bs_bs_banner

Archaeometry 57, 5 (2015) 869–878 doi: 10.1111/arcm.12128

A PR EL I MINARY ST UDY ON T H E LO S S O F I R O N A N D
A R SE N IC IN T HE RE -M E LT IN G O F I R O N - BEA R I N G
ARSE NICAL CO PPER *

J.-S. PARK†
Department of Metallurgical Engineering, Hongik University, Jochiwon, Sejong 339-701, Korea

and E. GELEGDORJ
Institute of Archaeology, Mongolian Academy of Sciences, Jucov street 77, Ulaanbaatar-51, Mongolia

Re-melting experiments were conducted with specimens made of high-arsenic copper alloys
containing lead, iron and sulphur. The melting treatment at 1000°C in ambient conditions for
5 min was found to cause an almost complete removal of iron and sulphur along with a notable
reduction of arsenic levels. Evidence was found that the preferential oxidation of iron and
arsenic was responsible for this change in composition. The treatment also brought about
perfect dissolution of speiss particles scattered throughout the original specimen, suggesting
the possibility of the addition of speiss to molten copper for the production of arsenical copper.

KEYWORDS: RE-MELTING, COPPER ALLOYS, ARSENIC, IRON, SPEISS

INTRODUCTION

In their work on the evidence for mass production of speiss and arsenical copper in Iran during
the Early Bronze Age, Rehren et al. (2012) predicted that iron would preferentially be lost in
molten Cu–As–Fe alloys before the arsenic level would start to be affected. This prediction
suggests that a considerable change in composition can occur when such alloys are subjected to
a series of re-melting treatments, whether for casting and alloying or for recycling. The loss of
iron and arsenic is then inevitable (Pernicka 1999) and the composition of the final products could
significantly depart from that of the starting material. This variation is extremely difficult to
predict accurately from either a theoretical or experimental basis. It may therefore act as a serious
barrier to the understanding of ancient bronze technologies, when employing data obtained from
the examination of chemical compositions of the finished products. If properly studied, however,
such variation may provide valuable information on the engineering processes involved and the
extent to which an ancient community depended on the production of fresh metals versus the
circulation of recycled materials. One of the best examples of this possibility was provided by
Bray and Pollard (2012) in their re-examination of the existing data sets accumulated for the
chemistry of bronze objects from the early Bronze Age of the British Isles. They could derive
from the variation in arsenic contents important clues as to the methods of fabrication and
diffusion of raw materials from their production centre. The success in their research emphasizes
the need for understanding the behaviour of alloying elements in copper in a range of engineering
processes such as casting, re-melting and thermomechanical treatments. This subject has been in

*Received 4 February 2014; accepted 17 June 2014

†Corresponding author: email jskpark@hongik.ac.kr
© 2014 University of Oxford
870 J.-S. Park and E. Gelegdorj

Figure 1 A map showing the Mongolian provinces. The arrow indicates the Xiongnu burial site at Duurlig Nars in
Khentii province, from which the bronze object under consideration was retrieved.

discussion (McKerrell and Tylecote 1972; Charles 1980) but is generally unresolved to this day.
In light of this concern, we performed a set of re-melting experiments with small specimens taken
from a high-arsenic bronze object from a Xiongnu burial site in Mongolia dated between the first
century bc and the first century ad. The site is located at Duurlig Nars in Khentii province, which
is marked by the arrow in Figure 1, a map showing the provinces of Mongolia. The object under
investigation was recovered from the 2006–7 Mongol–Korean collaborative expedition called the
Mon-Sol Project (Yun and Chang 2011). Figure 2 shows its general appearance, suggesting that
the object had served as a dish before it was broken. It was retrieved from a spot where a group
of bronze objects had been broken and burned, apparently as part of a unique burial ceremony.
Figure 2 illustrates only the major fragments, which were relatively large in size, with the inset
showing two tiny spherical specimens obtained from the re-melting experiments, as described in
the next section.

