lead and silver production

sites do not have such large estimates of total slag, so the mining districts cen-

tred at both Rio Tinto and Mt. Kosmaj must still represent the highest Roman

production levels for silver. Davies (1935: 94) recognized a trend that silver

production from Iberian Peninsula was beginning to be replaced after the end

of the 1st century from other Roman provinces in the Balkans. Imperial Ro-

man production of silver from Spain also seems to decline at many mines into

the 4th centuries (Domergue 1990: 213–214), but precise dating of production

levels is often difficult. Few Spanish mines have had total weights of ancient

slag estimated for lead and silver production, so other worthwhile comparisons

are not possible.

45


Fig. 3. This map of the Pruten valley and position of the auxiliary fort also illustrates the distri-

bution of 16 test pits (small black squares) excavated principally by P. Craddock and the author

outside the settlement. The test pits were made to obtain slag samples, check any associated

metallurgical remains and confirm dating to the 2nd to 4th centuries (map by B. Craddock).

For comparison to Laurion, an early estimate of the total quantity of slag

was 1,550,000 tons, but estimates for production by Conophagos (1980: 145)

and Bachmann (1982b: 246) were 1,400,000 tons lead with 3,500 tons silver for

the Classical Period in Greece. Davies (1935: 251) reported a high estimate of

some 2,000,000 tons of ancient slag. Rehren, Vanhove and Mussche (2002: 43)

use the reported value of 1,500,000 tons of ancient slag at Laurion. The esti-

mate of Roman slag weight near Stojnik and Mt. Kosmaj would be lower, but

the estimate for the mining district including Avala and Rudnik would seem to

have been on a similar scale.

The scale of silver production in the mining district centred on Stojnik

must rely upon evaluation of the quantity of ancient slag as well as the quality

of ore. Several published analyses of the ore grade from modern mining pres-

ents some guide for assessing ancient silver production. Argentiferous galena

from the Crveni Breg workings at Avala were reported from about 1901 to

have yielded 0.183% to 0.4308% Ag (Wray 1921: 65). A higher figure of 6000 g

of silver per ton of ore (0.6% Ag) is reported by Tomovi} (1995: 208) for ex-

ceptionally rich ore in the district. From the nearby modern mines at Rudnik

(Radi} and Vukovi} 1957: 128–130), under modern working conditions, a value

of 0.33% Ag in the galena concentrate has been published (Jankovi} 1967). No

compositional data have been available for the ore mined by the Romans

nearer Stojnk, Guberevac and Babe. Reported modern assays reflect lower

economic silver concentrations (Grèti} and Jelenkovi} 1995: 20–22).

In his article, Du{ani} (1977a: 93) stated, “The main conclusion of the

present study is that Roman mining in the Danubian provinces – and, no

doubt, also elsewhere – was more of a system than is commonly thought.” This

conclusion was based upon historical evidence for the administration of the

mining in these Provinces of the Roman Empire. Did the Imperial Roman

“system” for the exploitation of mineral wealth in Upper Moesia extend to ad-

ditional technical aspects? What perspective can be added from technical in-

vestigation of the metallurgical remains, such as the smelting slag from near

Stojnik? There was an obvious need for new, integrated archaeometallurgical

research for this significant production centre in Upper Moesia in comparison

to other centres (Tomovi} 1995: 204), including Rio Tinto where silver-rich

jarosite (hydrated sulphates with variable metal concentrations) ore was

smelted (Rothenberg and Blanco-Freijeiro 1981; Craddock et al. 1985) and

Laurion where argentiferous galena with other mixed sulphides was smelted

(Conophagos 1980; Pernicka, 1981; Bachmann 1982b). The technical studies of

the archaeometallurgical remains, especially the smelting slag and lead metal,

near Stojnik, Guberevac and Babe present the opportunity to interpret the re-

sults within a well documented context for Imperial Roman economic develop-

ment of Upper Moesia.


46

Excavations, Sampling and Dating

The excavated remains of the Roman fort near Stojnik are now covered

with trees and thick undergrowth. Due to reforestation, the perimeter walls are

not visible from the Pruten river valley on the crest of the hilltop. However, on

some adjacent, steep hillsides, slag dumps have eroded and large quantities are

readily examined (pl. I/1) near the settlement. During 1983 to 1988, research by

V. Kondi}, M. Werner, M. Tomovi} and the author, again focused upon aspects

of mining and metallurgy as well as technical investigations of the ancient re-

mains near Stojnik.2 Sample selection from the archaeometallurgical remains

anticipated the common production steps for ancient lead smelting of silver-rich

lead ores followed by cupellation to recover silver from the lead metal as out-

lined by Tylecote (1962: 75–82). The expected remains would include smelting

slag, litharge and lead metal. Other more exceptional finds might possibly in-

clude remains of smelting furnaces, tuyeres, cupels or cupellation hearths, possi-

bly “lead pigs” and silver ingots. Roasting the ore to decrease the sulphur con-

centrations would be unnecessary for low concentrations of galena in a relatively

silver-rich lead carbonate ore. Representative samples of all of the expected

metallurgical production steps were actively sought in the excavations.

Samples of slag were collected in 1983–1984 from archaeological excava-

tions – principally Trial Trenches 1, 2, 3, 7 and 8 – of the civilian settlement

and directly outside the fort. Whole tapped slag blocks from the excavations

near the fortification and from within the settlement exhibit clear flow struc-

tures on the upper surface of the plano-convex form. The flow structure away

from a single point (pl. I/2) indicates the molten slag was tapped from a fur-

nace. Almost complete blocks of tapped slag weigh approximately 20 kg, so the

operational volume of the furnace would have been approximately 5–8 litres

up to the level of the tuyere(s). No intact furnaces have been found in the ex-

cavations. There are, however, numerous isolated examples of slagged stone,

brick and clay lining from the excavated trench, presumably debris from smelt-

ing furnaces. Davies (1935: 213) described furnace remains from the village of

Babe (approximately 2 km northeast from Stojnik): “The use of brick [used for

furnace construction] at Babe South and Rudnik Jezero is almost certainly Ro-

man. The Romans also normally lined their furnaces with clay, as at Babe

47


2 Archaeological excavations of the military and civilian complex were undertaken under the di-

rectorship of V. Kondi}, Archaeological Institute, Belgrade, with M. Werner, State University of New

York at Albany. A preliminary report has been published by Werner (1990), but a full excavation re-

port with detailed plans of the site, structures, stratigraphy and finds remains unpublished. Fieldwork

and technical investigations of the lead and silver productions remains were addressed by the author as

part of the larger research project and co-ordinated through M. Werner. The regional survey of ancient

mines, with documentation and provisional dating of several underground working, was conducted by

M. Tomovi}, Archaeological Institute, Belgrade, and published in 1995. The aim of the current report

is to contribute new analytical data on technical aspects of lead and silver production to the investiga-

tion of Imperial Roman exploitation of mineral resources in Upper Moesia.

South, Blagaj, and Rudnik Jezero. Babe North has no brick and little clay lin-

ing, Majdanpek Tenka and Rudnik Village stone unlined furnaces.”

No fragments of ceramic tuyeres have yet been found in any of the slag

dumps around the castellum and vicus near Stojnik. However, excavations of

the smelting slag from the Roman settlement produced many examples of

curved thin fragments of slag in the dumps. These curved fragments suggest

the use of iron pipes as tuyere(s) in the furnace. The iron pipe would protrude

slightly into the furnace interior and direct air from the bellows toward the bot-

tom of the furnace for combustion of the charcoal fuel. The slag fragments are

relatively thin at 0.5 cm with a ropy exterior surface. The interior is smooth and

does not exhibit any evidence of contact with a ceramic. The lack of ceramic

tuyeres and similar curved fragments of slag from Site 47 at Rio Tinto was also

interpreted as suggesting that iron pipes had been used (Rothenberg and

Blanco-Freijeiro 1981: 108). More positive evidence comes from excavation of

a cupellation area at Sardis, dated 6th century B.C., where an unusual “iron

blowpipe nozzle; heavily corroded and encrusted with sand” was discovered

(Ramage 2000: 231). However, many more ceramic tuyere fragments were

found associated with cupellation at Sardis. Similar curved slag fragments were

produced in copper smelting experiments, which also used iron pipes as

tuyeres, by Merkel (1990). The use of copper pipes as tuyeres for smelting fur-

naces is mentioned by Agricola in the 16th century (1950: 376), but apparently

the practice of using metal tuyeres can be much earlier.

An interesting slag specimen from

Trial Trench 7, excavated in 1985 from

a layer 60–80 cm, is interpreted as rep-

resenting a partial blockage or accre-

tion (Ferruminata or ’sow’) of an in-

clined tuyere into a lead smelting fur-

nace. As the level of molten slag in the

furnace reached the tuyere, slag would

begin to solidify and block the tuyere.

The almost complete specimen (fig. 4)

had been discarded into a slag dump.

The inner surface, with a diameter of

about 25–27 mm, is relatively smooth.

The outer surface is more irregular and

typical of molten slag exhibiting various

flow structures and drips. The vent di-

ameter is about 13 mm and directed

downward judging from the direction

of the slag drips oriented to the bottom

in this drawing.

In 1984, Trial Trench 8, measur-

ing two by five metres, was excavated


48

Fig. 4. A fragment of slag interpreted as a

blockage around a tuyere in the smelting

furnace. The opening on the left measures

25–27 mm. Note the solidified ’drips’ of

slag next to the vent. The downward direc-

tion of the ’drips’ help the interpretation

of this specimen (drawing by B. Craddock).

immediately outside the overgrown outer defensive wall of the Roman fort,

principally to assess the association between the fortification and smelting slag.

In Level 8, among the slag fragments and larger blocks of tapped slag, it was

most unusual to discover three sherds of terra sigillata and a silver denarius,

dated to mid-2nd century (Popovi} 1984) resting directly upon undisturbed soil

under the slag. The coin from Trial Trench 8, Level 8 has a laureate head of

Antoninus Pius with the legend, ANTONINUS AVG PIUS.P.P. on the ob-

verse and the reverse has a figure of Felicitas holding caduceus and cornucopia

with the legend TR POT COS II (loc. cit.). This is attributed to a third issue of

silver denarii in 139 A.D. (cf. Mattingly 1968) which would predate the pro-

posed date of 167–169 A.D. for foundation of the fort. The condition of the

cleaned coin was very good. Thus, the earliest Imperial Roman and best dated

smelting slag from lead and silver production was found directly outside the

outer fortification wall, presumably where security was greatest. The excava-

tion of Trial Trench 8 recorded Levels 2–8 below the surface layer, but no stra-

tigraphy was apparent in the slag dump. The slag dump excavated in Trial

Trench 8 is interpreted as representing a single dump of principally large

blocks of tapped slag. It is significant that small pieces of limonite (hydrated

iron oxide) were also observed with the slag. Enlarging Trial Trench 8 may

have eventually found smelting furnace remains in situ and confirmed the pres-

ence of the working surface up to the perimeter wall of the fort or even adja-

cent slag dumps. Since the interior of the Roman fort had been excavated, no

additional work was undertaken within the outer defensive walls.

