A comparison between 'total petrography' and geochemistry using portable X-ray fluorescence as provenancing tools for some Midlands axe-heads.

R.A.Ixer1, O.Williams-Thorpe2, R.E.Bevins3 and A.D.Chambers1

(1) School of Earth Sciences, University of Birmingham, Edgbaston, B15 2TT
(2) Department of Earth Sciences, The Open University, Milton Keynes, MK6 7AA, UK
(3) Department of Geology, National Museums and Galleries of Wales, Cardiff, CF10 3NP, UK

Twelve axe-heads and fragments from a small area of north Staffordshire have been grouped and classified macroscopically with a hand lens, microscopically in transmitted and reflected light ('total petrography') and geochemically using a non-destructive field-portable X-ray fluorescence spectrometer (PXRF). Both major methods were successful in grouping the axes, in matching two of these groupings with established axe Groups VI and VII and in identifying the coarse-grained ash and porphyritic lava as coming from different and separate sources.

Petrographically the axes fall into a group of four tuffs belonging to the Lake District Group VI axes; five microdiorite axes that are very close to the Welsh, Graig Lwyd Group VII axes; two porphyritic, andesite lavas belonging to no known axe-group but probably from the Lake District; and a single, coarse-grained ash that, despite gross petrographical similarities, is not a member of the Midlands axe-group Group XX and also probably from the Lake District.

Geochemically the axes could be divided into five macroscopic-geochemical groups. Two groups of four axes can be matched with rocks from Graig Lwyd (Group VII), and with intermediate composition rocks from the Borrowdale Volcanic Group (taken as equivalents of Group VI). Of the single axes, one may also be Group VII; one (a coarse-grained ash) is chemically dissimilar to Group XX axes; and one an unsourced porphyritic lava possibly from North Wales or the Lake District.

A correct lithological identification followed by the accurate inclusion into an axe-group, are the first essential steps to be taken before any discussion of the archaeological significance of an axe is possible. The three standard approaches for undertaking lithological identifications are macroscopical (hand lens), microscopical (petrological microscope) and lithogeochemical (XRF, Atomic absorption spectophotometry AAS, Neutron activation analysis NAA, etc).

Although, like the Group IX porcellanite axes from Northern Ireland, it is possible to recognise and classify a few axe-groups macroscopically, based on surface features including colour, texture and grain size, most axes have been grouped using standard thin sections and transmitted light petrography. Indeed this has been the most successful and long-standing method in stone axe studies and has been the basis for the recognition of the thirty-five or so major lithological axe-groups. These groups, described by members of the Implement Petrology Committee (IPC), originally a specialist committee of the CBA, have been codified in 'Stone Axe Studies' and 'Stone Axe Studies Volume 2' (Clough and Cummins 1979; 1988; Davis 1997). More recently The Irish Stone Axe Project using all three identification methods has classified in excess of 13,500 plus axes retaining the established IPC axe-groups where appropriate (Mandal 1996; Cooney and Mandal 1998; Mandal et al this volume).

