6 Roman Copper Alloys: A General Examination

6.1 Introduction

This section describes the copper metallurgy of the Roman period as a whole while other sections consider the influence of artefact typology, date, provenance, etc. on alloy composition. An examination of the alloy composition of all the Roman samples at once allows the characterisation of Roman copper metallurgy more from a metallurgical perspective than an archaeological one. The results for the alloy elements (zinc, tin and lead) are initially dealt with individually, and then considered together (to examine the inter-relationships between them). In particular, use is made of 3-D plotting to examine the complex relationship between zinc and tin in Roman copper alloys. The results are compared with previously published analyses of objects from other parts of the Roman Empire. Finally, this section sets out the analytical results for some of the metal impurities (iron, nickel and arsenic).

6.2 Sampling

Almost all previous studies of Roman copper alloys have examined a limited range of object types (e.g. statuary, brooches). In order to gain a more representative picture of Roman copper alloys in the study area (northern Britain), samples were collected from all parts of the study area, from all the periods of occupation (with dating primarily by context but supplemented with typological information), from all the types of sites present, and including all the sorts of objects found on these sites.

While the overall chronology of the samples reflects the known archaeology of northern Roman Britain (most of the samples come from 1st or 2nd century contexts), there is some variation between different types of site. Most of the samples from forts come from the 1st and 2nd centuries, reflecting the intensive construction and moving activities the army carried out in the early Empire. Most of the evidence for extramural civil settlements has been dated to the 2nd century. The relative scarcity of 1st century samples suggests that there may have been a delay in the setting up of vici. There is little 3rd century (and no 4th century) evidence from vici, but the evidence from towns is greatest in this late period. Villas are virtually absent before the 2nd century.

There are overall biases in the evidence and material from different sorts of sites is not always strictly comparable (consisting of a variety of objects from all types of contexts of different dates); this largely reflects the biases of the archaeological record.

6.3 Alloying in the Roman Period

The work of Craddock (1975; 1978; forthcoming), Picon et al. (1966; 1967; 1968), Beck et al. (1985), Bayley (1992), Riederer (1974a; 1974b) has shown that Roman copper alloys were made using the addition of zinc, tin and/or lead to copper. The results presented in this chapter support many of the findings of this earlier work. Zinc is characteristic of the Roman period as this is the first period in Europe when it is regularly used (Craddock 1978). It is of some chronological significance when examining late Iron Age artefacts or determining the authenticity of certain finds. Nevertheless, tin remains the prevalent alloying element. Many objects are made from alloys which can be loosely referred to as gunmetals (they contain appreciable levels of both zinc and tin). Modern gunmetals are deliberately made but ancient gunmetals may have arisen accidentally through the mixing of bronze and brass scrap metal. Lead is added in varying quantities (usually to cast objects). The presence of each alloying element (zinc, tin and lead) is dealt with separately below. The results are then brought together to examine the relationship between the different alloying elements.

6.3.1 Zinc

Forty percent of all Roman alloys had at least 5% zinc. The distribution of zinc in all Roman alloys is fairly flat between 5 and 25% (Figure 31). This apparently even spread of zinc contents is an over-simplification. Zinc content varies with time - high zinc alloys belonging to the early Roman period. In addition zinc is strongly correlated (inversely) with tin (see Figure 34). The alloy type classification discussed below (see Figure 35) defines brasses as those alloys with 15% or more zinc. The method of brass production at this time was the cementation method (Craddock 1978) which could yield brass with a maximum zinc content of c. 28%. The paucity of such alloys (those with more than 23% zinc) in all the samples analysed here is striking.

Fig.31 Distribution of Zinc contents in all Roman alloys

6.3.2 Tin

Tin is found more frequently than zinc in Roman alloys (54% of all alloys have more than 5% tin). The distribution of tin contents is distinctly bi-modal, with one peak around zero (indicating the brasses and more-or-less pure coppers) and another at c. 8-10% (Figure 32). There is a small number of alloys with relatively high tin contents (tin content over 16%). Many of these are mirrors made of speculum (Craddock 1975). For the alloy type classification used here (see Figure 35) bronzes are alloys with 5% or more tin (except those up to 10% tin where, tin + zinc > 10%). The remaining alloys, except for those with very high copper levels, are classified as gunmetals.