EXPERIMENTS

A small fragment of the object, not shown in Figure 2, was chosen to provide specimens for both
metallographic examination and re-melting experiments. Two small pieces taken from the frag-
ment were placed on a flat tray and heated in a box furnace set at 1000°C for 5 min, followed by
either slow cooling in the furnace or quenching in water. This treatment produced two spherical
specimens, each weighing approximately 250 mg and 3.5 mm in diameter, as can be seen in the
inset of Figure 2. Both the original and the treated specimens were mounted and polished
following standard metallographic procedures and then etched, if needed, with a solution of
100 ml methyl alcohol, 30 ml hydrochloric acid and 10 g ferric chloride. Their microstructures
were examined using an optical microscope and a scanning electron microscope (SEM). The
chemical composition was measured using an energy-dispersive X-ray spectrometer (EDS), fitted

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

 Loss of iron and arsenic in re-melting of iron-bearing arsenical copper 871

Figure 2 The general appearance of the bronze object examined. The inset shows two spherical specimens solidified
after a re-melting treatment carried out at 1000°C in an atmospheric environment for 5 min.

in the SEM instrument. The chemical information is reported in weight fraction to one tenth of
a per cent. The average alloy composition was calculated from the EDS spectra taken in a raster
mode, mostly from a rectangular area of approximately 0.65 mm × 0.45 mm.

RESULTS

Figure 3 (a) is an optical micrograph showing the original structure of the object before the
treatment. The absence of notable dendrite structures in this figure indicates that the microstruc-
ture determined in casting was substantially modified at some point by a thermal treatment given
in a solid state. The structure, however, shows no evidence of forging and now consists primarily
of the α phase and numerous dark areas corresponding either to cavities or to particles of lead.
In Figure 3 (b), where an SEM micrograph covering approximately the same area as Figure 3 (a)
is shown, the lead particles appear bright while the cavities remain dark. The average alloy
contents inferred from Figure 3 (c), a spectrum taken from the entire area of Figure 3 (b), are
9.6% As – 4.2% Pb – 1.9% Fe, with traces of tin and sulphur. Figure 3 (d) presents an SEM
micrograph magnifying the same area neighbouring the arrows in Figures 3 (a) and 3 (b). Arrows
1–6 in Figure 3 (d) point to the spots where the EDS analyses were performed to obtain the
spectra presented in Figures 4 (a) – 4 (f), respectively. Table 1 summarizes the alloy compositions
that were similarly determined at arrows 1–4. The precipitate at arrow 1, one of many similar

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

872 J.-S. Park and E. Gelegdorj

cps

300

200

100

0
Energy (keV)
0 5 10

Figure 3 Optical and SEM micrographs with an EDS spectrum illustrating the results obtained from the original
specimen: (a) an optical micrograph; (b) a SEM micrograph covering approximately the same area shown in (a), with
the arrows in both referring to the same spot; (c) an EDS spectrum taken from (b); (d) a SEM micrograph enlarging the
spot marked by the arrow in (b), with arrows 1–6 identifying the spots from which EDS spectra were taken and presented
in Figures 4 (a)–4 (f), respectively.

precipitates scattered over Figure 3 (a), corresponds to the copper – iron arsenide phase called
speiss. The speiss composition was found to vary significantly but the arsenic level was fairly
consistent at around 40%. Figure 4 (b) shows that the neighbouring precipitate at arrow 2
contains arsenic as the major alloying element by approximately 33.3%. This composition is near
that of the γ phase in the Cu–As alloy system. The spectra in Figures 4 (e) and 4 (f) indicate that
the particle at arrow 5 consists of copper sulphide, approximating Cu2S in composition, while that
at arrow 6 corresponds to lead.
Figures 5 (a) and 5 (b) present optical micrographs covering the cross-section of the treated
specimens solidified at two different cooling rates. The difference in contrast between them is due
to the etching applied only to Figure 5 (b), where much finer dendrites were precipitated as a
result of the rapid cooling rate. Figure 5 (c), an optical micrograph magnifying the area sur-
rounding the arrow in Figure 5 (b), consists of dark α dendrites and the bright interdendritic
regions where a phase greatly enriched in arsenic exists, along with numerous dark particles of
lead. No precipitates of speiss or sulphide, however, were observed in this specimen. Figure 5 (d),
an EDS spectrum taken from the area near the arrow in Figure 5 (b), shows the presence of copper

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

 Loss of iron and arsenic in re-melting of iron-bearing arsenical copper 873
cps cps

150

300

100

200

50
100

0 0
Energy (keV) Energy (keV)
0 5 10 0 5 10
cps cps
500

400
400

300
300

200 200

100 100

0 0
Energy (keV) Energy (keV)
0 5 10 0 5 10
cps cps

400

300

300

200
200

100
100

0 0
0 5 Energy (keV) 10 0 5 Energy (keV) 10

Figure 4 EDS spectra: (a)–(f) spectra taken from arrows 1–6 in Figure 3 (d), respectively.