Numismatic evidence for dating mining and metallurgical activities is ex-

ceedingly rare. The find locations must be clearly associated with metallurgical

remains to be of value for dating; unexcavated surface finds can only provide a

general guide. Davies (1935: 213) noted that coins dated to the 2nd and 3rd cen-

turies had been reported in association with slag heaps from Babe and

Guberevac, but without providing more detail. The use of silver from the min-

ing districts within the province for coinage has been investigated by Du{ani}

(1995b). The mint issues are complicated by evidence for ancient forgers’ dies

(loc. cit.). The Imperial Roman mint at Viminacium which was the provincial

capital of Upper Moesia also produced silver coins (Ra{kovi} 1995). Large

quantities of lead metal were also produced, based on dated “lead pigs” from

the district (Du{ani} 1995a: 221). Before military occupation of the interior of

Upper Moesia, no smelting sites have been confidently dated any earlier dur-

ing the mid-2nd century in the Mt. Kosmaj area. No earlier or medieval mining

and silver production has been documented on a significant scale (Tomovi}

1995: 204).

The archaeometallurgical investigation of the large production area

around the Roman fort and settlement at Stojnik, with such enormous quanti-

ties of smelting slag, required strategic sample selections and limited research

aims. The essential objective for fieldwork was to improve dating of the spread

49


of slag from nearby the fort to the attached civilian settlement and contiguous

level areas now used as pasture. The present surface of level areas on the hill-

tops and ridges has been contoured for agricultural use. Hillside surface re-

mains of slag were not investigated, due to possible mixing from weathering

and erosion. In the 1986 mining survey by Tomovi} (1995: 204), slag was not

found around the mining shafts. Furthermore, excavation permits did not ex-

tend much beyond the immediate area of the fort and settlement.

In 1985, a series of 16 test pits in selected areas just outside the settlement

(fig. 3) were excavated principally by P. Craddock and the author to provide

samples for further technical investigation with associated ceramic evidence

for dating. These pits measured 2 x 1 m and were excavated to a depth of ap-

proximately 1 m. The excavations tried to follow some stratigraphy within the

slag dumps, but most test pits were excavated simply in 20 cm levels to provide

some control on sample selections from the tapped slag. Slagged and vitrified

stone and clay were interpreted as furnace fragments in the slag dumps. No

working surfaces were found. It was noted that the slag seemed to extend much

further in depth in almost all of the selected test trenches. In the accompany-

ing analytical tables, the samples are designated by test pit number and depth

level. Fragments of undiagnostic ceramic were found in some of the test pits,

but there was no other evidence to date remains from the test pits to periods

other than broadly 3rd and 4th centuries. No sherds of terra sigillata were found

among these more outlying slag dumps. The later approximate dates for lead

smelting slag extends production beyond the immediate security of the Roman

fort, presumably as military conditions improved. Additional archaeometallur-

gical fieldwork to obtain confirmation for this general model for extending

production, perhaps further toward Guberevac and Babe, however, became

impossible in subsequent years. Investigation of mining, smelting and settle-

ment remains was not possible at Avala or Rudnik within the mining district.

Our quick inspections of several accessible mines, opencast and adits, in

1984–1985 did not produce any alternative dating. No samples of appropriate

high-grade silver-rich lead ores were found in the mines or excavations. It is

significant; however, that iron oxide is also readily available in association with

the lead mineralization and mines in the region. Samples examined were pre-

dominantly low-grade limonite, not hematite. According to the local workmen,

many Roman lamps and other items had been removed from known mines.

The 1986 mining survey by Tomovi} (1995) investigated several mines and pro-

duced detailed plans of the workings which followed veins of ore. However, no

new data were obtained for the quality or quantity of argentiferous cerussite

with some galena. Several other mines were identified where high-grade iron

ore was removed (ibid.: 211), presumably for use as a flux added to the lead

ore to produce the observed slag compositions most useful for lead smelting.

As a waste product from smelting, large quantities of slag were discarded and

accumulated as “dumps” to be investigated by archaeometallurgists. Heavy


50

blocks of tapped slag would not be moved very far from the smelting furnace.

The chemical composition of the slag is characteristic of the efficiency of the

smelting process. It reflects the quality and requirements of the lead ore for

smelting. The technical investigation of metallurgical slag from an ancient site

needs an interdisciplinary approach with contributions from archaeologists and

historians as well as geologists, mining engineers and metallurgists. Each can

provide a different academic perspective on the complex “industrial” remains

from ancient metal production.

Technical Studies of Selected Metallurgical Remains

From a technological perspective, the identification of metallurgical pro-

cess-related steps and materials used for lead and silver production needed to

be verified. Technical studies are used to document production steps using sev-

eral different analytical techniques;3 each with its own advantages and disad-

vantages. Compositions of slag, litharge and corroded lead samples have been

investigated using atomic absorption spectrometry (AAS), x-ray fluorescence

(XRF) and electron probe microanalysis (EPMA). Phase constituents of sev-

eral slag samples have been confirmed with XRD and metallographic sections.

Eight samples of slag submitted for analysis at the British Museum were

selected to represent tapped slag compositions. Tapped slag compositions are

judged to be the most representative parameter to estimate consistent, desired

routine production values. Tapped slag compositions can be controlled, to a

degree, empirically by simply adjusting proportions, by weight or volume, for

ore, flux and fuel charged into the smelting furnace as demonstrated in copper

smelting experiments by Merkel (1990). The tapped slag samples were pre-

pared and analyzed by Freestone (1990) using atomic absorption spectrometry

(AAS) following procedures for silicates published by Hughes, Cowell and

Craddock (1976). The results are presented in tables 1–2, recalculated from el-

emental concentrations as slag-forming constituent oxides. Two samples of

tapped slag in (table 1) are from Trial Trench 8, Level 2, higher in the single

dump than the sherds of terra sigillata and coin dated 139 A.D. The results for

four samples from the excavations of the settlement range from 20–27% FeO

with about 5–8% CaO. The lead concentrations in the tapped slag are variable,

but concentrations for three samples fall between 3% and 7% Pb with one high

value of 22%, considered as an outlier from the distribution. Silver in the

51


3 Data have been assembled for this report from unpublished manuscripts on file from Merkel

(1984), Craddock (1985), Freestone (1990) as well as a supervised (by the author) M.Sc. dissertation

by Mizota (1995) at the Institute of Archaeology, UCL. Analytical work on the metallurgical samples

has utilized equipment at the Department of Anthropology/Peabody Museum at Harvard Univer-

sity, the Department of Scientific Research at the British Museum and the Wolfson Archaeological

Science Laboratory at the Institute of Archaeology, UCL.

tapped slag is low with an observed range from 0.002% to 0.018% Ag. There is

no correlation between Pb and Ag concentrations for these samples.

Table 1. Compositional analysis using AAS of slag samples collected from excavations

around the settlement, including public buildings and near exterior walls of the fort.

Sulphur and barium were not sought for these samples. Data provided by Freestone

(1990).

SAMPLE FeO CaO MgO MnO K2O SiO2 Al2O3 Pb ZnO Cu Ag

Trench 1,

St.85.X.525,8 7.48 1.76 1.16 2.28 36.6 8.00 4.96 5.85 0.019 0.008

St.83.X.10 26.6 6.45 1.37 1.16 1.67 37.2 6.96 6.69 5.53 0.026 0.003

St.84.8L2a 24.3 8.20 2.01 1.19 2.89 37.4 10.1 2.57 5.83 0.010 0.018

St.84.8L2b 20.3 4.62 0.91 0.40 0.71 33.9 5.02 22.0 3.49 0.043 0.006

The compositions of four samples of tapped slag from the more outlying

test pits (table 2), and provisionally later in date, have a range of 28–29% FeO,

which is slightly higher and more consistent than the samples from the settle-

ment excavations. Interestingly, the lead concentrations of the tapped slag all

fall within an observed range from about 4–7% Pb.4 The concentrations of sil-

ver are slightly lower with about 0.003% to 0.007% Ag. The measured concen-

trations of lead are consistent with lead concentrations of 6–7% reported by

Wray (1921: 66) as a production figure for Roman slag reworked from Babe in

1910 and 1911. This lead concentration relates to recovery from processing of

some 3000 tons of Roman slag, so it can be interpreted as a calculated average

from a normal distribution of lead concentrations for the Roman smelting slag.

Table 2. Compositional analysis of slag samples collected from selected test pits exca-

vated in 1985. Samples were selected on a representative basis to investigate tapped

slag compositions from across the nearby level industrial areas. Sulphur and barium

were not sought for these samples. AAS data provided by Freestone (1990).


Test 2, 0–20 cm 29.0 5.02 1.35 2.93 1.55 35.8 5.93 5.89 5.17 0.026 0.003

Test 2, 50–80 cm 28.7 5.93 1.27 1.14 1.83 35.5 7.13 7.21 4.38 0.026 0.003

Test 2, 50–80 cm 28.2 6.07 1.51 2.06 2.23 33.7 7.82 6.06 5.98 0.028 0.002

Test 6, 0–20 cm 27.9 6.38 1.73 1.25 2.24 33.4 8.05 6.87 3.90 0.029 0.007


52

4 These Pb concentrations are similar to those reported by Wray (1921: 66) for ancient slag from

Babe (cf. fig. 1).

Petrographic examination and XRD of the selected samples showed prin-

cipally fayalite (Fe2SiO4) with leucite (KAlSi2O6) in a glassy matrix present as

well as galena and some zinc-iron sulphide (Craddock 1985). The glassy matrix

seemed complex and variable in concentration, but not investigated in detail.

Samples of non-tapped slag, in other words “furnace slag” possibly with inclu-

sions of refractory furnace wall or lining, would be expected to be more vari-

able in composition and possibly higher in lead and silver concentrations.

Sections of smelting slag were made for microanalytical investigation by

Mizota (1995). Metallic lead prills and entrapped galena in the smelting slag

were checked for silver concentrations using EPMA. Minor lead sulphide com-

ponents (still exhibiting cubic cleavage surfaces in section characteristic of ga-

lena) from the ore charge are often found unreacted in the slag as the result of

reducing conditions in the smelting furnaces. In twenty-five selected slag sam-

ples, the silver values from EPMA were unexpectedly low (between 0.01–0.06%

Ag) in the metallic lead prills and lead sulphides. The detection limit for Ag in

the lead sulphide and lead metal was, at best, approximately 0.01% using

EPMA. Entrapped, unreacted lead sulphides within the molten iron-rich slag

were characteristic of the microstructures; low silver concentrations did not vary

significantly between the lead prills and lead sulphides.

Non-destructive XRF was used to identify constituent elements in the

lead-rich slag (fig. 4) on the tuyere ’blockage’ from Trial Trench 7. The ap-

proximate ratio of iron to silicon and aluminium (table 3) links this specimen

with fayalite-type lead smelting slag compositions. The relatively high iron

would suggest a higher melting temperature and greater viscosity than the

other examples of tapped slag. The relatively high concentration of lead and

silver in the iron-rich slag with 25% Pb and more than 0.08% Ag as well as al-

most 4% sulphur also link this specimen to primary smelting of argentiferous

lead ore. These compositional data for Fe, Pb, Zn and S also support the inter-

pretation of this specimen as a blockage at a tuyere within the smelting fur-

nace, rather than another production step, such as cupellation for silver recov-

ery from lead metal. The non-destructive XRF results are not very precise for

the reported concentrations, but the data are significant. These compositional

data illustrate a relevant point for sampling and interpretation of the slag com-

positions. More viscous slag partially reacted with the furnace walls or a slag

blockage on a tuyere contains more lead and silver than the tapped slag. Thus,

the ratio of silver to lead from this sample indicates a higher level near 0.32%

Ag in the Pb as an indicator of a high ore quality. This ratio is not supported by

observed concentrations of silver in the lead sulphide entrapped in the tapped

smelting slag. Most of the Imperial Roman smelting slag from the district pro-

cessed historically for recovery of silver was tapped slag which is correspond-

ingly lower in silver concentrations.