Fig.32 Distribution of Tin in all Roman alloys

6.3.3 Lead

Lead is the least common of the alloying elements in Roman copper alloys (only 25% of all Roman samples had more than 1% lead). In addition, the distribution of lead contents decreases logarithmically (only 15% had over 9% lead - see Figure 33). Most Roman alloys therefore had relatively low levels of lead, even though 63% of all Roman alloys (where the method of fabrication could be determined) were cast - see Table 6.2.

Smythe (1936) noted that lead was found in greater quantities in cast objects than wrought ones. This is confirmed by the results of this project (summarised in Table 6.3). A small number of the analysed samples had lead contents which were extremely high (20% or more). Such alloys are not now used because of their poor mechanical strength. They would only be suitable for casting decorative objects which would have to take no strain (including strain during finishing). Lead is always present as discrete globules, but at high levels the excess lead will tend to segregate into a core which may have more lead than copper. There is, therefore, some doubt concerning the accuracy of the lead contents at these levels, since the establishment of lead determinations for such alloys depends on the number of samples and their siting and depth. A further degree of inaccuracy is introduced as the calibration equation is extrapolated beyond the lead content of the standards. An accurate estimation of lead content can in this situation only be obtained through total wet chemistry analysis of the entire artefact.

Fig.33 Distribution of Lead in all Roman alloys

6.4 The relationships between different alloying elements

Smythe (1936) and Craddock (1975) noted that some of the alloying elements in Roman copper alloys were correlated with each other. In particular, tin and zinc levels were inversely proportional to each other. It is valuable, therefore, to examine the inter-relationships between tin, zinc and lead. Craddock (1975) and Caple (1986) achieved this through 2-D plots of zinc against tin (see Figure 2.2). Caple has also used 2-D plots to indicate the gaps in the distribution of alloy type, e.g. alloys containing high levels of zinc and tin are almost unknown. In order to examine the relationship between alloy elements, and identify peaks and troughs in the distribution of alloy types, a 3-D surface graph is used here (Figure 34). The two horizontal axes show the zinc and tin content while the vertical axis shows the frequency of any particular alloy type.

Fig.34 Smoothed 3-D surface chart showing the zinc and tin contents of Roman copper alloys

Figure 34 clearly shows two main peaks: one around c. 8% tin (with zinc approaching zero) and a second around c. 18% zinc (with tin approaching zero). These two peaks are bronze and brass respectively. It can be seen from this figure that bronzes were more common than brasses. This is somewhat surprising given the importance (among modern researchers) that is attached to the presence of zinc in alloys of the Roman period.

There is a spread of results between the two main peaks of brass and bronze. This range of alloys comprises the gunmetals which contain zinc and tin. The lack of any distinct peak in this region suggests that no one intermediary composition was particularly favoured over others. The mixing of bronze and brass in varying proportions could have produced the observed pattern.

The almost complete absence of alloys containing moderate amounts of zinc (5-15%) and almost no tin is striking. This suggests that brasses were rarely recycled on their own (if they had then there should be more samples with 5-15% zinc (and no tin). If brasses were recycled they were almost always mixed with bronze.

A third (smaller) peak can also be seen around c. 2% tin (again, zinc approaching zero). This metal is technically a bronze (the only alloying element present is tin) but the tin content is very low. Many of the objects made of this alloy are sheet items (for which a low tin content would be appropriate as the metal would be more malleable). In addition, this alloy does not have the pinkish-brown colour of normal bronze but is more like the colour of pure copper. This alloy is referred to here as 'more-or-less pure copper' (or 'copper' for short).

Fig.35 Boundaries for the alloy types defined from Figure 6.4

The 3-D surface chart (Figure 34) allows the identification of distinct and favoured alloy types (brass, bronze, gunmetal, and 'copper'). The drawing of boundaries between these four types is not that easy, however. Cluster analysis is of little help here in identifying the centres and boundaries of these types for a number of reasons. Most important is the fact that the centres of three of the alloy types rest against an axis. This makes the distribution of results around them hemispherical rather than spherical. Cluster analysis assumes that all its clusters are spherical (Baxter 1994: 155). Cluster analysis would not, for example, place the centre of a 'bronze' cluster near the tin axis (which is where a visual inspection of Figure 34 would suggest it should go). Ultimately, the placing of the boundaries for the four alloy types has been subjective and based largely on a visual inspection of Figure 34. The boundaries of the four alloys are shown on a 2-D plot of zinc and tin contents (Figure 35). The proportions of these alloy types can be shown more clearly on a barchart (Figure 36).