However, by their very nature, axes are required to take a good, uniform polish and make and retain a sharp edge and so many of the axe lithologies are fine-grained and homogenous and less than ideal for study by thin section petrography. In addition many of the axe-groups, especially those from 'highland Britain', comprise rocks that are macroscopically and microscopically difficult to identify. This is because they are composed of major amounts of clay and clay-like silicates or have suffered low-grade or very low-grade metamorphism leading to the alteration of their primary assemblages into fine-grained, intimate mixtures of similar looking silicates. As a consequence of this the rocks have been considered to be difficult and 'scruffy' and so were poorly described and little investigated by petrographers/petrologists. However in the last two decades detailed work, itself based on high powered optical and electron microscopy allied with equally careful small-scale geochemistry, has characterised many of these rocks. It is now clear that there are significant mineralogical variations between apparently similar lithologies and that these can be related to variations in primary lithogeochemistry and/or slight differences in metamorphic grade that in turn can be matched to specific geographical areas. For example, the primary geochemical character of an igneous rock, whether it is calc-alkaline or tholeiitic in affinity, can be determined from its primary mineral assemblage and associated textures. Calc-alkaline basalts and andesites are found within the Lake District volcanic rocks whereas the same lithologies are tholeiitic in the Welsh Basin (Bevins 1998). In a similar manner the presence and paragenesis of secondary calc-silicates and amphibole minerals found in many low-grade metamorphic rocks of the United Kingdom can be used to distinguish between different areas (Bevins and Robinson 1993; Bevins 1998; Robinson and Bevins 1999). Of particular relevance to stone axe studies is the relative abundance of epidote and actinolitic to hornblendic amphibole and the scarcity of prehnite and pumpellyite in metamorphosed Palaeozoic igneous rocks from the English Lake District compared with similar rocks from Wales (Bevins et al. 1984). Transmitted light petrography can only investigate translucent minerals not opaque ones and most rocks carry up to 5% opaques with their complex and often individual textural relationships. Ixer (1994) demonstrated that, for fine-grained and/or highly altered lithic artefacts, reflected light petrography, namely describing the mineralogy and textural relations between opaque minerals and especially iron-titanium oxides and common sulphides, was a powerful tool in axe provenance studies. He further suggested that even better results could be obtained by combining transmitted and reflected light microscopy into 'total petrography'. The merit of this approach has been demonstrated by its successful use in the Irish Stone Axe Project (Mandal 1996; 1997; Cooney and Mandal 1998).

Although petrography remains the most common method for the investigation of lithic materials, whole rock geochemistry, primarily using XRF and databases from the geological literature, has proved effective in classifying and recognising both lithic artefacts and the raw materials from which they are made and is a complementary approach. Stillman (1996) gives a good and concise overview of the role of geochemistry in lithic provenancing and both Mandal et al. (1997) and Williams-Thorpe et al. (1999) have shown its effectiveness in axe provenancing.

The benefits of combining 'total petrography' with lithogeochemistry to provenance lithics have been briefly discussed by Ixer (1996) but more fully by Mandal (1997). From a limited number of case studies, including unaltered basaltic millstones and metamorphosed dolerites (Stonehenge bluestone monoliths) it was found that by combining the results of both techniques detailed, outcrop-level provenancing is possible and that each technique acts as a check upon the other (Ixer 1994; 1996; 1997).

The above discussion has concentrated on the geological aspects/problems associated with axe-provenancing but there are a number of archaeological ones. The source outcrops for a number of the axe-groups have not been recognised and so they remain essentially unprovenanced. In addition approximately one third of all British axes that have been sectioned do not belong to an established axe-group and so are listed under broad lithological headings and are also unprovenanced; any patterns or groupings within them are lost in their numbers. Even those that are assigned to a group may, due to 'petrographic-drift', be incorrectly classified, for with time there has been a gradual broadening of the lithological definition of some of the axe-groups as seen by the erection of sub-groups or by use of such terms as 'GroupV-like'. The causes of 'petrographic-drift' include the discovery of additional factory sites on new outcrops, but more commonly is the result of different petrographers putting their own interpretation on the original descriptions and setting their own limits as to what can be encompassed in a group. Finally, the petrographical and geochemical methods that so far have been discussed generally require material to be removed from the artefact so damaging it. Although slicing of axes for thin sections is no longer standard practice and cosmetic reconstruction after coring reduces visual damage there remains a strong reluctance from museums and private owners to offer axes for detailed analysis other than by non-destructive means.

There is, therefore, a need to reassess the limits of a 'total petrography' based within the broader contexts of igneous and metamorphic petrology as a provenancing tool for axes and to determine if the extra data that are generated have a practical use. At the same time these petrographical data and results from them are used to determine the effectiveness of a modern, non-destructive, geochemical method, namely the portable XRF, in axe provenancing studies.