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

874 J.-S. Park and E. Gelegdorj

Table 1 The chemical composition of the phases located at arrows 1–4

in Figure 3 (d) as determined by the EDS analyses carried out in a point
mode

Location in Figure 3 (d) Chemical composition (wt%)

Cu As Fe Sn Sb Ni

Arrow 1 26.4 43.1 29.4 – – 1.1

Arrow 2 64.6 33.3 0.4 – 1.7 –
Arrow 3 86.9 11.7 0.4 1.0 – –
Arrow 4 89.3 9.0 1.0 0.7 – –

–, None detected.

(Cu), arsenic (As), lead (Pb) and tin (Sn). The average alloy contents as inferred from this
spectrum are 7.1% As – 5.0% Pb – 0.7% Sn. The iron and sulphur levels, however, were below
the detection limit of the EDS instrument, indicating that these elements were mostly lost during
the treatment. It should also be noted that there was a notable reduction in arsenic levels from
9.6% down to 7.1%. The apparent increase in lead contents from 4.2% to 5.0% probably resulted
from an inaccuracy due to the noticeably uneven distribution of lead particles in the original
specimen, as shown in Figure 3 (b).
Figure 6 (a) presents an SEM micrograph showing the surface of the specimen that was cooled
at a slow rate. The surface seen in Figure 6 (a) consists of two different regions, as noted at arrows
1 and 2. Their highly magnified views are provided in the SEM micrographs of Figures 6 (b) and
6 (c), respectively. Both regions are covered with numerous fine grains. The morphological
variation, however, is evidently associated with the agglomeration of these grains, which
occurred only in Figure 6 (c) to form large clusters. Figures 6 (d) and 6 (e), containing EDS
spectra taken from Figures 6 (b) and 6 (c), respectively, reveal that the peak intensity of iron (Fe)
significantly increased in both regions, suggesting that there was preferential oxidation of iron
atoms on the specimen’s surface. This reaction most probably accounts for the depletion of iron
in the interior, as is verified in Figure 5 (d). Also, the loss of arsenic is anticipated from its relative
peak (As) height, which is higher in both spectra than that of the interior, as noted in Figure 5 (d)
as well. The most conspicuous difference between Figures 6 (d) and 6 (e) is found in the peak
intensity of lead (Pb), which is much stronger in the former. It is certain, therefore, that the visible
difference in surface topography resulted from variations in lead content.

DISCUSSION

The results presented above demonstrate that re-melting treatments can cause an almost complete
removal of iron and sulphur (see Fig. 5 (d)), if any, from arsenical copper alloys, along with a
notable reduction in arsenic content. Depending on the conditions given, therefore, iron and
sulphur as well as arsenic could be substantially removed from copper alloys even in a single
brief re-melting operation and be further depleted with repeated casting for recycling. The
absence of iron and sulphur in final products, therefore, should be used with caution in predicting
the chemistry of specific ores used in smelting. According to the composition of the treated
specimen in question, the removal of these elements, which are highly prone to oxidation, was
driven by preferential surface oxidizing of the molten alloy. It is significant, however, to note that

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

 Loss of iron and arsenic in re-melting of iron-bearing arsenical copper 875

(b)
cps

300

200

100

0
Energy (keV)
0 5 10

Figure 5 Optical and SEM micrographs with an EDS spectrum illustrating the results obtained from the specimens
solidified after a re-melting treatment carried out at 1000°C in an atmospheric environment for 5 min: (a) an optical
micrograph showing the structure of the specimen solidified very slowly inside the furnace; (b) an optical micrograph
showing the structure of the specimen solidified very rapidly inside the water; (c) an optical micrograph enlarging the
area marked by the arrow in (b); (d) an EDS spectrum taken from the area near the arrow in (b).