Fig.36 Barchart showing the proportions of Roman alloy types

Each of the four alloy types defined above can now be divided into two groups: leaded and unleaded. The amount of lead for this division is 1% (this reflects the difference between unleaded wrought alloys, and cast alloys [often leaded]).

The 3-D surface chart (Figure 34) cannot, unfortunately, show the relationships between three alloy elements simultaneously. In order to examine the relationship between zinc, tin and lead, Figure 34 is repeated for unleaded alloys (Figure 37) and for leaded alloys (Figure 38). The unleaded alloys contain a higher proportion of brass and 'copper'. This is probably because these two alloys are often used for wrought sheet and wire work where a leaded alloy would be inappropriate. The leaded alloys, however, have almost no brass or 'copper'. Bronze and gunmetal are the commonest alloy types for leaded alloys. The proportion of gunmetals is much lower for unleaded alloys than it is for leaded ones. If gunmetals are usually formed as a result of recycling of scrap metal (as discussed above), then lead would often have been added during recycling. The speculum used for mirrors appears in Figure 38 but not 37 - speculum always contains some lead.

Fig.37 Zinc and tin contents for unleaded Roman alloys

Fig.38 Zinc and tin contents for leaded Roman alloys

The 3-D charts showing the distribution of alloy types for leaded and unleaded alloys can be simplified into barcharts (Figure 39). Barcharts of this type also allow the alloys from different sites, or of different dates, to be compared with each other.

Fig.39 Barcharts showing leaded and unleaded alloy compositions

(compare with Table 6.2)

6.5 Comparisons with previous analyses

A large number of analyses of Roman copper alloys have been carried out but the collection of samples has usually been constrained in some way. Craddock's work (1975; forthcoming) has largely been restricted to those objects available from museum collections and concentrated on cast objects such as statuary, musical instruments, military equipment, etc.

Picon et al. (1966; 1967; 1968), and Beck et al. (1985), have examined a very large number of statues and statuettes from across France (the latter also examining a limited number of other objects (mostly casting waste and ex votos). Riederer has presented the results of the analysis of drop handles, needles, and other objects found in the Tiber during the late19th century (Riederer 1974a; 1974b; Riederer and Briese 1974; Laurenze and Riederer 1980). Few of these objects could be closely dated and the associated coins range from the 1st through to the 4th century AD. Bayley (1992; Bayley and Butcher forthcoming) has analysed a large number of brooches (mostly from southern England).

The results presented here are based on a wider survey of Roman artefacts and should give a more representative picture of copper metallurgy as a whole. Detailed comparisons with other programmes will tend to highlight differences related to typology and method of construction, rather than differences due to social and economic factors in different Roman provinces. Riederer's results, for instance, show very little use of leaded alloys (only 11% had more than 5% lead, compared to 28% of the samples analysed for this project). This could be interpreted as showing that Britain had greater access to lead than Italy. The lower lead levels in Riederer's objects are, however, a reflection of their method of construction - most needles and drop handles were wrought rather than cast. The work on Gallic statuary suggests relatively little use of brass in Gaul, but Craddock's work has shown that statuary tends to be made from leaded bronze rather than brass. While Bayley's analysis of brooches has concentrated on a single artefact type (Bayley 1992), the semi-quantitative analysis of a large number of everyday objects from Gorhambury has produced results (Bayley 1990; see also Bayley 1992 for a range of other sites) comparable with those presented here (i.e. roughly 20% of all objects are brasses).

As most previous analyses have been carried out in a typological framework the considerable body of comparative data is of more use in understanding the relationship between alloy composition and object typology.