53


Table 3. Non-destructive analysis of irregular slag surfaces using XRF is only semi-

-quantitative. The data should be interpreted only as low and high values; whole weight

percent are presented as a guide. Nevertheless, this Ag concentration is among the

highest for slag from the site suggesting a calculated ratio of about 0.3% Ag to the lead

metal alone. This gross approximation would seem a circumstantial estimate of the best

grade lead ore with some 0.3% Ag in a galena concentrate. The iron is higher and the

calcium much lower. Ba was not detected. This slag specimen interpreted as a blockage

from a tuyere. There was no adhering ceramic. It is suggested that iron pipes were used

as tuyeres at the site.

SAMPLE Fe Ca Mg Mn K Si Al Pb Zn S Cu Ag

Trench 7, 60–80 cm 41.7 nd 1.02 1.38 nd 13.5 6.68 25.0 5.28 3.63 0.26 >0.081

From massive slag dumps near the ruined 20th century industrial buildings

along the Pruten river valley identified with the reworking of Roman slag

(Tomovi} 1995: 205), two other modern samples were collected. Analysis of

these modern slag samples by Freestone (1990) indicated lower lead and silver

concentrations (table 4) than Roman tapped slag samples from the excavations

(tables 1–2). Otherwise, the compositions of the modern slag are not very dif-

ferent from the Roman slag; there is only slightly more CaO and Al2O3, per-

haps from additional fuel and reaction with modern refractory from remelting.

The ZnO concentration is slightly lower, suggesting some loss of zinc with

higher modern temperatures for remelting. The type of modern furnace used

for remelting is uncertain, but possibly could have been a reverberatory fur-

nace simply for remelting the slag. More documentation and industrial re-

search is required on modern reworking of the Roman slag.

Table 4. Compositional analyses of two slag samples from industrial remelting of Ro-

man lead slag reportedly in the late 19th and early 20th centuries (British Museum Sam-

ples numbers St.85.X.54). Surface collection from large dumps associated with ruined

modern industrial buildings in a nearby Pruten river valley. Sulphur and barium were

not sought for these samples. AAS data provided by Freestone (1990).


St.85.X.54a 29.5 8.52 1.48 1.56 2.31 34.5 9.73 1.49 4.07 0.018 0.001

St.85.X.54b 27.5 9.43 1.81 1.01 2.33 35.8 9.60 1.35 4.02 0.016 0.002

A location with a large quantity of litharge and other evidence from

cupellation to recover silver from the primary smelted lead metal was anticipated,

but only a few pieces of litharge were found in the slag dumps from Trial Trench 5

and Trial Trench 10 located within the settlement. The samples are dense with a

light coloured weathered surface, upon fracture the structure is crystalline and

light red/pink in colour. Non-destructive XRF of several litharge samples did not

detect silver on a clean, flat sample surface. Other approximate concentrations

were >85% Pb, 5% Zn, >7% Si, 2% Al and 1% Fe. Minor and trace element


54

concentrations included Mg, Ti, Cr, Mn, Co, Ni, As and Sb. The zinc concentra-

tion in the litharge from Stojnik is unexpected. During cupellation, zinc concentra-

tions should have been decreased. Zinc concentrations in litharge from Sardis

were quite low, under 0.02% (Middleton et al. 2000: 160). Litharge, however, can

be quite variable in composition reflecting characteristics of the ore and lead

metal (Tylecote 1987: 139); one example of litharge with 2.01% Zn was reported

from Nord Eifel. Zinc was not sought in any of the other reported analyses. Alter-

natively, litharge samples from Stojnik were recovered from excavations of slag

dumps. The composition of porous litharge might be affected from the burial con-

ditions, so further investigation needed.

Litharge can also be unexpectedly scarce from excavations of major silver

production sites. The melting temperatures for PbO and Ag are 880oC and 960oC,

respectively. No cupels or fragments of cupellation hearths have been identified

yet among the excavated slag dumps. No litharge was found in the outlying test

pits excavated in 1985. The presence of several specimens of litharge with the slag

from the settlement does not necessarily indicate that cupellation took place

alongside the smelting outside the Roman fort. Litharge with the slag and furnace

remains could suggest that some was recycled back into smelting. It seems more

likely, however, that cupellation may have been undertaken under greater admin-

istrative control, possibly even within the fort or perhaps at another location

within the district. It would seem less likely that Singidunum or Aureus Mons may

have played some technical role in transport or intensive recovery of silver from

lead produced in the Mt. Kosmaj mining district. Litharge was not sought as a

metallurgical by-product in the earlier excavations of 1911–1913 within the Ro-

man fort. Often, litharge looks simply like another piece of white, corroded metal-

lic lead. However, litharge is brittle and easily broken to examine the interior. The

presence of litharge serves to document cupellation as a production step, but a

convincing location has not yet been identified near Stojnik.

Nine specimens of corroded lead metal were selected for compositional

analysis from Test Pits 2, 5, 11 and 13 and from the settlement Trial Trenches 2

and 3. Most lead metal samples were interpreted as ’scrap’ because there were

no clearly distinguishing features for corroded ’spills’ and fragments of sheet

and rod-like forms. Samples of lead metal were collected from under the cor-

rosion layers and analysed by the author using AAS to measure concentrations

of Pb, Ag, Fe, Cu, Sn, Zn, As, Ni, Au, Mn, Co, Cr, Mg and Sb. The concentra-

tions of Au, Mn, Co, Cr, Mg and Sb were below detection limits and not re-

ported in the tables. Detection limits were approximately 0.005 %. Undetected

elements are marked 0.00 in the table.

For AAS, the detection limits vary by element, but for silver the approxi-

mate value was about 0.005% for the selected sample size and dilution. The

precision also varies by element, for example, the precision for Ag is much

better than that for As. A more complete discussion is provided by Hughes,

Cowell and Craddock (1976). An accuracy value of ±1% is cited for major con-

55


centrations, but a value of ±5–15% is cited for minor and trace concentrations

(loc. cit.). Standard reference materials were used to monitor element calibra-

tions, detection limits and precision. Silver was measured on separate dissolu-

tions of 10 mg lead samples using 1:1 nitric acid diluted to 25 ml. The method

for analysis of lead published by Hughes, Cowell and Craddock (1976) system-

atically underestimates the silver concentrations in lead metal when there is

any corrosion present with the sample. Rehren and Prange (1998: 186) publish

another dissolution method to measure low silver concentrations in lead using

Inductively Coupled Plasma spectroscopy (ICP-OES). AAS data (table 5) for

lead metal suggest two groups with low ranges of silver compositions:

0.005–0.02% and 0.03–0.07%. The average for silver concentrations is 0.022%

Ag for the nine samples of lead metal from slag dumps in the settlement and

test trenches.

Table 5. Lead metal ’scrap’ was not identifiable as more than uncharacteristic ’spills’ of

metal as well as corroded fragments of sheet and rods mixed in the slag dumps from the

test pits and settlement. These selected examples were discovered in the trial trenches

from the settlement as well as more outlying test pits. Other elements were not sought.

Based upon these AAS data by the author, such low concentrations of silver in the lead

metal would suggest that production was inconsistent from the smelted lead ores.

SAMPLE Pb Ag Fe Cu Sn Zn As Ni

Test Pit 13, 80–100 cm 95.5 0.009 0.00 0.064 0.04 0.00 0.00 0.009

Test Pit 2, 50–80 cm 95.6 0.005 0.00 0.053 0.00 0.00 0.00 0.005

Test Pit 5, layer 3 74.2 0.069 0.93 0.068 0.00 0.057 0.63 0.069

Test Pit 11, 50–80 cm 69.5 0.064 0.14 0.024 0.00 0.00 0.04 0.00

Test Pit 2, 50–80 cm 73.1 0.031 0.08 0.064 0.06 0.014 0.59 0.01

Test Pit 13, 80–100 cm 95.6 0.005 0.00 0.032 0.00 0.00 0.00 0.005

Test Pit 13, 60–80 cm 94.5 0.000 0.00 0.044 0.00 0.00 0.00 0.00

Trial Trench 2, St.6.83 96.9 0.008 0.00 0.043 0.00 0.00 0.00 0.00

Trial Trench 3, St.12.83 92.5 0.005 0.00 0.036 0.00 0.00 0.00 0.00

To assess the lowest levels of silver in lead metal, presumably after

cupellation, sixteen more samples of lead metal were analyzed by the author using

AAS (tables 5–7) from other nearby Roman necropolis sites at Guberavac,

Grobnica and Gomilica near the village of Stojnik. Lead coffins with simple cast

decoration have been found in the cemeteries. The necropolis at Guberavac, for

example, consists of multi-chambered mausolea along a 4–5 km stretch of a mod-

ern road apparently following earlier pathways. The tombs date to 2nd and 3rd cen-

turies (Werner 1990). Certainly, it could be assumed that the lead for production

of coffins for local use was produced from the local mines and smelting furnaces.

Furthermore, it could be assumed that lead metal from these sites would most

likely represent lead from which silver had been removed by cupellation. Analysis


56

of the lead and ’solder’ join from one coffin was undertaken to test the second as-

sumption. The observed compositions, under 0.005% Ag, for the two samples

from the lead coffin (table 7) may serve as a baseline for desilvered lead used in

quantity during Imperial Roman times. It was technically possible to desilver lead

by cupellation to a very low value. The sample of the lead “solder” along one of

the edge joins had a similar composition, without any addition of tin. It is signifi-

cant to note that sixteen lead metal samples (tables 6–7) from the necropolis near

Guberevac have an average concentration of 0.010% Ag. The lowest concentra-

tions are near the detection limits of 0.001% for AAS given the selected sample

sizes and dilutions. The average for silver in the lead metal from the necropolis

sites in for tables 6–7 is not much different from the average for table 5, given

the standard deviation. The difference, between the respective averages, would

represent a recovery value of about 100 g of silver per ton of lead metal, with a

comparable amount of the silver still lost in the lead.

Table 6. Lead metal samples selected from the necropolis (Grobnica 2) near Gube-

revac and analysed with AAS by the author. The first sample in this table is interpreted

as an example a crucible spill because of the low silver concentration, weight and finds

location away from primary lead smelting. All other samples were unidentifiable lead

metal ’scrap’ from the excavations, perhaps fragments from damaged objects.


Grobnica 2, � 60, C255 95.1 0.012 0.005 0.045 0.00 0.00 0.00 0.005

Grobnica 2, C185 95.3 0.000 0.00 0.043 0.00 0.00 0.00 0.00

Grobnica 2, Quad 2 96.1 0.000 0.005 0.037 0.00 0.00 0.00 0.00

Grobnica 2, � 62, C252 96.5 0.026 0.006 0.074 0.00 0.00 0.00 0.00

Grobnica 2, C186 96.2 0.016 0.00 0.036 0.00 0.05 0.05 0.00

Grobnica 2, � 47, C229 97.5 0.005 0.005 0.043 0.03 0.00 0.00 0.00

Grobnica 2, C183 94.6 0.017 0.00 0.038 0.00 0.00 0.03 0.00

Grobnica 2, C202 97.6 0.009 0.005 0.039 0.00 0.005 0.14 0.00

Grobnica 2, � 196, C262 95.7 0.008 0.005 0.037 0.00 0.00 0.00 0.00

Grobnica 2, � 196, C269 93.7 0.007 0.005 0.052 0.06 0.00 0.00 0.005

Grobnica 2, � 44, C209 94.0 0.018 0.005 0.048 0.05 0.00 0.00 0.006

Grobnica 2, � 60, C255 96.6 0.010 0.005 0.040 0.04 0.00 0.00 0.00

Grobnica 2, � 48, C228 96.4 0.014 0.00 0.054 0.00 0.005 0.00 0.00

Grobnica 2, � 22, C150 95.4 0.016 0.00 0.029 0.00 0.00 0.00 0.00

A most unusual example of ’tapped’ lead or probably a crucible spill was

discovered from nearby Grobnica 2, Ä 60, C 255. This specimen was not found

in a lead smelting context, but it could have been recovered and moved to ne-

cropolis for further processing. The weight is about 2 kg. A sample was re-

moved and analyzed by the author with AAS. The analysis measured only

57


0.012 % Ag and detected no tin (table 6). Based upon compositional data,

weight and find spot in the necropolis (away from primary smelting), it would

seem more likely this specimen would represent desilvered lead possibly asso-

ciated with coffin manufacture, not primary lead from smelting. For example,

lead metal (without tin) of the same composition as the coffin had been used

to ’solder’ the edge joins. This type of ’soldering’ using the same metal compo-

sitions would probably have used several kilograms of molten lead poured

from a crucible along the edges to be joined.