Twelve complete or fragments of axe-heads collected by D.Walters from surface find spots in the vicinity of the Weaver Hills, north-east of Wootton in north Staffordshire, were sent for identification to RAI. Most were found within a square kilometre of each other and since they constitute a random sample of axes they were chosen to be used in a comparison between macroscopic (hand lens) plus microscopic ('total petrography') and macroscopic plus non-destructive geochemistry (portable XRF) methods of analysis. Two members of the team (RAI and OWT) classified the axes macroscopically, REB, ADC and RAI petrographically described the axes (concentrating on metamorphic, igneous and opaque mineralogy respectively) and OWT produced and interpreted the geochemical data.

The archive containing the detailed descriptions of the axes is held by the Implement Petrology Group and can be obtained from RAI.

Petrographic and macroscopic study

Methods and results

A nine-millimetre core was removed from each axe and a slightly thick, thin section prepared from the core in the standard way. However, instead of a protective glass cover slip being placed onto the top surface of the section it was polished with progressively finer grade diamond pastes ( six, one and finishing with ¼ micron paste) until optically flat. The resulting polished thin section was suitable for either transmitted or reflected light microscopy. As a bonus, the optically flat surface makes it possible to distinguish between fine-grained silicates, only a few microns in size, by using a combination of their properties in transmitted and reflected light as suggested by Ixer (1990, 126). Notably polishing hardness and reflectance in reflected light combined with transmitted light properties make it possible to distinguish between clay minerals (fine-grained muscovite/sericite) and prehnite; between chloritic minerals and pumpellyite; and even between fine-grained quartz and feldspars and clinozoisite. In all three examples the latter mineral has a higher reflectance, is better polished and is an important and characteristic mineral in low-grade metamorphic assemblages. It is of note that a very similar discrimination can be obtained by using the SEM in back-scatter electron mode (Schiffman and Day 1999).

Petrographically, using macroscopic and microscopic methods, the twelve samples fall into four groups as detailed below:-
WP1/187, WL90/123, WP91/80, WP91/36
WP1/25, WP1/44, WP2/18, WP2/28, WP3/2
WP3/1, WP3/3

Epidotized Tuffs
The epidotized tuffs (WL90/123; WP1/187; WP91/36 and WP91/80), although they differ in detail are all fine-grained to very fine-grained, non-banded to banded, graded (WP1/187) andesitic tuffs. All are clastic and extensively epidotized, indeed epidote makes up between 5-20% of the rock, and they have no primary iron-titanium oxides but do carry secondary sulphides. In detail they are more varied; WP/187 and WP91/80 are banded and may be epiclastic tuffs, whilst WL90/123 and WP91/36 are probably ash fall/ ash-flow tuffs. Rhombic carbonate patches are present in WP91/36.

Altered, angular clasts and abundant small feldspar laths showing simple or multiple twinning, are set in a fine or very fine-grained (WP91/80) indeterminate matrix, containing chlorite (WP1/187). Feldspar clasts range in composition from albite to oligoclase and show extensive alteration to epidote and, to a lesser extent, 'white mica'. Other clasts now comprise very fine-grained mosaics of a low birefringence mineral or epidote. In WP1/187 very fine-grained basalt and vesicular lava clasts and rare, 'large' oligoclase crystals are present. Prehnite, pumpellyite and amphiboles were not recognised.

Primary iron-titanium oxides are absent but secondary sulphides include pyrite as the main sulphide in WL90/123 and WP1/187 and pyrrhotite± chalcopyrite, associated with epidote, in WP91/36 and WP91/80. The four tuffs are characterized by their clastic nature, their lack of opaque oxides, by the abundance of epidote and absence of prehnite and pumpellyite.

Comparisons with assigned Group VI axes from Staffordshire
Group VI is a broadly defined group of epidotized tuffs and ashes from the Lake District (Keiller et al. 1941), the petrography of which has been revised by Woolley (in Claris and Quartermaine 1989).