the loss of iron was almost completed while there was less than 30% reduction in arsenic levels.
This result is in support of the theory proposed by Rehren et al. (2012) that the depletion of iron
would be accomplished before the beginning of the notable loss of arsenic.
The numerous iron – copper arsenide particles scattered in the α background (Fig. 3 (a))
were probably formed during smelting, while the level of iron in the melt was still significant
or introduced later in the alloying operations. The complete absence of such particles in the
treated specimen indicates that they existed only as an unstable transient phase that dissolved
in molten copper under the given oxidizing atmosphere of ambient conditions. It can be
deduced that the selective removal of iron from molten copper alloys, as discussed above,
accelerates the dissolution of the particle. This will in turn raise the arsenic level in the sur-
rounding area, resulting in a depression of the melting points. This region is then solidified last
and arsenic can be further enriched due to elemental segregation attending the solidification

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

876 J.-S. Park and E. Gelegdorj

cps

60

40

20

0
Energy (keV)
0 5 10

cps
60
(e)

40

20

0
Energy (keV)
0 5 10

Figure 6 SEM micrographs and EDS spectra illustrating the results obtained from examining the surface of the
specimen solidified very slowly inside the furnace after the re-melting treatment: (a) a micrograph showing the surface
of the specimen; (b), (c) micrographs enlarging the areas marked by arrows 1 and 2 in (a), respectively; (d), (e) spectra
taken from the entire area of (b) and (c), respectively.

reaction. This may lead to the precipitation of arsenic-rich phases such as the γ particles around
the dissolving speiss particles, as can be verified in Figure 3 (d). The speiss particles present
in the original specimen were iron – copper arsenide based. Their arsenic levels were higher
and more consistent than the iron and copper levels. It is not clear whether their initial com-
position was maintained or subsequently modified. It is plausible, however, that the depletion
of iron in the neighbouring molten alloy would have caused a selective removal of iron atoms
from the speiss particles, leaving room for compositional variation within the particles
themselves.

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

 Loss of iron and arsenic in re-melting of iron-bearing arsenical copper 877

In their smelting experiments with mixed oxide-based copper ores and arsenic-bearing ores,
Lechtman and Klein (1999) verified the formation of copper – iron arsenide speisses of variable
composition. They found that speiss particles occurred only in alloys smelted in a furnace, but not
in those smelted in a crucible, where ores remained isolated from burning charcoal. As a direct
result, the former had significant iron contents as opposed to the latter, whose iron concentrations
were negligible. By contrast, Rehren et al. (2012) reported evidence of mass production of iron
arsenide speisses as a target product, not as an unwanted by-product, at the Early Bronze Age site
in Iran. This evidence led them to propose processes in which speiss was added either to oxide
copper ores during smelting or directly to molten copper in alloying operations. As for the object
under investigation, it is uncertain how its alloy composition was determined. Even so, the
presence of speiss precipitates and the fairly high level of iron are consistent with some of the
processes proposed for the production of arsenical copper, suggesting that they were likely in
practice. The re-melting experiment carried out in this study did, however, reveal that alloy
composition as determined by such processes could be altered considerably depending on the
frequency and conditions of re-melting treatments as applied subsequently to casting.

CONCLUSIONS

Re-melting experiments were performed with high-arsenical copper specimens containing iron,
sulphur, lead and tin. The specimens were taken from a bronze object retrieved in the excavation
of a Xiongnu burial at Duurlig Nars in Khentii province, Mongolia. The object contained arsenic
as a solute in the background α phase and as a major constituent of the iron – copper arsenide
speiss precipitates, which were scattered over the α background. The re-melting treatment in
ambient conditions given to the initial object specimens was found to cause preferential removal
of iron and sulphur from the molten alloys before the arsenic level began to decrease. The
significant enrichment of iron and the fairly high arsenic content observed in the surface of the
treated specimen indicate that the early removal of iron, followed by the reduction in arsenic
concentration, from inside the molten alloy, resulted from a selective oxidation reaction on the
specimen’s surface. The re-melting process also caused the complete dissolution of speiss
particles and no such precipitates were retained in the treated specimen.
The high iron content and the presence of speiss precipitates observed in arsenical copper
alloys could be indicative of certain processes employed for their production (Lechtman and
Klein 1999). According to the results reported herein, however, even a single re-melting treatment
can remove both iron and speiss precipitates from alloys, making it virtually impossible to
estimate their original composition and microstructure. This variation has been discussed
(McKerrell and Tylecote 1972; Pernicka 1999; Bray 2012), but little attention has been given to
the microstructure of arsenical copper objects. In fact, the majority of the reported composition
data has had little or no such microstructural information (Barnard 1961; Chernykh and
Kuzminykh 1989; Chase and Douglas 1997). The presence of speiss precipitates, therefore, has
not yet been properly tested and their significance in arsenical copper alloys remains underesti-
mated. If their ready solubility in molten copper had been realized in antiquity, however, speiss
would have been produced and circulated as a valuable commodity for use in the making of
arsenical copper alloys (Rehren et al. 2012). Subsequently, it would also have played a significant
role in ancient bronze industries. We expect that careful examination of the variation in iron and
arsenic concentrations along with their ratio and the fraction of speiss phase present, as a function
of the re-melting conditions, could lead to better estimations of the treatments applied to copper
alloys.