Table 7. A sample of ’solder’ from along the join to the lower right was taken for

compositional comparisons and specifically to check for any tin in the solder composi-

tion using AAS by the author. The silver concentrations are some of the lowest inter-

preted as representing deliberately refined lead produced by converting litharge back

to lead metal.


Gomilica, lead coffin 96.2 0.005 0.000 0.035 0.00 0.005 0.00 0.005

Solder from coffin 95.0 0.002 0.007 0.035 0.00 0.005 0.00 0.005

In the present investigation, no new specimens of silver-rich lead ore were

found. No “lead pigs” have been analyzed during this current programme of in-

vestigation, but some fourteen are known from the Mt. Kosmaj region. The

Imperial Roman “lead pigs” noted by Davies (1935: 214) and discussed by

Du{ani} (1995a: 221) have not yet had compositional analysis to determine the

silver concentrations. Silver alloy jewellery has been excavated from burials in

the Guberavac cemetery, but examples have not been analysed. No silver in-

gots have been analysed in the present study. If the investigation were to have

been continued with archaeometallurgical fieldwork beyond 1986, then addi-

tional samples of such specimens and objects would have been actively sought

for analysis. Partitioning and recovery of various element concentrations has

been investigated in cupellation experiments by Pernicka and Bachmann

(1983). “Pure” silver metal from cupellation would be expected to have been

alloyed with variable concentrations of copper metal to produce the composi-

tions of alloys used for Imperial Roman jewellery, plate and coins.

DISCUSSION

The emphasis of the present investigation of Imperial Roman metallurgi-

cal activities near Stojnik has been upon technical evidence for smelting,

cupellation and lead metal production. The observed compositions and miner-

alogy of the Imperial Roman lead smelting slag near Stojnik are unexpectedly


58

consistent (tables 1 and 2). Controlled weights of iron ore as a flux would need

to be charged with the lead ore to produce the observed slag compositions. In

part, however, some of the uniformity is due to the representative sampling of

tapped slag. Molten tapped slag that readily flows out of the furnace is charac-

teristically of a specific compositional range. There is a slight trend toward us-

ing higher concentrations of iron in the tapped smelting slag for the later peri-

ods, but essentially the slag compositions do not change over some 200 years.

The presence of fayalite (Fe2SiO4), and not pyroxene (CaFeSi2O6), is also sig-

nificant and due to the concentration of CaO in the range of about 5–8%. This

concentration could be attributed to calcium carbonate mixed with the ce-

russite and galena as well as calcium oxide from the fuel ash and a smaller pro-

portion from the furnace wall. While it is noted that there is some limited min-

eralogical association between available iron oxides (limonite) and reported

argentiferous cerussite with galena, it is concluded that a higher-grade iron ore

was added separately as a flux for smelting to consistently produce such uni-

form tapped fayalite-type slag compositions. It would also be reasonable to as-

sume liquid lead metal was also tapped from the furnace, perhaps at intervals.

The molten lead would readily separate in a smelting furnace under molten

fayalite slag with an approximate specific gravity of 3–5 g/cm3 (Bachmann

1982a: 4). The free-running temperature for the fayalite slag compositions

would be approximately 1200oC; characteristic for smelting in a small furnace

capable of producing tapped slag of some 20 kg using charcoal as fuel. Slag left

behind as accretions on tuyeres or adhering to the furnace walls is not as repre-

sentative of target operating conditions as the tapped slag. The accretions have

higher concentrations Pb and Ag (table 3). Samples of slag adhering to furnace

wall would incorporate higher proportions of silica and alumina. Selection of

such variable furnace slag samples to characterize the operating process would

be misleading.

It has been noted that fayalite-type slag was also used at Rio Tinto for sil-

ver production from argentiferous jarosite ores (Rothenberg and Blanco-

-Freijeiro 1981: 312; Craddock et al. 1985: 205), but with several variations,

such as the addition of lead to improve recovery of precious metals from the

jarosite ore and possibly the iron-rich gossan used as flux for smelting.

Rothenberg and Blanco-Friejeiro (1981: 174) argue for state administration

and technical management for large scale production at Rio Tinto, but dating

such huge quantities of slag to the various major periods of production seems

somewhat tentative due to the complexity of the evidence. Greene (1990: 147)

does not agree with argument of scale at Rio Tinto and argues for a degree of

lease-holder activity. Rio Tinto has evidence for Republican and Imperial Ro-

man production as well as other earlier and later periods. Therefore, develop-

ments in technological detail are more difficult to resolve by date at Rio Tinto.

For their notes on the interpretation of compositional data for fayalite

(Fe2SiO4) slag samples from Rio Tinto, Rothenberg and Blanco-Freijeiro

59


(1981: 312) agree that “All slags are essentially 2Fe(Mn)O SiO2”. Craddock et

al. (1985: 202–205) observed that fayalite was the present in all samples of slag

from Rio Tinto Site 19A. However, the mineralogy was more complex, with

other mineralogical phases also present due to smelting jarosite ore. Funda-

mentally, it remains clear that fayalite-type slag compositions were very useful,

indeed, and prevalent at Rio Tinto during the Roman periods. Iron-rich slags,

principally fayalite-type slag compositions, were apparently well-understood

and readily adapted for smelting various types of ore. It is significant that faya-

lite-type slag is characteristic also for copper smelting and iron smelting known

from other ancient sites.

Lead smelting experiments by Hetherington (1980: 38–39) also used

iron-rich slag compositions, but operated on a small scale with total weights of

about 5 kg total charges of approximately 4 kg galena concentrate with 78%

PbS to 1 kg iron ore flux. The measured slag weights were about 3–4 kg with

high PbO losses in the slag of about 15–20%. Very little lead metal was recov-

ered from most smelting experiments; complications can be attributed to poor

control of iron-rich slag compositions. CaO concentrations were not deter-

mined for the lead ore, iron ore flux or smelting slag, so the slag mineralogy

based upon reported concentrations is somewhat uncertain. It seems, there-

fore, based upon the table of variable iron concentrations and related compli-

cations for tapping, that fayalite-type slag compositions were not produced

consistently in the experiments. Based upon reported iron ore flux charges,

however, fayalite slag compositions remained an unachieved objective for the

research. With copper smelting experiments by Merkel (1990), consistent pro-

duction of fayalite slag required additional charges of iron oxide as flux to

compensate for silica contributions from the fuel ash and furnace wall to slag

compositions. In the slag, CaO was derived principally from the iron ore flux.

The quality of the flux, including CaO concentration, has a significant affect on

the slag compositions. The slag-type mineralogy produced in the lead smelting

experiments by Hetherington (1980) was probably pyroxene (CaFeSi2O6)

based upon published viscosity descriptions at reported temperatures, rela-

tively high PbO losses and lack of CaO analyses in the experimental simula-

tions.

Nevertheless, using some of these experimental data from Hetherington

(1980) to assess the Imperial Roman lead smelting slag compositions from Mt.

Kosmaj, several estimates of scale can be proposed. With a simple mass bal-

ance of FeO and depending on estimates of charcoal fuel quantities and fur-

nace refractory qualities, the charged cerussite ore alone would not be suffi-

cient to produce such consistent and iron-rich fayalite slag compositions. Addi-

tional iron ore charges as a flux would have been required, as in the experi-

ments reported by Merkel (1990). As an estimated range for the available li-

monite ores would normally be expected to have 25–30% Fe. An observed

weight of 20 kg of tapped fayalite-type slag with 25% FeO would represent an


60

original charge of about 5 kg of FeO along with the charge of cerussite, which

would require a minimum of about 20 kg of limonite. Unfortunately, the corre-

sponding weight of lead produced from a smelting furnace with an estimated

volume of 5-8 litres is difficult to model without using the assumed experimen-

tal ratio of 4 kg galena concentrate to 1 kg iron ore flux (Hetherington 1980:

36–37). A total weight of 80 kg cerussite (PbCO3) would require some 20 kg of

limonite to still produce some 20 kg of tapped slag as a waste product. A

high-grade lead ore concentrate could contain as much as about 70% Pb, but

losses in the slag of about 6-7% Pb (Wray 1921: 66) would account for about

1.5 kg lead in 20 kg tapped slag on a mass balance estimate. Thus, an estimate

of lead metal produced from single smelting furnace could be 50 kg, but proba-

bly lower. Based on the specific gravity of lead, each litre of lead metal in the

furnace would weigh about 11 kg. A lower estimate of lead production based

upon of total capacity in the furnace would be approximately 10 litres of mol-

ten slag and lead metal, distributed by weight between some 20 kg fayalite type

smelting slag with perhaps 40 kg lead metal. A more precise scale model can-

not be proposed without better preserved furnace remains. However, the lower

estimate may provide a useful guide.

For comparison, much earlier lead smelting slag samples from 5th–4th cen-

tury B.C. silver production at Laurion, also indicate routine use of iron-rich

slag compositions with an important difference. The mineralogy of lead smelt-

ing slag from Laurion was consistently pyroxene (CaFeSi2O6), as reported by

Bachmann (1982b: 249), with only secondary fayalite (Fe2SiO4). The slag com-

positions from Laurion were lower in FeO and higher in CaO than the Impe-

rial Roman slag compositions from the near Stojnik. For Laurion, the reported

range for CaO was 10–42% with an average of about 20% (loc. cit.). The aver-

ages for iron oxide and total lead were about 15%. Rehren, Vanhove and

Mussche (2002: 44) interpreted the data to indicate that the lead was present

as an entrapped compound, not metallic lead. The higher viscosity of the

pyroxene-type slag at temperatures of 1200oC (Bachmann 1980: 130) may be

responsible for the higher lead losses in the smelting slag from Laurion. From

analysis of nineteen slag samples from Laurion, Bachmann (1982b: 248–249)

reported seventeen samples with under 0.005% Ag and only two other samples

with 0.05% and 0.59% Ag. For comparison to Rio Tinto, Rothenberg and

Blanco-Freijeiro (1981: 312) report compositional data for one hundred and

twenty slag samples with ninety-four silver values under 0.005% Ag. From the

remaining analyses, only seven samples represent exceptions with higher con-

centrations from 0.06% Ag up to 0.27% Ag. Frequency plots of sample num-

bers by published silver concentrations reveal the skewed distribution toward

the lowest values. The smelting experiments by Hetherington (1980) seem

more representative for slag compositions reported from Laurion, not Rio

Tinto or Stojnik.

