Chemical analysis of pottery is the study of either presence/absence or, more usually, quantitative data on the chemical content of a ceramic body. It is almost always used in a comparative study, normally to establish an artefact's source, but can also be used to study the technology of ceramic bodies and glazes and the use of ceramics in metallurgy or glass working.
NB These notes do not cover the application of organic chemistry to archaeology, which is now a very important branch of ceramic analysis.
Chemical analysis can be carried out using very small samples. Indeed, results can be obtained non-destructively using XRF. It is thus a less destructive technique than thin-sectioning, the principal method of study used in ceramic petrology.
Despite the ability to get results from extremely small samples, there are disadvantages, of which the most obvious is that chemical analysis of a heterogeneous body will give wildly varying results, depending on which particular rocks or minerals are present in the sample. This, and other potential sources of error (or, to be more accurate, variation) are discussed below.
XRF is a bulk characterization technique for the rapid, simultaneous, and non-destructive detection of all elements heavier than fluorine. The sample is irradiated with x-rays and re-emits x-rays characteristic of its composition. The measurement process itself is non-destructive and thus in theory could be used on a complete or displayable artefact.
The X-Ray beam can either be focussed to analyse a small area, in the order of 100 microns across, or defocussed to analyse a wider field. However, even in this mode the sampled area will be minute in comparison with that of other methods. This can be an advantage, in that a specific inclusion or feature can be studied, but for characterisation it is a disadvantage.
XRF is useful for identifying the colourants used on painted pottery and glazes. However, unless the machinery has been adapted it is not possible to analyse complete vessels or large fragments.
When measured in air the results are at best semi-quantitative but greater accuracy is obtained when measuring in a vacuum. However, producing the vacuum slows down the measuring process.
One way around the problem of variability and small sample area is to crush a sample of pot and produce a homogeneous pellet either by compression or fusing. However, the method is then no longer non-destructive, or quick, thus negating the major strengths of the method.
One unique feature of XRF is the ability to measure silica and there is therefore a case for using this method for analysing refractory clays.
Both ICP-AES and ICP-MS work by vapourising a sample of pottery dissolved in acid. This produces a stream of excited atoms which can either be measured through their mass (ICP-MS) or the wavelength of light (ICP-AES). Mass spectroscopy is much more powerful than Atomic Emission spectroscopy, but the measurement process is also up to four times as expensive.
ICP-AES analysis is the standard method of chemical analysis carried out on consultancy projects. The analysis takes place at Royal Holloway University of London using the NERC ICP-AES Facility under the supervision of Dr. Nick Walsh but I supervise the sampling strategy and the first stage of sample preparation.
The standard programme I use is the Silicate Rock and Mineral Analysis Programme, with the addition of lead.
Samples of British Museum Standard Pottery are included in large batches so as to allow comparison with data from the Department of Scientific Research at the British Museum and other laboratories.
ICP-MS is an extremely accurate method of measuring element concentrations in the parts-per-thousand and parts-per-million range and can not only measure elements but separate isotopes.
A large number of elements are measured by default and to a greater precision than with ICP-AES.
In comparison with ICP-AES the increased accuracy has to be balanced against the fact that the major elements, measured in percent, cannot be measured. For ceramics, this could be a big disadvantage depending on the objectives of the analysis.
Neutron Activation Analysis uses powdered samples of pottery, as with ICPS. These samples are irradiated in an atomic reactor and the radioactivity of the samples is measured in two stages, to measure both short-lived and longer-lived radioactive isotopes.
The choice of analytical method will depend on the budget and the exact nature of the problem to be addressed. Simple XRF is the quickest way of identifying colourants, paints and glazes. SRF is also the main method for determining silica content (although a good guess can be obtained by subtracting combined percentage of the major measured elements using ICP-AES from 100%).
For trace element analysis, the choice is between ICP-AES, ICP-MS and NAA. The number of elements measured is highest for ICP-MS, which is also the most accurate of the three methods. However, if identical results can be obtained from ICP-AES then this is by far the cheapest of the three methods.
For studying the metallurgical use of ceramics, as moulds, furnace lining and crucibles, one needs to measure gold, silver, zinc, copper and lead. Of these, none of the methods considered here measures all five as standard. There is, in any case, considerable doubt as to the validity of studying metal content using a bulk analytical technique and the preferred option would be to analyse metal blebs embedded in the ceramic surface using EDXRF (energy dispersive x-ray fluorescence) on a polished block or polished thin-section in a scanning electron microscope. Such analysis is expensive and can only be justified if research questions are clearly formulated that could only be answered through the use of this method.
All methods produce numerical data which can be analysed in a variety of ways.
My approach is to analyse separately the major elements, those elements which are likely to be mobile during burial (Ca, P, Na, Fe, Mn), metals (Cu, Zn, Ni, Pb and, if measured, Ag) and the remaining trace elements and rare earths.
I use Principle Components Analysis for each subset, calculating the maximum number of components (in my software this is 14, but it is rare to find any patterning below PC3). Pairwise plots can be produced to examine variations in the ratios of one element against another.
For each chemical, archaeological or petrological group of samples I calculate means and standard deviations for each element and use these mean values as comparanda for further studies.
Although geochemists are naturally concerned with the accuracy of their measurements, for most pottery the largest probable sources of error will be:
Some of these sources of error can be either removed or controlled through judicious sampling. Others can be addressed by the use of ceramic petrology alongside chemical analysis, which is the recommended standard procedure.
The distribution of elements varies in nature but there are few cases where the chemical content of a sample will lead to the identification of its source, unlike ceramic petrology. Furthermore, the obvious approach - accumulate a mountain of data from known production sites with which to compare the samples under investigation - is only as good as the archaeological evidence used to create the database. Chemical analysis is really testing similarities and whilst it can negate a suggested source it can only show that within the data compared there is no discernable difference between the unknown samples and the controls.
Furthermore, PCA produces different results depending on what data is submitted for analysis. Including standards which are very different from the samples under study will have the effect of minimising differences between similar but discrete groups. Consequently, it is best to use controls which are relevant, in other words ones that visually and petrologically have similar characteristics to the unknown samples.
Most ceramic vessels were probably used for the preparation or serving of food and drink. This can affect the non-organic chemistry of the vessel, in particular where acidic liquids have been in contact with the fabric. However, non-organic chemistry is a very blunt weapon with which to attack this problem.
Pottery vessels and other ceramics were, however, also used in a variety of industrial processes. Some of these processes may have led to the enrichment of the ceramic body. In particular metallurgical use, as moulds, crucibles, parts of furnace or kiln structures and so on. If such a use is suspected then it is best to tailor an appropriate analytical method for the specific project. Nevertheless, the incidence of copper and zinc is measured as standard with ICP-AES and lead and silver can be added at extra cost. Therefore, if an analytical programme is planned in any case then for a small extra cost the presence of metallurgical/industrial activity can be flagged.
© Alan Vince 1999