SECTION 3.5:CLAYDON
PIKE COPPER ALLOY METALLURGY by Peter Northover
INTRODUCTION
THE ALLOYS
Bronze
Gunmetals
Brass
CRUCIBLES
METALLOGRAPHY
CONCLUSION
BIBLIOGRAPHY
Table
1: Copper alloy metalwork from Claydon Pike, ordered by composition
Table
2: Copper alloy metalwork from Claydon Pike, ordered by class
Introduction
Thirty eight samples from twenty-nine copper-based alloy small finds
were taken for analysis from metal waste, scrap and vessel fragments recovered
from the Claydon Pike excavations. Of these samples one proved to be of
corrosion products only (9887) and was not analysed, but a second from
that item was successfully analysed. Of the remaining thirty-seven, fifteen
were from unstratified material but their compositions were generally
compatible with those of stratified material and can be assumed to be
of the same date range. There are three possible exceptions to this general
scheme which could be from much later date: the copper alloy lump (9881)
which has a low tin content and high arsenic and antimony, and the brass
lace-tag, (348) and strip (448) which latter two could be both made of
modern brasses.
The samples were collected in two tranches; initially twelve samples of
metal waste that showed evidence of melting were taken. More recently,
given the amount of fragments of copper alloy sheet among the small finds
it was decided to analyse a sample of these to discover whether they had
any connection with the other evidence of metalworking. At the same time
large sections of sheet bronze from a vessel or vessels were sampled,
to determine whether they might be a source of scrap for making smaller
wrought items. Finally, some of the waste samples were re-analysed because,
although the same analysis technique was used, the instrument first employed
had been replaced, and the data would link the two sets of results. Also,
such waste can be rather heterogeneous and the additional sampling can
give a measure of this.
Analysis in all cases was by electron probe microanalysis with wavelength
dispersive spectrometry.
The data have been assembled and sorted in two ways, by composition (Table
1) and Class (Table 2). For the former table the data were sorted experimentally
in both descending order of tin content and ascending order of zinc content.
They were then divided into low and medium/high tin bronzes, gunmetals
and brasses. The brasses were then divided into those containing tin and
those with tin only as a trace element.
Top of Page
The alloys
Only three samples contain alloy levels of lead : sheet fragment 536
(2.8%), binding or edge fragment 468 (6.0%) and waste 504 (3.4%). Lead
was increasingly used as an alloying addition in copper alloys destined
for cast products from the very late pre-Roman Iron Age (Northover 1992),
while earlier Iron Age copper alloys were unleaded. Thus the very small
number of leaded bronzes could be indicative of either date or function,
or both. All the other alloys analysed can be described as brass (Cu-Zn),
bronze (Cu-Sn), and ternary Cu-Sn-Zn alloys. These last would normally
be classified as gun-metals; they generally derive from the mixing of
bronze and brass scrap and in Roman times need not necessarily be deliberately
created alloys although some almost certainly were.
Bronze
The majority of bronze samples come from unleaded medium tin bronzes
with 9-14% tin; there are also four low tin bronzes with less than 5%
tin and one rather heterogeneous high tin bronze lump with 13.3% and 17.9%
tin in the two separate samples taken from it (9879).
Of the four low tin bronzes one sample differs in having much higher levels
of impurities than the others, for example 0.39% arsenic, 1.12% antimony,
0.28% silver and 0.15% nickel. Roman copper and bronze in general have
much lower levels of impurities because the raw copper was heavily oxidatively
refined (cf. The impurity values for the other objects analysed from Claydon
Pike). Outside the Roman period such an impurity pattern could date from
the Bronze Age to the late medieval, although in both those cases a leaded
bronze would be more usual. In the context of Claydon Pike another possibility,
a late pre-Roman Iron Age date in the 1st century BC or early 1st century
AD would both be possible and appropriate, while the absence of lead would
also be consistent.