Twelve axes, assigned to Group VI, are recorded from Staffordshire in Stone Axe Studies 2 (Clough and Cummins 1988); indeed Group VI is the largest recorded group from the county. Transmitted light petrography for seven 'epidotized intermediate tuffs' (St 10/c; St 12/c; St 15/c; St 22/c; St 23/c; St 29/c and St 33/c) and one acid tuff (St 19/c), held in the Lapworth Museum Collection, Birmingham University, has been compared with that for the four tuff sections. There appears to be few differences between St19c (the acid tuff) and the other seven and so it was included in the comparison.

They show variations in grain size from very fine- to fine-grained, in the presence and degree of banding, from pronounced to absent, and in the amounts of epidote present. Specimen St 29/c, from the Manifold Valley and closest in find spot to the present axes, shows very strong banding and grain size variations as great as those seen throughout the rest of the group. Collectively the eight samples show rounded to angular clasts and common, small feldspar laths and epidote, much of which replaces feldspar-rich clasts. Some (St 10/c; St 22/c) show quartz crystals and St 12/c has carbonate as part of its alteration assemblage.

The degree of variation between the eight previously assigned axes means that the present four may also be classed as Group VI. The degree of epidotization is consistent with a Lake District origin (Bevins et al. 1984) and grain size variation is probably not a very useful criterion for provenancing this axe-group.

Macroscopically, WP1/25, WP2/18 and WP2/28 (a tight sub-group) and WP1/44 and WP3/2 comprise fine-grained, heavily altered rocks, with 1-2mm long, pale-coloured feldspars and dark coloured, often tabular, ferromagnesian minerals set in a fine-grained matrix.

Pale, zoned, twinned plagioclase phenocrysts, up to 2mm in length, are extensively altered to fine-grained epidote and prehnite (WP1/44). Smaller, tabular, ferromagnesian phenocrysts, up to 1mm in length, show a wider range of alteration minerals including much alteration to epidote, accompanied by pumpellyite and minor actinolite/hornblende (WP1/25) or by carbonates and sulphides (WP3/2). In WP2/28 euhedral ferromagnesian minerals are altered to soft, limonite-stained clay minerals and minor amounts of quartz.

The phenocrysts are present in a very fine-grained to fine-grained groundmass dominated by stubby crystals of altered plagioclase, rare to abundant, unaltered to altered clinopyroxene, orthopyroxene that has altered to chlorite, acicular apatite, altered titanomagnetite and ilmenite laths and quartz. Epidote, clinozoisite and carbonate have a localized distribution. Within the groundmass, late-stage, quartz-rich segregations occur, although only in trace amounts in WP3/2 and comprise subhedral quartz mosaics with unaltered to altered plagioclase, pumpellyite and radiating prehnite (WP1/25 and 2/28).

All samples carry discrete, equant titanomagnetite, often with oxidation-exsolution lamellae of ilmenite and discrete ilmenite laths in the groundmass. Both phases show extensive alteration to TiO2 minerals and titanite. Trace amounts of chromite are present in WP1/25 and WP2/28 and traces of chalcopyrite in WP2/18 accompanied by bornite in 2/28, by galena in WP1/44 and by pyrrhotite, galena and sphalerite in WP3/2.

WP1/44 and WP3/2 form a poorly defined sub-group based on a greater degree of lithological heterogeneity. Both have zoned plagioclase-ilmenite-epidote-chalcopyrite-rich areas but in addition WP3/2 has a vein-like epidote-carbonate-base metal sulphide-clinopyroxene segregation on its worked surface. Axes WP1/25, WP1/44, WP2/18, WP2/28 and WP3/2 are made from an intermediate lava/hypabyssal rock of microdiorite to microgranodiorite affinity. Petrographically they share many similarities with the microdiorites of Graig Lwyd, North Wales including having the same low-grade, sub-greenschist facies metamorphic assemblage.

Comparisons with assigned Group VII axes from Staffordshire.
Group VII is well-established and well-documented and was first defined by Keiller et al. (1941) and later described as an augite granophyre in Stone Axe Studies 2, (Clough and Cummins, 1988); however a better lithological definition would be a microdiorite (Ball and Merriman 1989, 11)). The group belongs to a series of axe-factory sites in the Penmenmawr area of North Wales that are associated with fine-grained or very fine-grained igneous rocks cropping out at Graig Lwyd, Dinas and Garreg Fawr. Despite a number of petrographical studies neither opaque nor sub-greenschist facies mineral assemblages have been documented in any detail in the archaeological literature to date.