61


The concentrations of lead “lost” in the slag from Stojnik, generally be-

tween 6% and 7% (tables 1–2), represent a more consistent and efficient pro-

cess using fayalite-type slag compositions. From the metallographic sections of

the slag used for microanalysis, much of the lead is lost as lead sulphide. Faya-

lite-type slag compositions and pyroxene-type slag compositions were investi-

gated in detail for copper smelting experiments by Merkel (1990). The faya-

lite-type slag was observed to be more free-running for tapping and provided

better liquation within the furnace of the copper metal, as indicated by the

data published by Bachmann (1980: 130). Molten and tapped fayalite slag com-

positions were visibly less viscous than pyroxene slag compositions at the tem-

peratures achieved in the charcoal fuelled smelting furnaces. Use of iron-rich,

fayalite-type slag compositions has similar benefits for lead smelting; concen-

trations between 5–8% CaO represent a narrow range in the fayalite tapped

slag from Stojnik. Production and control of fayalite-type slag compositions are

interpreted as deliberate at Stojnik. However, deliberate control of CaO to

avoid pyroxene in the slag represents another level of technical difficulty. Un-

expectedly, the CaO values are observed to have been consistent at Stojnik

suggesting additional control of other factors contributing CaO to the slag,

such as fuel consumption as well as ore and flux selections. Concentrations of

zinc (reported as ZnO) in the smelting slag from near Stojnik can also be of

use for inferring smelting parameters, such as ore grade and operating condi-

tions in the smelting furnaces. Relatively consistent values of 4–6% ZnO in the

smelting slag (tables 1–2) suggest an association of lead and zinc in the ore

charges as does the observation of galena and zinc-iron sulphides in the slag by

Craddock (1985). In lead ore deposits of Serbia, the presence of mixed miner-

alization containing lead and zinc together is quite common (Jankovi} 1967;

Jankovi} and Petkovi} 1980). These few percent ZnO, however, are not

enough fundamentally to alter the fayalite-type slag compositions and proper-

ties. Wray (1921: 66) and Davies (1935: 214) noted the presence of lead

sulphides with zinc sulphides in the ores mined near Kosmaj. Lead-zinc miner-

alization is present as hydrothermal deposits at Rudnik (Radi} and Vukovi}

1957: 130). It is not unexpected to detect several weight percent of zinc in the

lead smelting slag. Under reducing conditions in the copper smelting furnace

using fayalite-type slags in copper smelting experiments, based upon mass bal-

ance calculations, about 70% of the charged zinc was lost as fume while about

25% partitioned into the slag (Merkel 1990: 90, 108). Such a model would sug-

gest much higher zinc inputs with the lead ore and iron ore flux charges to pro-

duce consistent values of 4–6% ZnO in the slag from Stojnik. However, zinc

concentrations are characteristically low in the lead metal after smelting and

cupellation.

Sulphur is not included in the slag compositional data reported in tables

1–2. Using AAS analysis, sulphur cannot be successfully measured. It repre-

sents one of several disadvantages for AAS of ancient slag samples. As a check,


62

however, XRF analysis (table 3) can measure sulphur concentrations and also

detect several significant elements not routinely sought with AAS. The slag ac-

cretion from Trial Trench 7 included approximately 4% sulphur. Based upon

the bulk slag compositions, the ore charges would seem to have included a

mixed assemblage of sulphide minerals. Barium was not detected with XRF.

Documentation of the Roman mines near Stojnik (Tomovi} 1995) needs fur-

ther investigation of the mineralogy, but clearly galena (PbS) and possibly

sphalerite (ZnS) were constituents of the ore charges based upon the common

presence of lead sulphide and zinc-iron sulphide entrapped in the fayalite-type

smelting slag. Proportions of cerrusite (PbCO3) and other possibly associated

iron carbonates (FeCO3) and calcium carbonate (CaCO3) in the ore charges

remain uncertain.

The compositional data for zinc, barium and sulphur in the slag from

Stojnik, also make interesting contrasts to results for the pyroxene-type slag

composition from Laurion (Conophagos 1980: 284; Bachmann 1982b: 248–249)

and the fayalite slag compositions from Rio Tinto (Rothenberg and Blanco-

-Freijeiro 1981: 314–317; Craddock et al. 1985: 205). The Laurion slags have a

zinc concentration range of 2–10% with consistently low values of sulphur. BaO

was not detected in the Laurion slag. From Rio Tinto, sulphur ranges in the slag

analyses from about 0.1–4.0% (Rothenberg and Blanco-Freijeiro 1981: 314–317).

These data for sulphur were determined on bulk samples using a combustion

and titration technique (Matterson, Peers and Shettle 1981: 299–302). From Rio

Tinto, zinc concentrations are all under 2% with most published values much

lower. BaO is much higher, with variable concentrations up to about 11% BaO

reported in the Rio Tinto slag due to the complex jarosite ore (Craddock et al.

1985: 204–205) or additional fluxing with barytes (BaSO4).

From total concentrations of 6–7% Pb remaining in the smelting slag

(Wray 1921: 66) and tables 1–2, lead sulphides observed microscopically to be

entrapped in the slag are evidence of galena as a constituent in the ore

charges. The presence of lead sulphide in the smelting slag is interpreted as

unreacted galena from the ore charges. Lead sulphide is not readily converted

directly to metal under reducing conditions. The melting temperature for ga-

lena is 1115oC. If galena in the smelting charge is first not adequately oxidized

to PbO in the upper zone of the smelting furnace, then unreacted galena will

simply melt and in part become entrapped in the fayalite-type slag. The silver

concentration in the charged galena would not be expected to decrease simply

from melting. Metallic lead prills in the slag, analysed with EPMA for preci-

sion and accuracy, had similar silver concentration to the lead sulphide (be-

tween 0.01 and 0.06% Ag) in twenty-five selected samples (Mizota 1995).

Microanalysis would be expected to provide a better measure of silver concen-

trations and distributions in lead smelting slag by detecting partitioning of high

silver concentrations between mineralogical phases and lead metal prills, than

bulk compositional analysis of gram size samples of slag. The documentation

63


of numerous inclusions of lead sulphide suggests that the ore charges were not

dead-roasted first as a preliminary processing step to convert the galena to

lead oxide. Some approximation of higher grade ores was expected from

microanalysis of lead sulphide inclusions, representing galena, or silver-rich

metallic lead prills in the smelting slag. However, the observed microanalysis

values between 0.01% and 0.06% Ag indicate that low concentrations of silver

were more common in the galena component of the lead ore grades actually

being processed in the smelting furnaces near the Imperial Roman fort and

settlement.

A wide range of lead ores from the many different mines would need to

be assayed first by Roman prospectors to identify ore grades suitable for smelt-

ing and silver recovery. Of course, there could be relatively great differences

between silver concentrations in the mined areas from Mt. Kosmaj and Avala

as well as possibly Rudnik during the Imperial Roman period. Wray (1921:

65–66) noted in his geological report that the modern Crveni Breg workings at

Avala produced galena with a range of 0.183% to 0.4308% Ag. At Kosmaj,

Wray (1921) noted that lead and zinc sulphides (PbS-ZnS) were associated in

the ore mined. However, Davies (1935: 214) claimed argentiferous cerussite

(PbCO3) with some galena (PbS) was mined by the Romans at Kosmaj.

Jankovi} (1967) reported a value of 0.33% Ag for galena concentrate which is

interpreted as an average value for Rudnik based upon modern standards for

crushing and beneficiation. Tomovi} (1995: 208) cited an exceptionally rich ore

in the district with 0.6% Ag. The problem with these published data concerns

variation. Should an average, range or highest value be reported? What con-

centrations of silver in the lead ore are most representative? Without new sam-

ples of silver-rich ores from the mines or smelting sites near Stojnik, there are

aspects of silver production, such as ore grades, that are difficult to verify with

much precision. A more complete report on the mineralogy of the lead-zinc

deposit at Rudnik is published by Radi} and Vukovi} (1957) with further ex-

amples of hydrothermal lead-zinc deposits in the [umadija District outlined by

Jankovi} (1967), Jankovi} and Petkovi} (1980) and Grèti} and Jelenkovi}

(1995). However, citation of exceptionally high silver concentrations seems bi-

ased and modern average concentrate values seem misleading. Thus, the re-

ported range of values from Avala by Wray (1921: 66) is considered most rep-

resentative to gain an impression of ancient selection of available ore grades

for smelting. Further ore samples from the mines and smelting sites would be

required to further resolve this fundamental issue of silver concentrations in

the lead ores accessible to Roman miners. Descriptive statistics of compo-

sitional data would be worthwhile. The lead ore would be improved by

hand-sorting or washing out unwanted constituents to concentrate the heavier

argentiferous cerussite and galena. However, additional minor constituents in

a mixed lead-rich ore, such as pyrite (FeS2) and sphalerite (ZnS), would not be

readily removed in beneficiation. Such minor constituents in a beneficiated


64

mixed lead-rich ore would partition into the slag, as observed unreacted

sulphides. Beneficiation represents a critical step for preparation of lead ore

charges for smelting. Many small rivers and streams cross the mining district,

but there is no evidence yet for Imperial Roman lead ore concentrates near

Stojnik.

For comparison, a range of ore concentrations and mineralogical constitu-

ents are published for smelting at Laurion. Pernicka (1981: 398) reported sev-

eral high values between 0.4% and 0.5% Ag for argentiferous lead ore from

Laurion, but most values were lower. Bachmann (1982b: 250) proposed an av-

erage value of 0.1% Ag for the galena smelted at Laruion. Rehren, Vanhove

and Mussche (2002: 36) report normalised values to calculate an average silver

value in ore samples from Thorikos, near Laurion, of 0.095% Ag. Their highest

reported value was 0.156% Ag. Bachmann (1982b: 248–250) concluded that

galena, not cerussite, was smelted at Laurion. Photos-Jones and Jones (1994:

357) claim cerussite was the primary ore mined at Laurion. Along with galena

as a constituent of the primary ore, Rehren, Vanhove and Mussche (2002:

45–46) have identified and documented a second type of lead-rich fluorspar

ore, also richer in silver, from samples of the ore beneficiation sites at

Lavriotiki. Cerussite was interpreted as a weathering product in the samples.

For further comparisons, the principal silver-rich ore at Rio Tinto was jarosite,

which necessitated the addition of lead into the smelting charge to collect and

concentrate the silver (Craddock et al. 1985: 210). Craddock et al. (1985: 207)

suggested an economic lower limit of approximately 0.02% Ag in the jarosite

ore from Rio Tinto, but note high values up to 0.31% Ag. These data from

Laurion and Rio Tinto give some indication of economic values for argentifer-

ous lead ore grades.

Excavated lead metal samples from near Stojnik and the nearby necropo-

lis at Guberevac can provide another perspective on technical aspects of lead

production and silver recovery by cupellation. Using the observed upper range

for silver concentrations (0.03–0.07% Ag) in the lead metal from selected sam-

ples from excavations of the settlement and outlying test pits (table 5), it could

be estimated that maximum production of 50 kg of primary lead from the

smelting furnace would correspond to only some 15–35 g silver with total re-

covery from cupellation. At these low values, however, cupellation of one kilo-

gram of lead metal could have produced less than 1 g silver. No examples of

silver-rich lead “bullion” have been identified from the site, but such compari-

sons help emphasize aspects of human labour and fuel consumption to pro-

duce silver under these limitations. Nevertheless, higher recovered silver

weights should be expected from the silver-rich lead produced from smelting

the best quality argentiferous lead ore.