6.6 Trace elements in Roman Alloys

The Roman alloys are relatively 'clean' compared to Iron Age alloys - generally the levels of impurities in the metal are lower in Roman alloys. This might be taken as indicative of the differences between the Iron Age and Roman societies as a whole. This would, however, impose modern expectations onto ancient metallurgy. Modern alloys are usually very 'clean' (especially if they are to be used for electrical work) and modern metallurgists have come to expect relatively pure metal. Even quite high levels of trace elements, however, are not detrimental to copper alloys when used for the manufacture of decorative castings. Moderate levels of trace elements (0.5% or possibly more) such as arsenic and antimony could lead to the formation of a range of highly attractive surface finishes (perhaps in conjunction with inverse segregation) including alloy and patina colours. The lower levels of trace elements in Roman alloys could only be achieved through the use of hotter or more oxidising smelting conditions, repeated smelting, or more rigorous fire-refining purification of the finished metal. All of these processes would tend to reduce the overall yield of copper as more copper would be lost to the slag at each step. Thus Roman copper smelting may have been in this sense less efficient than that in the Iron Age. The Iron Age smelting may have used a more 'appropriate technology' and the Roman period should not necessarily be seen as a technological advance on the Iron Age.

6.6.1 Iron

Figure 40 shows the distribution of iron contents for all Roman alloys. Iron was the only impurity which was regularly detected in all Roman alloys. Unfortunately, iron is possibly the least useful of the elements for use in provenancing copper alloys. Iron would be present in the ore, the flux and the furnace lining, and so it is not possible to relate iron levels in objects directly with ore sources. The level of iron may, however, tell us something about the smelting process used. Craddock & Meeks (1987) argue that the low levels of iron in early prehistoric copper alloys suggests the use of a smelting procedure which did not use a free-running slag (this would explain the lack of copper smelting slags in the Bronze Age). The iron levels in Roman alloys are considerably higher than in Bronze Age alloys and suggests that Roman copper smelting used a free-running tap slag method.

Fig.40 Distribution of Iron in all Roman alloys

The production of brass by the cementation process occurs under reducing conditions and the brass will tend to absorb any iron present. If the zinc ore used was the sulphide, then it would be pre-roasted before the cementation. This would remove some of the iron impurities present in the ore. Oxide ores would not need pre-roasting and so could contribute more iron to the brass. Figure 41 shows that there is no correlation between zinc and iron contents. It is likely that the major source of zinc ore for Roman brass was the sulphide rather that the carbonate ore.

Fig.41 Scatter chart showing zinc and iron contents of Roman alloys

6.6.2 Nickel

Nickel is chemically similar to copper and may be a more useful indicator of the ore source used. However, the nickel levels in Roman alloys are so low as to make it of little use (Figure 42). The nickel content of Roman alloys is lower than that for Iron Age alloys. While nickel was detected in a quarter of Iron Age alloys, it was detected in only 10% of Roman samples.

Fig.42 Distribution of Nickel in all Roman alloy

6.6.3 Arsenic

Arsenic was detected in only 15% of all Roman alloys (Figure 43). This contrasts strongly with the results for the Iron Age, where arsenic levels are much higher (arsenic was detected in 62% of all Iron Age alloys).

Fig.43 Distribution of Arsenic in all Roman alloys

6.7 Summary

Roman copper alloys were manufactured through the addition of zinc, tin and/or lead to copper. The presence of zinc is often seen as important because it first appears in copper alloys on a routine basis in the Roman period. The results presented above, however, suggest that it was not the most important of the three alloying elements. Tin bronze (with varying amounts of zinc and/or lead) still constituted the most ubiquitous alloy type.

The distribution of zinc and tin contents (Figure 34) show three distinct peaks: brass, bronze and copper. These three alloy types were probably those which would have been available to smiths. The inverse relationship between zinc and tin suggests that scrap bronze and brass were often mixed together before remelting. Most gunmetals may have been formed in this way. The absence of low zinc brasses shows that brass was rarely recycled on its own. The failure to recycle brass on its own is curious and indicates the possible use of recipes by Roman smiths.

Trace element levels in Roman copper alloys are very low and it is unlikely that they will be of much use in provenancing the ore sources used. They do illustrate, however, some of the differences between prehistoric and Roman smelting and alloying. Most striking is the higher arsenic levels in Iron Age alloys compared to Roman ones.

© Internet Archaeology
Last updated: Thu Apr 3 1997