Of the other three low tin bronzes two are fragments of sheet. Only one
(516), with 3.53% tin, has a datable context which is late Roman (Phase
4). A second, a piece of melting waste (9884) is very similar indeed in
composition with 4.31% tin and a slightly raised antimony impurity (0.13%
against 0.11% in 516) and, in the absence of other evidence, it would
be reasonable to see some link between the two in terms of metal use in
the settlement. The third is another sheet fragment (531) with a very
low tin content at 1.33% and a silver content of 0.99%; it also has a
zinc impurity of 0.16%. Another medium tin bronze piece of melting waste
(9881) also has raised silver contents at 0.35% and 0.70% in the two samples
taken, together with a zinc impurity (0.15% and 0.09% respectively). It
is entirely possible that scrap containing silver had been remelted but
it is also of interest to observe that some early medieval bronze from
Welsh princely sites such as Llangorse and Dinas Powys also had silver
and zinc as their principal impurities (Northover unpublished).
As noted above, the bulk of the bronzes contains tin between 9% and 14%,
while only one has an alloy addition of lead. Impurity levels are low
except for the melting waste already described (9881). The analyses come
from both classes of material - metalworking waste and sheet scrap including
the large vessel fragments (501). With such a small dataset and with virtually
no in situ evidence of metalworking facilities there is a danger of circular
arguments, but it is reasonable to conclude that unleaded bronze in the
form of sheet bronze scrap from large vessels was being melted but it
is not possible to decide from the present evidence whether sheet bronze
was also being produced. Even so, whatever was being made a wrought product
can be assumed from the bronze. In as much as there are dated contexts
the bronze spans much of the Roman occupation appearing on Phases 3 and
4.
Two other variations on the theme of bronze may be noted. The first consist
of two analyses from another piece of melting waste (9879) with 13.3/17.9%
tin and arsenic (0.12%/0.21%), antimony (0.13%/0.18%) and nickel (0.13%/014%)
impurities. The associated context is in Phase 3 but the composition is
probably more appropriate in the late Iron Age where such compositions
may be found in the cast bronze coinage. The second variation is bronze
with a modest zinc impurity (0.42%, 0.43%) which probably represents a
small amount of zinc-containing scarp being included in the melt. One
piece, a sheet fragment(456), was found in a post-hole of Phase 3, while
the second, a piece of melting waste (9880) was from a n undated context.
Top of Page
Gunmetals
This group of analyses is defined by gaps in the distribution of zinc
contents between 0.43% and 2.32% and between 8.67% and 11.53% - below
is bronze and above is brass. The group is further divided by a gap in
the range fo tin contents between 1.86% and 5.20%. Above that gap the
alloys are typical of many Roman gunmetals and consist of a mixture of
available bronzes and brasses. It is noticeable that they have higher
levels of the impurities iron, arsenic and antimony than the bronzes.
The iron is certainly brought to the alloy by the brass, being a by-product
of the cementation process. The other impurities could have been in the
copper used to make the brass component of the alloy - the brasses analysed
have a mixture of higher and lower antimony contents, and the brass ingot
(9875) is in the group with low values of these elements As with the bronze,
there is a rough balance between melting waste and scrap sheet. For the
one context where dating is available (for sheet fragment 503) the dating
is rather broad as Phases 3/4.
A group of 3 analyses from two objects (edge fragment 468 and melting
waste 9877) falls well under the main trend of tin against zinc with under
2% tin against 5-9% zinc when perhaps 5-6% tin would be more typical.
The dribble (9877) was found on a cobbled road surface (context 2335)
of Roman date, but not very precisely dated (AD 100-300). Even if it is
as early as the 2nd century it may still be a precursor of early Anglo-Saxon
copper alloys as seen, for example, in unpublished analyses of 5th-6th
century saucer brooches from the Oxford area (T. Dickinson pers. comm.).
Thus we have identified a second alloy group which potentially shows some
continuity into post-Roman times.
Top of Page
Brass
When the compositions were sorted in ascending order of zinc content
two items, both with uncertain contexts, were separated from the rest
by having, by some margin, the highest zinc contents, at 24.4% and 29.1%.