Analyses of lead metal from the excavations in the present study are hypo-

thetically interpreted to show that all of the lead metal has been subjected to

cupellation. There are few concentrations of silver in the lead metal from the

65


present investigation which would exceed a threshold value of 0.06% for eco-

nomic recovery proposed by Tylecote (1986: 61). Secondary evidence of

cupellation would be the presence of larger quantities of litharge from the ex-

cavations. In litharge, the lead oxide has very low silver concentrations. If all of

the lead metal from smelting were routinely processed to recover the silver,

then it is reasonable to conclude that practically all of the lead metal converted

to litharge was also recycled and smelted back to produce lead metal; large

quantities of lead metal from the excavations have very low silver concentra-

tions. Such a conclusion has implications for scale of recycling litharge back

into smelting furnaces as well as fuel and labour requirements. Litharge would

represent a readily available source to produce lead metal for other uses. The

lack of high silver concentrations in all twenty-five samples of lead metal from

the settlement, test pits and necropolis, suggests that huge quantities of

litharge were apparently being recycled back into the smelting furnaces to pro-

duce lead metal. Apparently, organization for cupellation and recovery of sil-

ver from the lead metal was very thorough, possibly under very close adminis-

trative control, consuming additional quantities of fuel, labour and other re-

sources. Few examples of litharge, perhaps even just kilograms, have been re-

ported from Laurion (Rehren et al. 1999: 302) and Rio Tinto (Craddock et al.

1985: 208). Recycling litharge back into the smelting furnace, litharge dumps at

other locations as well as medical uses have been suggested to explain the rela-

tively small quantities of litharge found at ancient lead and silver production

sites. Metallic lead was still a valued product, even with the silver removed.

The silver values in the lead metal analysed in this investigation of the

archaeometallurgical remains from Stojnik are lower than might have been ex-

pected, but not that different compared to analytical data published from other

sites such as Laurion (Conophagos 1980: 284) and Rio Tinto (Rothenberg and

Blanco-Freijeiro 1981: 314–317; Craddock et al. 1985: 205, 208). A few exam-

ples are relatively high, but most slag analyses reveal unexpectedly low silver

concentrations. Lead metal from Laurion had observed values between 0.015

and 0.019% Ag (Conophagos 1980: 331–332). Examples of lead metal from

Rio Tinto, interpreted as refined lead, were reported with values between

0.007 and 0.03% Ag (Craddock et al. 1985: 208). Large quantities of Roman

slag at Rio Tinto have been systematically reworked by the modern companies.

Interpretation of these low silver concentrations is not straightforward be-

cause the lowest values can represent technological capabilities, such as noted by

Healy (1978: 180; idem 1999: 324) for the ’ability’ to desilver lead down to

0.01% Ag. Thus, the capability existed, but apparently was not always used to re-

cover as much silver as technologically possible from the lead metal. Cupellation

of silver-rich lead produced from smelting was probably done just once for a

given batch. The recovery and loss of silver was somewhat variable, but repeti-

tion of cupellation consistently to recover the lowest levels of silver from lead

metal would also skew a frequency distribution more toward the lowest values. It


66

can be concluded that cupellation was used thoroughly on all silver-rich lead be-

cause no high concentrations were observed in any lead metal samples. No sil-

ver-rich lead or “bullion” seems to have escaped cupellation under these specific

conditions of Imperial Roman production and administration.

The more difficult threshold value to estimate is the lowest “economic”

concentration of silver in lead metal produced from smelting that could be re-

covered profitably under the political, military and economic conditions of Up-

per Moesia during the later 2nd century A.D. In other words, under what con-

ditions would silver recovery be profitable from any lead metal? Such an esti-

mate would reflect issues of political, military, administrative control of labour

and other resources. For example, transportation of people between provinces

for compulsory work in the mines of interior Upper Moesia (Du{ani} 1977a:

93) suggests economic development may have been more politically important

than the actual concentrations of silver in the ore and lead metal. However, it

is difficult to encompass such issues within a single numeric estimate for silver

concentrations. Nevertheless, a value above 0.06% Ag was reported by Tyle-

cote (1962: 82; idem 1986: 61; idem 1987: 140) in lead metal for the Roman pe-

riod. However, at one point, Tylecote (1976: 61) published a lower estimate of

above 0.01% Ag for the Roman period. Healy (1978: 180; idem 1999: 324) also

uses this value; Rehren and Prange (1998: 189) and Rehren, Vanhove and

Mussche (2002: 38) also use a threshold value of 0.01–0.015% Ag for desil-

vered lead metal. These different estimates need to be assessed within a spe-

cific context. Concentrations above a value of 0.06% Ag in lead metal would

certainly seem “profitable” under given conditions, while lower values at vari-

ous times would not have been economic for recovery. However, it is not nec-

essarily correct to infer that values under such a concentration of 0.06% Ag

have all been desilvered or refined, unless it is clear all of the lead ore from a

specific mined deposit had been sufficiently silver-rich. All three mining areas

within the district, Avala, Mt. Kosmaj and Rudnik, are clearly identified in

modern geological surveys as having produced argentiferous lead ores as illus-

trated by Du{ani} (1977a: 55). Under specific conditions of Imperial Roman

administration of precious metal resources in Upper Moesia and other prov-

inces, the demand for silver may have driven the value for economic produc-

tion of silver even lower, perhaps toward 0.01% Ag. The organization of the

district as a territorium metalli (Mücsy 1974: 63, 195; Du{ani} 1977a: 79–93) in-

dicates that the mines were initially the property of the Emperor. Administra-

tion began, in the case of Mt. Kosmaj, presumably with the establishment of

the auxiliary military fort near Stojnik in the centre of the district, to provide

better security for silver production and transport. “Economic” values change,

but only as a result of changes in other political, social and external conditions.

Various factors, such as the Marcomannic Wars, theft, forced transport of la-

bour into the mining district, extensive use of slave labour, possible municipali-

sation of the administration, limited private production and more, would have

67


affected the “economic” value for recovery of silver, pushing a threshold value

for silver recovery even lower still. Ultimately, political and military consider-

ations seem to have prevailed in the need to increase silver production from

the mid-2nd century in Upper Moesia.

Furthermore, based upon the compositional data for modern slag from

near the ruined 20th century industrial buildings along the Pruten river valley, it

would seem that modern reworking of the Roman slag also would not have

been very profitable, despite other claims (Wray 1921: 66). Using the highest

silver concentration for Roman slag from the table 1 of 0.018% or 180 g/ton,

and the lowest silver concentration for the modern slag from table 4 of 0.001%

or 10 g/ton, the corresponding maximum estimate would be approximately 170

g silver recovered from remelting each ton of Roman lead smelting slag. The

average assay values reported for the Roman slag by Davies (1935: 213) would

indicate higher levels of lead recovery, but even lower silver levels at only

0.0037% Ag. Based upon the industrial data for modern reworking in 1907–1908

of the Roman slag, reported by Tomovi} (1995: 205), from 42,245 tons of Ro-

man slag processed, the recovery was 1767 tons of lead and only 0.488 tons of

silver. For comparison in weight percent to the new compositional data for the

slag and lead metal, these industrial data (loc. cit.) would represent a value of

4% lead and only 0.001% Ag in the Roman slag. The mass of 0.488 tons silver

divided by 1767 tons lead represents a value of about 0.025% Ag in the lead

metal. These value estimates are lower than expected for modern recovery, but

these data also represent silver “lost” in Imperial Roman processing. Several

lines of investigation now suggest the quality and quantity of argentiferous

cerussite with some galena were highly variable. Microanalytical evidence from

lead sulphide inclusions and lead prills in the Imperial Roman slag does not

support the consistent use of exceptionally rich argentiferous lead ores.

Technical studies of the production remains are essential to interpret

archaeometallurgical sites. For example, for Roman lead production in the

Mendip Hills in Somerset, England, Todd (1996: 15) concluded that evidence

for cupellation and silver recovery was negligible or inconclusive without tech-

nical studies. Investigations of the archaeometallurgical remains from Laurion,

Rio Tinto and Stojnik allow an increasingly technical assessment of ancient sil-

ver production which can provide new perspectives on other lines of historical

research.

CONCLUSIONS

The region around Stojnik, Guberevac and Babe, south of Belgrade in

Serbia, has not been recognized widely for its importance as a major Imperial

Roman centre for silver production from the lead ore deposits. Estimates of

Imperial Roman silver production are based on the reported 1,000,000 tons of


68

smelting slag and some 5000 mining shafts in the district. Further metallurgical

aspects of silver production concerning ore grade and smelting efficiency have

been based upon modern geological reports and reworking of the ancient slag.

Technical studies of the smelting slag, litharge and metallic lead from archaeo-

logical excavations near Stojnik were required for better comparison to other

ancient silver production sites at Laurion and Rio Tinto.

Based upon the consistent use of iron-rich, fayalite-type slag at Stojnik for

lead smelting (tables 1–2) and observed weights of tapped slag near 20 kg, it is

concluded that Imperial Roman lead production utilized a separate addition of

iron ore as flux. The use of fayalite slag required a melting temperature of

about 1200oC, so the lead smelting furnaces reliably operated at temperatures

achieved for Roman copper smelting and iron smelting. The technological key

to the operation was the utilisation of controlled additions iron ore flux to pro-

duce consistent slag compositions for lead smelting.

The weights of tapped slag blocks serve to estimate operational furnace

volumes, up to the level of the tuyere(s). Thus, the furnaces must have been

numerous based upon the observed distribution, depth and dating of smelting

slag dumps near Stojnik. As expected, Imperial Roman lead smelting repre-

sents an improved efficiency, with generally lower lead losses in the slag, over

earlier practice. Technical studies of the archaeometallurgical remains from

the site provide sufficient detail to begin addressing new questions about the

origins, development and diffusion of specific aspects and Imperial Roman

lead smelting and cupellation techniques. In some ways, the technology reflects

earlier practice, similar to earlier production at Roman Rio Tinto with the use

iron-rich fayalite smelting slag compositions. However, there are significant

differences, too, due to the complex silver-rich, lead-poor jarosite ores encoun-

tered at Rio Tinto. In the mid-2nd century, silver production centred on Stojnik

and Mt. Kosmaj in Upper Moesia was beginning to intensify during the reign

of Marcus Aurelius. Developed knowledge and utilization of iron-rich faya-

lite-type slag compositions were uniformly and successfully applied to new con-

ditions for smelting argentiferous cerussite and galena containing ores in Up-

per Moesia. Silver and lead production increased rapidly on an “industrial”

scale under an Imperial Roman administration with military security of the

mining district. With interdisciplinary research, perhaps, further understanding

of Imperial Roman technology will be achieved for their “political-military-

-industrial” complex. This research attempts to take a wider view of the diffu-

sion and innovation using specific technological detail for a specific metallurgi-

cal process and the exploitation of interior natural resources of Upper Moesia.

The Imperial Roman lead smelting technology centred at Stojnik in the

mining district used reliable fayalite-type slag compositions for lead and silver

production. The beginning and development of this technology took place else-

where. The smelting technology seems to have been introduced alongside the

initial military occupation in the mid-2nd century. The smelting technology rep-

69


resents a significant part of the rapid expansion and increasing production of

lead and silver. There seems to have been no development of the technology in

the mining district. Technical problems had already been overcome for the ar-

gentiferous cerrusite and galena ores encountered around Mt. Kosmaj. Indeed,

technical aspects of smelting furnace operation do not change over some 200

years. The dating evidence for the earliest slag from Levels 2–8 in Trial Trench

8 near the outer perimeter defensive wall of the Roman fort is critical to this

conclusion. The choice of technological approach using a fayalite-type slag was

decisive from the beginning of Imperial Roman lead and silver production

near Stojnik. It is concluded that the Imperial Roman “system” for develop-

ment of lead and silver production (Du{ani} 1977a: 93) also included techno-

logical aspects, such as a prescribed production “target” for smelting slag com-

positions with readily observable physical properties, such as slag viscosity and

practical, monitored weights of lead and silver recovered, for quality control of

the process steps. The desired compositional values were quickly implemented

and maintained, based upon the iron, calcium, lead and silver concentrations

of the slag (tables 1–2), with deliberate additions of iron ore as flux into the

smelting furnaces. Cupellation could also be controlled by observation based

upon the appearance and quantity of silver at the end of the process, as de-

scribed by Pliny. The litharge could be recycled to produce more lead metal

which would be cast into “lead pigs” which would be weighed before transport.