Group VII is the second most common axe type in the Midlands (Shotton 1988). There are six Group VII axes with find spots in Staffordshire (Clough and Cummins 1988), namely St 11c, St 17c, St 20c, St 26c, St 30c and St 32c. These, now in the Lapworth Museum collection, plus three unpublished axes, St 38c, St 44c and St 45c, from Wooton, found very close to the five axes of this study, were compared petrographically to each other. All six designated axes have feldspar phenocrysts partially or totally replaced by epidote, often very fine-grained epidote, accompanied by prehnite and carbonate (St 11c) and chlorite (St 30c). Ferromagnesian phenocrysts show a more varied set of alteration products. Most have altered to chlorite accompanied by quartz and calcite (St 26c, St 30c) and epidote (St 11c, St 32c). In addition, in one slide (St 11c) chlorite, titanite and prehnite are present alongside epidote. Most axes retain unaltered clinopyroxene. Late-stage quartz is rare in St 11c but in all the other sections quartz is accompanied by prehnite and commonly pumpellyite (absent from St 32c) and calcite in St 17c and St 26c.

Axes St 38c, St 44c and St 45c share many characteristics with all the other axes of the group. All have plagioclase phenocrysts, now altered to epidote plus chlorite, or epidote with clinozoisite; in these intergrowths clinozoisite rims epidote. All have unaltered clinopyroxene but also ferromagnesian minerals that have altered to chlorite and limonitically stained chlorite plus, in St 38c, epidote, clinozoisite and pumpellyite. Large, glomeroporphyritic clusters of clinopyroxene and plagioclase are a feature of St 45c. All have late quartz, altered feldspar and radiating prehnite and clinozoisite associated with chlorite and pumpellyite (St 45c).

The present five axes, the six axes that have previously been assigned to Group VII and the additional three share enough primary igneous and secondary metamorphic petrographical characteristics to be included within a single, cohesive group, namely Group VII.

Porphyritic Andesitic Lavas
Macroscopically WP3/1 and WP3/3 are similar, with 3-4mm long feldspar and small, dark, ferromagnesian phenocrysts set in a pale matrix. In thin section although both are highly altered, with abundant epidote, white mica, and chlorite and, in WP3/1, pale green amphibole, they show significant differences.

In WP3/1 large, weakly zoned feldspars are altered to white mica and epidote within very thin feldspar rims. No ferromagnesian phenocrysts were encountered in the section but small masses of epidote within pale green amphibole and chlorite margins may represent them. The phenocrysts are set within an even-grained matrix comprising altered plagioclase laths, titanite replacing titanomagnetite, epidote, chlorite, amphibole and, locally, quartz mosaics. Pale blue-green amphibole (probably actinolite to hornblende in composition), with minor amounts of epidote and clinozoisite, form late-stage veinlets cross-cutting phenocrysts and groundmass. There are no primary iron-titanium oxides only a little pyrite and chalcopyrite.

In WP3/3 strongly oscillatory-zoned feldspar phenocrysts are extensively replaced by coarse-grained 'white mica' plus minor amounts of epidote and chlorite, the latter two minerals probably replacing trapped glass in the feldspar. A little relict feldspar lies close to the crystal edges. Euhedral ferromagnesian crystals, that were probably orthopyroxene, are now totally altered to zoned epidote with clear cores and yellow rims, accompanied by chlorite together with lesser amounts of clinozoisite and titanite. The groundmass comprises altered plagioclase laths, altered ferromagnesian minerals, epidote, chlorite and titanite. Titanomagnetite with ilmenite oxidation-exsolution lamellae is altered to titanite and discrete ilmenite laths are altered to fine-grained mixtures of TiO2 minerals, haematite and titanite.