It is probable that they are modern in origin: both have iron contents
which are low for Roman cementation brasses in England (0.08% compared
with 0.25-0.50%), while the high nickel content of the lace tag (348)
and the high zinc content of the strip (448) would both be unusual in
a Roman or medieval context. They will not be discussed further.
A rare and significant find was a section of a plano-convex brass ingot
(9875), uncontaminated with any tin. This has an iron content within the
range to be expected of Roman cementation brass (0.41%) together with
0.64% lead; other impurities are at rather low levels but this is not
uncommon in Roman brass since Roman copper was often heavily oxidatively
refined. The zinc content was 20.6% Zn; this is very similar to two ingots
from the Seven Sisters, Glamorgan hoard and to a large variety of Roman
brass artefacts. The composition immediately prompts the question of how
much control Roman brass-makers had over the composition of their products.
Certainly higher zinc contents were produced as is shown by many brooch
analyses (Bayley 2000) but the very large dataset there now is of Romano-British
copper alloy compositions needs careful analysis to determine whether
among uncontaminated brasses there are specific clusters or simply a continuum
of zinc contents.
In the absence of any evidence for brass making at Claydon Pike, and given
that this is the only plain brass ingot fragment recovered, we can assume
that this cake was the ingot form in which brass was brought to the site
for further processing. It is not the only possible form since the Seven
Sisters hoard ingots have other shapes such as bar and pig. Given that
the ingot represents raw material brought directly to the site, can we
identify any by-products that might indicate how it was being used? Brass
was preferred for wrought products and the only other plain brass analysed,
a strip with a rivet-hole (461) is very similar indeed with 19.8% zinc,
0.08% tin and broadly similar impurities, although much less lead than
in the ingot. The strip has a late mid-1st century AD context (732) while
that of the ingot is uncertain. Both could represent an early post-Conquest
use of brass on the site.. Whether the strip itself was made on site is
another question: in its passage though the workshop the brass will have
flashed off some zinc during re-melting and can pick up slight traces
of other alloys from small debris and crucible residues, perhaps hinted
at by the tin content going from 0.01% in the ingot to 0.08% in the product.
This is not to say that the strip was necessarily made at Claydon Pike
but is certainly representative of what could be done.
The five other brass samples again comprise a mixture of metalworking
waste and fragments of wrought products. There is some evidence that the
Romans did make a choice of brass-based alloys with a few percent tin
to modify both colour and mechanical properties. Such alloys are still
used today for products like cheap jewellery and evidence from the finds
at Winterton Roman Villa currently being prepared for publication shows
that the Romans did the same. A wider range of alloys of this type could,
then, be expected from the analysis of other artefacts form the Cotswold
Water Park excavations.
Top of Page
Crucibles
Three sherds, presumably from crucibles, which had metalworking residues
attached, were submitted for analysis.
4788: A small fragment, possibly refractory fabric with a patch of dark
vitrified material adhering. X-ray flourescence (XRF) showed that the
only significant metallic peak was iron. The sherd was sectioned for metallographic
examination.
The polished section indicated that the vitrified material was in fact
completely glassy and had been deposited on the surface of the sherd and
had not been formed from it. There is no evidence to indicate the type
of high-temperature process involved.
4789: A sherd with the sand filled fabric and typical cross-section of
a crucible very much in the Iron Age tradition. The sherd included a portion
of the rim; much of the sherd was coated in red vitrified material, in
places a thick deposit corroding green. XRF analysis suggested that the
crucible had been used for melting bronze. Again a section was prepared
for metallographic study.
The section confirmed that the crucible had been used for melting bronze,
showing features identical to those observed in bronze-melting crucibles
from Hengistbury Head (Northover 1987). The matrix of the vitrified layer
is formed by the reaction of the crucible fabric at high temperatures
with metal oxides produced by the melting of bronze under oxidising conditions.