The administration and security would have closely monitored silver produc-

tion data.

This technological experience with lead and silver production as well as

utilisation of iron ore flux apparently came from elsewhere in the Imperial Ro-

man world, presumably accompanying the mining administration and initial

military presence. Beyond security and possible construction, insufficient detail

exists to demonstrate any clear military participation in the actual technical

work of silver production. For example, Tacitus (Tac. Ann. XI.20) refers to the

opening of silver mines and construction work by Roman soldiers. His descrip-

tion, however, specifically mentions only mining work and construction of wa-

ter channels, but the mined ore had a limited extent and probably a relatively

low silver content. No mention was made of other technical steps or metallur-

gical specialists for assaying, smelting and cupellation which obviously must be

inferred. Was it implied that the soldiers also did the extractive metallurgy?

Engineering work attributed to the Roman Legion, IIII Flavia, stationed at the

fortress in Singidunum seems to have been limited to construction (Wilkes

2000: 106). Within the auxiliary mounted cohort based in the fort near Stojnik,

certainly blacksmiths would have been present for maintenance of iron/steel

weapons, tools and other objects. Roman soldiers could be assigned to various

construction tasks, including deep mining below the surface, but within such

military units would seem unlikely to have normally included any metalworkers

with sufficient technical experience to undertake large-scale lead smelting and


70

cupellation to produce silver. Distinctions between mining and smelting/cupe-

llation activities are significant for assessing the technical capabilities of the

Imperial Roman military for the exploitation of mineral resources.

Along with the arrival of auxiliary mounted troops, I Ulpia Pannoniorum

equitata, transferred from Pannonia to Upper Moesia, other individuals must

have come with the necessary administrative and technological experience for

lead smelting and cupellation on a large scale. According to Du{ani} (1977a:

89), the mining administration consisted of several levels of imperial officials.

Names of some officials are known. It would seem these officials may have had

some authority or experience extending to technical aspects of mining and ex-

tractive metallurgy, too. The term metallarii is used to identify metallurgical

specialists and experienced workers, but this term would probably represent a

wide range of skilled workers with metals and possibly their dependents. Based

upon the Aljustrel tablets of the Lex metalli Vipascensis during the reign of

Hadrian, mining and smelting were undertaken by different groups of workers

(Louis 1921: 3; Healy 1978: 130–133). It would be expected that metallarii

needing security and monitoring, would live initially in the associated civilian

town, essentially huts, outside the auxiliary fort near Stojnik. Healy (1978: 132)

lists other officials with administration responsibilities. Epigraphic evidence ex-

ists for subsequent movement of military units, such as the mounted cohort II

Aurelia nova, to the auxiliary fort and centre of the administrative mining dis-

trict. Military support from a mounted cohort would seem more likely for se-

cure transport of silver. In the Mt. Kosmaj area, Illyrian and Thracian names

are associated with the beginning of mining during the second century (Mücsy

1974: 63–65), but other possibilities exist for a source of metallurgical expertise

for lead and silver production. “Foreigners” from other provinces documented

by inscriptions were attracted by the prosperous conditions around the lead

and silver production (ibid.: 217). Foreign names were also noted in inscrip-

tions from silver mining regions at Rio Tinto and Tharsis (Rothenberg and

Blanco-Freijeiro (1981: 18). Healy (1978: 133, 276) noted that some His-

pano-Iberian terms featured in the mining vocabulary used by Pliny, but there

are also many Greek terms. More inscriptions are known from the Mt. Kosmaj

mining region than other mining regions in Upper Moesia, but aspects of cul-

tural identification and the implications for lead and silver production have not

yet been investigated in detail. The presence of more Roman burials in the

nearby Roman necropolis sites at Guberavac and Gomilica may contribute fur-

ther perspectives on the influx of people associated with lead and silver pro-

duction into the mining district.

Mixed argentiferous and galena were mined and smelted, but the techni-

cal studies of the selected samples in this investigation do not yet document all

technical steps in detail. Only a few examples of litharge have been recovered

from excavations. The technical evidence for best estimate for Imperial Ro-

man silver production currently rests upon published data on silver-rich lead

71


ores with concentrations approximately between 0.18% and 0.43% Ag from

modern mining at Avala (Wray 1921: 65–66) as well as modern production fig-

ures for lead and silver from the early 20th century in Serbia from reworking of

the Roman slag for recovery of lead and silver in Serbia (loc. cit.). Production

levels were variable.

Low values of silver detected with the EPMA in the lead sulphide inclu-

sions in the many examples of tapped lead smelting slag, however, alternatively

suggests that much of the lead ore utilized for smelting was not very silver-rich.

From the excavations, twenty-five smelting slag samples had metallic lead prills

and lead sulphide inclusions with silver concentrations between 0.01% and

0.06% Ag (Mizota 1995). Certainly, higher values should have been readily

identified, if present, using this microanalysis approach. Only the non-de-

structive XRF data for the slag blockage (table 3) indicates a high ratio of sil-

ver to lead, like those published for argentiferous galena concentrate from

near Avala (Wray 1921: 65). Compositional analysis of the samples of tapped

slag are better for representing ancient lead smelting practice, but the more

variable compositional data for slag adhering to fragments of furnace wall may

serve better to reveal higher silver concentrations and silver to lead ratios by

physically entrapping more ore particles. Nevertheless, microanalysis of lead

sulphides in the tapped slag should still reflect the higher silver concentrations.

Compositional analysis of the lead metal and lead coffin (tables 6–7), supports

a theoretical lower threshold of 0.01% Ag, proposed by Healy (1999: 324) for

recovery of silver from lead by the Romans.

The new data from slag and lead metal analyses from the Stojnik samples

indicate a downward range from the highest values reported; perhaps only sil-

ver concentrations in the lead ore from Avala and Rudnik were actually

higher. Based upon the economic geology of the region, the quality of ore var-

ied widely within the three mining zones of the administrative district; some

lead ore had relatively high concentrations of silver, while other ore veins were

not so rich. Recovery of silver from primary smelted lead was variable and pro-

duction depended upon a wide range of lead ore grades. On a large scale with

simultaneous operation of multiple smelting furnaces to increase production,

large quantities of lead ore, flux and charcoal fuel would have been consumed,

so perhaps lead ores were not always assayed first to determine potential silver

yields. Since smelting and cupellation apparently took place in separate loca-

tions, the yield from smelting would have been monitored as lead weights. All

lead metal from primary smelting was apparently collected and transported to

another location for cupellation and subsequent recycling of the litharge back

to lead metal for casting of “lead pigs”. The low silver concentrations in the ex-

cavated lead metal samples (tables 5–6) indicate that cupellation had been

used on almost all lead metal found in the settlement and cemeteries, regard-

less whether the smelted ore contained much silver.


72

Technical studies of the archaeometallurgical remains for lead and silver

production help to illustrate Imperial Roman activities in the mining district as

well as to elaborate the complexity and extent of their “system” for the exploi-

tation of economic mineral resources of interior Upper Moesia. As a terri-

torium metalli (Mócsy 1974: 195), the emperor was the ultimate owner of the

mines and metals. The administration of lead and silver production was effi-

cient in its component parts. The auxiliary mounted cohort provided security

from the beginning for large-scale production, principally covering transport of

materials. It is concluded that metallarii were responsible for actual lead smelt-

ing and controlled slag compositions. Smelting furnaces were concentrated ini-

tially near to the fort, so secure deliveries of lead ore and iron ore from the

mines and charcoal fuel could be maintained. Without reliable delivery of the

required quantities for large-scale production, smelting operations would cease

operation. The mining district officials probably facilitated and monitored de-

liveries as well as production weights, and then collected all silver-rich lead

“bullion” from smelting. A location for cupellation remains uncertain, but it

too would have required metallarii with technical expertise. Cupellation would

have probably been under strict administrative control and monitored closely

by district officials because of the quantity, value and ownership of the silver.

The recovered silver metal was moved again, probably still under administra-

tive control with military protection. Litharge would have been recycled to ob-

tain lead metal, but a location for this step is also uncertain. The technical

steps for lead and silver production were controlled and separated, based upon

the observed distribution of lead smelting slag alone without expected remains

for cupellation or recycling litharge. Administration, transport and security

seem optimised, to variable degrees, within the overall system to increase lead

and silver production. Many aspects of this model, of course, still lack histori-

cal or epigraphic support. Nevertheless, documentation of expected produc-

tion steps for silver from argentiferous lead ore can provide a new interdisci-

plinary perspective. Technical studies of the remaining material evidence near

Stojnik – of a successful 200 year duration for Imperial Roman lead and silver

production on an “industrial” scale in Upper Moesia – reveal an unexpected

consistency in slag compositions.

REFERENCES

Agricola, G.

1950 De Re Metallica. Translated by H. C. Hoover and L. H. Hoover. New York:

Dover Publications.

Bachmann, H.-G.

1980 Early copper smelting techniques in Sinai and in the Negev as deduced

from slag investigations. Pp. 103–134 in Scientific Studies in Early Mining

and Metallurgy, ed. P. T. Craddock. British Museum Occasional Paper 20.

London: British Museum.

73


1982a The Identification of Slags from Archaeological Sites. Institute of Archaeol-

ogy Occasional Publication 6. London: University of London.

1982b Archäometallurgische Untersuchungen zur antiken Silvergewinnung in

Laurion. II. Charakterisierung von Bleiverhüttungsschlacken aus Laurion.

Erzmetall 35/5: 246–251.

Conophagos, C. E.

1980 Le Laurium Antique. Athenes: Ekdotike Hellados.

Craddock, P. T.

1985 Unpublished report. British Museum.

Craddock, P. T., et al.

1985 The investigation of a small heap of silver smelting debris from Rio Tinto,

Huelva, Spain. Pp. 199–217 in Furnaces and Smelting Technology in Antiq-

uity, eds. P. T. Craddock and M. J. Hughes. British Museum Occasional Pa-

per 48. London: British Museum.

Davies, O.

1935 Roman Mines in Europe. Oxford: Clarendon Press.

Domergue, C.

1990 Les Mines de la Péninsule Ibérique dans L’Antiquité Romaine. Roma: École

française de Rome.

Du{ani}, S.

1977a Aspects of Roman Mining in Noricum, Pannonia, Dalmatia and Moesia Su-

perior. Aufsteig und Niedergang der Römischen Welt 2: 52–94.

1977b Mounted Cohorts in Moesia Superior. Pp. 237–247 in Limes, Akten Des XI.

Internationalen Limeskongresses, ed. J. Fitz. Budapest: Akadémiai Kiadü.

1995a Late Roman Mining in Illyricum: Historical Observations. Pp. 219–225 in

Ancient Mining and Metallurgy in Southeast Europe, eds. P. Petrovi}, S.

\ur|ekanovi} and B. Jovanovi}. Belgrade and Bor: Archaeological Insti-

tute and Museum of Mining and Metallurgy.

1995b Minting in the Mining Districts of Roman Illyricum. Pp. 143–144* in Silver

Workshops and Mints, eds. I. Popovi}, T. Cvjeti}anin and B. Bori}-Bre{ko-

vi}. Belgrade: National Museum.