Neither axe can be matched to each other or to any recognised axe-group. Although the two axes are altered porphyritic andesites/basaltic andesites they do not have the same lithology as shown by their 'total petrography'. They differ in their primary mineralogy; for in WP3/3 titanomagnetite and ilmenite laths are present but magnetite alone is present in WP3/1; they also differ in the degree and style of alteration of the feldspar phenocrysts. They differ in their metamorphic history for the presence of actinolite/hornblende in WP3/1 suggests it has undergone a contact metamorphism that is absent from WP3/3. However the oscillatory zoning of the plagioclase feldspars and the original presence of both ortho- and clinopyroxene are suggestive of calc-alkaline rocks for example the Borrowdale Volcanic Group of the English Lake District.

In Ireland approximately 240 'porphyritic' axes (not all are andesites or indeed porphyritic) have been recognised macroscopically, a number that represents less than 2% of the total number of Irish axes (Cooney and Mandal 1998). The number of porphyritic axes in Britain is unknown. If they occur with a similar frequency to their Irish equivalents and given that the two present examples are from different rocks then it suggests that there is not a single, undiscovered factory site but more likely that porphyritic axes represent the adventitious use of erratics. The geological memoir for the Burton, Rugeley and Uttoxeter area (Stevenson and Mitchell 1955) lists altered porphyritic andesites from Borrowdale in boulder clay and other porphyritic rocks from northern England in fluvio-glacial deposits so although these axes may be made from erratic rocks they have a 'local' origin.

Coarse-grained pyroclastic
WP1/236 is a banded, andesitic, crystal-lithic tuff with coarse- and fine-grained layers. Within the coarse layers clasts are up to 8mm in length.

The main crystals are feldspars that range from being unaltered andesine to highly altered and replaced by fine-grained epidote and white mica. Ferromagnesian crystals, probably originally clinopyroxene, are totally replaced by a fibrous, pale-green amphibole but retain zoning, simple and multiple twinning and a single cleavage whereas original orthopyroxene crystals are now pseudomorphed by chlorite. Rock clasts mainly comprise plagioclase and ferromagnesian phenocrysts in a very fine-grained matrix; medium-grained plagioclase-rich clasts are extensively altered to epidote plus minor amounts of clinozoisite and a green amphibole that is possibly actinolitic hornblende.

Chlorite, showing anomalous brown or, less often, blue interference colours, and a little carbonate are present as a groundmass. Circular areas (vesicles?) have epidote rims and chlorite-rich cores. Minor amounts of equant, cubic magnetite and lath-shaped ilmenite, the latter altered to pale-coloured TiO2 and mantled by titanite are present. Importantly no quartz, sulphides or trachyte clasts were recognised. The axe does not belong to any of the established axe-groups, and although it has some similarities with Group XX its petrology, and in particular its calc-alkaline character, strongly suggests that it is not a Midlands rock but comes from the English Lake District.

Comparisons with assigned Group XX axes.
Group XX is defined as epidotized, ashy grits from Charnwood Forest, Leicestershire (Shotton 1959). Despite cropping out in the Midlands at Charnwood Forest, and there being very similar rocks close to Nuneaton (Evans et al. 1968; Bridge et al. 1998), Group XX is not very well represented in Staffordshire, with only one axe recorded in Stone Axe Studies Volume 2 (Clough and Cummins 1988). However, subsequently three additional and so far undescribed axes, St 47/c, St 48/c and St 52/c have been found that belong to Group XX and these also were compared with WP1/236. The three axes conform with the definition of Group XX and are similar to the type specimen St 13/c. They vary in composition from being a lithic-crystal tuff to crystal-lithic tuff and in grain size from fine-grained to medium-grained.

Characteristically all have strained quartz (in St 52/c the quartz is unstrained), and plagioclase feldspar plus abundant trachytic and fine-grained porphyritic lava clasts although they lack ferromagnesian minerals. Epidote, chlorite and carbonate are alteration products and St 48c has significant amounts of pyrrhotite, chalcopyrite, pyrite and sphalerite.