The vitrified material varies in colour from dark red to orange and from
translucent to opaque. Its appearance depends on the amount of copper
and tin oxides precipitated in it, the former as spheroids and dendrites,
the latter as large rhombs and needles of cassiterite. There are also
numerous prills of oxidised copper consisting of primary copper dendrites
surrounded by the Cu-Cu2O eutectic. Some of these structures also appear
in the residues from the manufacture of bronze by reducing cassiterite
in contact with molten copper in a crucible. However, some characteristic
features of this process are lacking, and it can be confirmed that the
crucible was used for melting bronze and that in the area of the sherd
conditions were highly oxidising. This process seems to be typical of
the Late Iron Age where triangular crucibles were effectively buried in
the fire and heated from the top giving slagged and vitrified rims and
often virtually unfired bases. Similarly oxidising melting probably continued
into the early Roman period: this fragment has a Late Iron Age/Early Roman
context putting it in the 1st century AD.
4790: The base of a pottery vessel, not necessarily a refractory, covered
with what appear to be metallic residues. XRF analysis showed that metal
was present, a copper based alloy containing tin, zinc, and a little lead..
The presence of an alloy containing zinc is consistent with the Roman
date of the context.
A metallographic section showed that the residues contained prills of
a copper based alloy in what may have been a black glassy matrix; however
the whole is so corroded that it is impossible to be certain. Whatever
the matrix, it has not been formed by reaction with the vessel fabric.
Either waste metal had been poured into the vessel at some time or it
had been used for melting a leaded gun-metal, perhaps only on one occasion.
The melting point of such an alloy might well have been low enough for
no visible vitrification of the vessel fabric to have occurred.
Top of Page
Metallography
After analysis the second tranche of samples was examined metallographically
with a primary objective of finding any patterns in the structure of either
the melting waste or the working of the copper alloy sheet. The data are
presented in tabular form in Table 3. In the end the conclusions were
quite limited but not without interest. Irrespective of alloy the copper
alloy sheet had been worked in a very consistent manner leaving it with
a fully recrystallised grain structure with a very fine grain size (<10?m)
and some residual cold work, often with increased deformation at the surface
from the final planishing of the sheet. The melting waste exhibited its
own characteristics in that very little was in an absolutely as-cast state
but had had a more complex thermal history, either being slowly cooled
or reheated by falling in a hearth or at some stage after solidification
being very rapidly cooled. Not much more can be said with the waste being
so scattered and removed from its original metallurgical context.
Top of Page
Conclusions
Overall, the recovery of copper alloy scrap and metallurgical waste from
Claydon Pike/Cotswold Water Park is very limited. There is certainly no
evidence that copper alloy metallurgy was ever more than a minor and episodic
component of the local metal economy. While episodes of metalworking could
well be scattered throughout the Roman occupation it seems to be concentrated
mainly in an early phase around the time of the Roman conquest, and a
very late phase prefiguring the metals used in post-Roman times. .
It is extremely unlikely that any alloys were actually manufactured from
their component elements at Claydon Pike but the possibility cannot be
ruled out on the data available. The piece of a brass ingot of course
suggests one way in which metal came to the site. Bronze was certainly
another input in the form of mixed scrap .Certainly some of the scrap
was in the form of sheet bronze and other copper alloy vessels and some
of the output could also have been sheet, although other wrought products
are likely. With the lack of leaded bronze it would seem unlikely that
objects that were normally produced as castings would be made there.
Top of Page
Bibliography
Bayley, J, 2000: The production of brass in antiquity with particular
reference to Roman Britain, in P.T. Craddock, ed., 2000 years of zinc
and brass, (London: British Museum Occasional Paper, 50), 7-26
Northover, J P, 1987: Non-ferrous metallurgy, in B.W. Cunliffe, Hengistbury
Head, Dorset, Vol. 1, The prehistoric and Roman settlement, 3500 BC-AD
500, (Oxford: OUCA Monograph, 13), 1987, 186-196, Fiche 7: A3-12, B1-4
Northover, J P, 1992: Materials Issues in the Celtic coinage, in M. Mays,
ed., Celtic coinage: Britain and beyond, Proceedings of the 11th Oxford
Symposium on coinage and monetary history, (Oxford: British Archaeological
Reports, 222), 1992, 235-300
Top of page
|