Freestone, I.

1990 Unpublished report. British Museum.

Greene, K.

1990 The Archaeology of the Roman Economy. Berkeley and Los Angeles: Uni-

versity of California Press.

Grèti}, I. A., and Jelenkovi}, R. J.

1995 Characteristics of silver and its findings in Serbia. Pp. 22* in Silver Work-

shops and Mints, eds. I. Popovi}, T. Cvjeti}anin and B. Bori}-Bre{kovi}. Bel-

grade: National Museum.

Healy, J. F.

1978 Mining and Metallurgy in the Greek and Roman World. London: Thames and

Hudson.

1999 Pliny the Elder on Science and Technology. Oxford: Oxford University Press.

Hetherington, R.

1980 Investigations into primitive lead smelting and its products. Pp. 27–40 in As-

pects of Early Metallurgy, ed. W. A. Oddy. British Museum Occasional Paper

17. London: British Museum.


74

Hughes, M. J., Cowell, M. R., and Craddock, P. T.

1976 Atomic absorption techniques in archaeology. Archaeometry 18/1: 19–37.

Jankovi}, S.

1967 Metalogenetske epohe i rudonosna prodru~ja Jugoslavije. Beograd: Rudar-

sko-geolo{ki fakultet.

Jankovi}, S., and Petkovi}, M.

1980 The main lead, zinc and copper deposits of Yugoslavia. Pp. 75–92 in Yugo-

slavia, Outline of Yugoslavian geology, ed. A. Grubi}. Belgrade: National

Committee for Mineral Resources of Yugoslavia.

Louis, H.

1921 Some aspects of mining laws under the Roman Empire. Proceedings of the

[Durham Philosophical] Society 6/2: 1–14.

Matterson, K. J., Peers, C. L., and Shettle, K. N.

1981 Estimation of sulphur by combustion. Pp. 299–302 in Studies in Ancient

Mining and Metallurgy in South-West Spain, eds. B. Rothenberg and A.

Blanco-Freijeiro. London: Institute for Archaeo-Metallurgical Studies.

Mattingly, H.

1968 Coins of the Roman Empire in the British Museum IV: Antoninus Pius to

Commodus. London: British Museum.

Merkel, J.

1984 Unpublished report. Peabody Museum, Harvard University.

1990 A reconstruction of copper smelting at Timna. Pp. 78–122 in The Ancient

Metallurgy of Copper, ed. B. Rothenberg. London: Institute for Archaeo-

-Metallurgical Studies.

Middleton, A. P., et al.

2000 Scientific Examination of Some Ceramic Materials and Samples of Litharge.

Pp. 156–168 in King Croesus’ Gold, Excavations at Sardis and the History of

Gold Refining, eds. A. Ramage and P. T. Craddock. London: British Mu-

seum Press.

Mizota, H.

1995 Lead-silver production in a Yugoslav (Serbian) site during the Roman period.

M.Sc. research. Institute of Archaeology, University College London.

Mócsy, A.

1974 Pannonia and Upper Moesia. London and Boston: Routledge and Kegan

Paul.

Orton, C.

2000 Sampling in Archaeology. Cambridge: Cambridge University Press.

Pernicka, E.

1981 Archäometallurgische Untersuchungen zur antiken Silvergewinnung in

Laurion. I. Chemische Analyse griechischer Blei-Silber-Erz. Erzmetall 34:

396–400.

Pernicka, E., and Bachmann, H.-G.

1983 Archäometallurgische Untersuchungen zur antiken Silvergewinnung in

Laurion. III. Das Verhalten einiger Spurenelemente beim Abtreiben des

Bleis. Erzmetall 36: 592–597.

75


Photos-Jones, E., and Jones, J.

1994 The building and industrial remains at Agrileza, Laurion (4th century B.C.)

and their contribution to the workings at the site. Annual of the British

School at Athens 89: 307–358.

Pliny, the Elder

1952 Natural History. The Loeb Classical Library, Vol. 9, Libri XXXIII–XXXV.

English translation by H. Rackham. London: Heinemann.

Popovi}, I.

1984 Unpublished report. National Museum, Belgrade.

Radi}, S., and Vukovi}, M.

1957 Pregled mineralizacionog ciklusa rudnika “Rudnik” u [umadiji. Vesnik

Zavoda za geolo{ka i geofizi~ka istraìvanja (Beograd) 14: 111–134.

Ramage, A.

2000 Inventory and descriptions of finds of equipment and supplies. Pp. 221–232

in King Croesus’ Gold, Excavations at Sardis and the History of Gold Re-

fining, eds. A. Ramage and P. T. Craddock. London: British Museum Press.

Ra{kovi}, D.

1995 Finds of Roman silver coins in the Kru{evac area. Pp. 206* in Silver Work-

shops and Mints, eds. I. Popovi}, T. Cvjeti}anin and B. Bori}-Bre{kovi}. Bel-

grade: National Museum.

Rehren, Th., and Prange, M.

1998 Lead metal and patina: a comparison. Pp. 183–196 in Metallurgica Antiqua:

In honour of Hans-Gert Bachmann and Robert Maddin, Der Anschnitt 8, eds.

Th. Rehren, A. Hauptmann and J. D. Muhly. Bochum: Deutsches Bergbau-

-Museum.

Rehren, Th., Vanhove, D., and Mussche, H.

2002 Ores from the ore washeries in the Laviotiki. Metalla 9/1: 27–46.

Rehren, Th., et al.

1999 Litharge from Laurion – a medical and metallurgical commodity from

South Attika. L’Antiquité Classique 68: 299–308.

Rothenberg, B., and Blanco-Freijeiro, A. (eds.)

1981 Studies in Ancient Mining and Metallurgy in South-West Spain. London: Insti-

tute for Archaeo-Metallurgical Studies.

Ruddle, R. W.

1957 Difficulties encountered in smelting in the Lead Blast Furnace. London: Insti-

tution of Mining and Metallurgy.

Tac. Ann. = Tacitus, Cornelius

1909 The Annals of Tacitus, Books XI–XVI. Translated by G. G. Ramsay. Lon-

don: John Murray.

Todd, M.

1996 Pretia Victoriae? Roman Lead and Silver Mining on the Mendip Hills,

Somerset, England. Münsterische Beiträge z. antiken Handelsgeschichte 15/1:

1–18

Tomovi}, M.

1995 Roman Mines and Mining in the Mountain of Kosmaj. Pp. 203* in Ancient

Mining and Metallurgy in Southeast Europe, eds. P. Petrovi}, S. \ur|e-


76

kanovi} and B. Jovanovi}. Belgrade and Bor: Archaeological Institute and

Museum of Mining and Metallurgy.

Tylecote, R. F.

1962 Metallurgy in Archaeology. London: Edward Arnold.

1976 A History of Metallurgy. London: The Metals Society.

1986 The Prehistory of Metallurgy in the British Isles. London: The Institute of

Metals.

1987 The Early History of Metallurgy in Europe. London: Longman.

Wray, D. A.

1921 The Geology and Mineral Resources of the Serb-Croat-Slovene State. Report of

the Geologist attached to the British Economic Mission to Serbia. London: His

Majesty’s Stationery Office.

Werner, M. R.

1990 The Roman Military-Civilian Site at Stojnik in the Mt. Kosmaj Mining Re-

gion, Moesia Superior: The 1984–1985 Field Seasons. American Journal of

Archaeology 90/2: 221.

Wilkes, J. J.

1981 The Illyrian Provinces: External Threat and Internal Change. Pp. 515–524

in The Roman West in the Third Century, eds. A. King and M. Henig. BAR

International Series 109. Oxford: British Archaeological Reports.

2000 Roman Legions and their Fortresses in the Danube Lands [first to third

centuries AD]. Pp. 101–119 in Roman Fortresses and their Legions: Papers in

honour of George C. Boon, ed. R. J. Brewer. London and Cardiff: Society of

Antiquaries of London and National Museums & Galleries of Wales.

* English Summary

YON F. MERKEL

PROIZVODWA OLOVA I SREBRA

U RIMSKOM CARSKOM PERIODU

U SEVERNOM DELU GORWE MEZIJE (OBLAST KOSMAJA)

Rezime

Arheometalur{ka prou~avawa proizvodwe olova i srebra tokom rim-

skog carskog perioda u blizini sela Stojnik na planini Kosmaj (oko 30 km

ju`no od Beograda) predstavqaju deo istraìva~kog projekta koji je ostva-

rila 1983–1988. srpsko-ameri~ka ekipa, ~iji su fokus ~inile metalur{ke

aktivnosti u blizini vojnog utvr|ewa i okolnog civilnog naseqa (sl. 1–3;

t. I/1) Uzorci metalur{kih aktivnosti sakupqeni tokom iskopavawa ispi-

tani su na Institutu za arheologiju Univerzitetskog koleya u Londonu i u

Britanskom muzeju. Tehni~ka prou~avawa obavqena su na uzorcima zgure,

olova i olovo-monoksida kako bi pokazala glavne faze proizvodwe olova i

srebra. Rezultati analiza predstavqeni su u tabelama 1–7. Najraniji uzor-

77


ci, datovani numizmati~kim nalazima, svedo~e o proizvodwi metala nepo-

sredno izvan tvr|ave osnovane 167–169. Kasnije, topqewe olova se poja~ano

vr{i i oko civilnog naseqa. Mada u sondama tokom iskopavawa nije

otkrivena nijedna pe}, najve}i primerci pe}ne zgure (sl. 4; t. I/2) bogate

gvo`|em teìli su ~ak 20 kg, {to je omogu}ilo pribli`nu procenu i

drugih parametara vezanih za rad pe}i. Minerolo{ka analiza olovne zgure

ukazala je na dominantnu koncentraciju fajalita (FeSiO4), koji je obi~no

kori{}en za topqewe bakra i gvo`|a. Mikroanaliza zgure potvrdila je

prisustvo neizreagovanog olovo-sulfida u pe}noj zguri fajalitnog tipa.

Takav sastav zgure prisutan je od samog po~etka rimske carske proizvodwe

olova u okolini Stojnika. Eksperimentalna rekonstrukcija proizvodwe

olova sugeri{e me{avinu srebronosnih olovnih ruda galenita (PbS) i

keruzita (PbCO3), kao najverovatnijih minerala koje su Rimqani eksploa-

tisali u regionu. Savremeni geolo{ki radovi o sadràju galenita isko-

ri{}eni su kao indikator raspona dostupnih ruda, ali u analizi zgure

rimskog carskog perioda konstatovane su i niè vrednosti. Olovo u formi

metala je kupelacijom daqe prera|ivano u ciqu dobijawa srebra. Usta-

novqeni osnovni hemijski sastav uzoraka olova sa lokaliteta i obli`wih

nekropola ukazuje, pored ostalog, da je izmerena koncentracija srebra u

olovu, karakteristi~no, ispod 0,06% Ag. Unutar celokupnog sistema, as-

pekti administracije, transporta i bezbednosti bili su, u okviru mo-

gu}nosti, optimizovani, u ciqu pove}awa proizvodwe olova i srebra.

Received: 9 May 2007

UDC 904-034(497.11 Stojnik):669(091)(37)


78

��

�� !" �� #� � �� $�� %�

�� &� �� '�� !(�& �� '�� " �� $�� %�

John Merkel - 2007 - Imperial Roman Production Of Lead And Silver In The Northern Part Of Upper Moesia Mt - Kosmaj Area

Documents

lead smelting

lead metal

lead ores

silver mining

roman lead ingots

lead pigs

silver deposits

finest silver