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878

878 J.-S. Park and E. Gelegdorj

ACKNOWLEDGEMENTS

This work was supported by the Korea National Research Foundation (NRF-
2013R1A1A2059128). Two anonymous reviewers are acknowledged for their valuable com-
ments and suggestions. We thank Mr Jason Buschman for his linguistic assistance with this
manuscript.

REFERENCES

Barnard, N., 1961, Bronze Casting and Bronze Alloys in Ancient China, Monumenta Serica Monograph 14, The
Australian National University and Monumenta Serica, Canberra.
Bray, P. J., 2012, Before 29Cu became copper: tracing the recognition and invention of metalleity in Britain and Ireland
during the third millennium B.C., in Is there a British Chalcolithic? People, place and polity in the later 3rd
millennium (eds. M. Allen, J. Gardiner and A. Sheridan), 56–70, The Prehistoric Society Research Paper 4, Oxbow
Books, Oxford.
Bray, P. J., and Pollard, A. M., 2012, A new interpretative approach to the chemistry of copper-alloy objects: source,
recycling and technology, Antiquity, 86, 853–67.
Charles, J. A., 1980, Recycling effects on the composition of non-ferrous metals, Philosophical Transactions of the Royal
Society of London Series A, 295, 57–68.
Chase, W. T., and Douglas, J. G., 1997, Technical studies and metal compositional analyses of the bronzes of the Eastern
Eurasian steppes from the Arthur M. Sackler collections, in Ancient bronzes of the Eastern Eurasian steppes from the
Arthur M. Sackler collections (eds. E. C. Bunker, T. S. Kawami, K. M. Linduff and W. En), 306–18, The Arthur M.
Sackler Foundation, Distributed by Harry N. Abrams, Inc., Publishers.
Chernykh, E. N., and Kuzminykh, S. V., 1989, Ancient metallurgy in northern Eurasia (Seyma–Turbino phenomenon),
Academy of Sciences of the USSR Institute of Archaeology, Nauka, Moscow (in Russian).
Lechtman, H., and Klein, S., 1999, The production of copper–arsenic alloys (arsenic bronze) by cosmelting: modern
experiment, ancient practice, Journal of Archaeological Science, 26, 497–526.
McKerrell, H., and Tylecote, R. F., 1972, The working of copper–arsenic alloys in the Early Bronze Age and the effect
on the determination of provenance, Proceedings of the Prehistoric Society, 38, 209–18.
Pernicka, E., 1999, Trace element fingerprinting of ancient copper: a guide to technology or provenance, in Metals in
antiquity (eds. S. M. M. Young, A. M. Pollard, P. Budd and R. A. Ixer), 163–71, British Archaeological Reports,
International Series 792, Archaeopress, Oxford.
Rehren, Th., Boscher, L., and Pernicka, E., 2012, Large scale smelting of speiss and arsenical copper at Early Bronze Age
Arisman, Iran, Journal of Archaeological Science, 39, 1717–27.
Yun, H. W., and Chang, E. J., 2011, Excavations of Xiongnu tombs at Duurlig Nars cemetery in eastern Mongolia, in
Xiongnu archaeology: multidisciplinary perspectives of the first steppe empire in Inner Asia (eds. U. Brosseder and
B. K. Miller), 261–74, Rheinische-Friedrich-Wilhelms-Universitat, Bonn.

© 2014 University of Oxford, Archaeometry 57, 5 (2015) 869–878