Based upon its 'total petrography' WP1/236 is believed to be a pyroclastic rock from the Lake District. It differs from Group XX in a number of significant respects, it has abundant ferromagnesian crystals, now altered to green amphibole, and it lacks quartz and trachyte clasts. The presence of quartz has been used to recognise the Midlands volcaniclastics, including those of Group XX and to distinguish them from similar rocks cropping out in the Lake District (Shotton 1959). Davis (1984) and Fell and Davis (1988), however, urge a more cautious approach towards provenancing Group XX as they note that quartz-bearing rocks very like those of Group XX crop out in the Central Lake District.

Geochemistry using non-destructive field-portable XRF

Method and results
Eleven of the axes were analysed using a field-portable XRF (PXRF) spectrometer which allows the quantitative determination of a range of major and trace elements. The instrument, a Spectrace TN9000 which employs a high-resolution mercury (II) iodide detector, has recently been assessed for use in lithic studies by Williams-Thorpe et al. (1999), who showed it to be a useful tool in stone axe provenancing. The PXRF probe is held against the untreated/unmodified surface of an artefact during analysis, avoiding the need for destructive sampling. Indeed the artefact requires no preparation before being analysed. Since it is the surface layer of the artefact that is analysed (mainly 0.03 mm to 2 mm in thickness, to a maximum of 10 mm), PXRF is only useful for characterization of the bulk composition if the surface measured is reasonably fresh and unweathered, and representative of the bulk. The measured surface must also be fairly flat (relief of less than 1mm ideally, and certainly less than 3 mm (Potts et al. 1997a)), and at least 2.5 cm in diameter. One of the axe fragments, WP3/1, a porphyritic lava, was too small for measurement by this method.

Two analyses each were carried out on eight of the artefacts, and three analyses on one artefact (all non-overlapping, independent measurements). These measurements gave average axe compositions with standard deviations of the means (sd/square root of n) (reflecting sampling error) mainly between 2 and 10 %. This agrees with the sampling precision reported by Potts et al. (1997b) for rocks of similar grain sizes. Two axes were analysed only once, because most of the surface in each case was judged to be too uneven for reliable measurement. In all cases, the measured surfaces have relief of less than 1 mm. Surface weathering is present but slight, with weathered depths much less than 1 mm and only WP2/18 showing any signs of significant weathering.

Results for three major elements, and six trace elements, are given in Table 1. Previous work using the PXRF on lithic artefacts has identified those elements which are most likely to be chemically altered within surface layers and therefore different from bulk composition, and those elements more likely to reflect bulk composition (Williams-Thorpe et al. 1999; Jones and Williams-Thorpe, 2001). The more reliable trace elements include Zr, Sr and Ba (Williams-Thorpe et al. 1999, 236). Major elements tend to be least representative of bulk composition in weathered rocks and artefacts, but in the axes measured for this study one of the major elements, iron, shows reasonable comparison with the potential sources discussed below. This is probably because the axes only have thin weathering crusts.

Interpretation and provenancing comments
Table 1 shows in bold those elements that are most likely to be reliable indicators of bulk compositions as noted above. Since many major elements are excluded from this category, it is difficult to determine the precise rock type from PXRF on its own. However, the low-medium concentrations of iron and relatively high amounts of trace elements such as Zr suggest that the axes most probably represent igneous rocks of intermediate composition.

The axes can be divided into five groupings on the basis of their compositions, groupings that are illustrated on Figures 1 and 2, and supported by data in Table 1: these groupings are: -
WP1/187, WL90/123, WP91/80, WP91/36
WP1/25, WP1/44, WP2/18, WP2/28

Figure 1

Figure 2

Although WP1/236 appears to belong to the WP1/187 grouping on Figures 1 and 2, it can be distinguished from the other axes by its particularly low Y concentration (17 ppm, averaged to one decimal place from two closely similar measurements of 16.4 and 16.9).

These groupings are almost exactly the same as those determined from the petrographic examination described above. The WP1/187 group equates with the epidotized tuffs, the WP1/25 group with the microdiorites, WP3/3 with the porphyritic lava, and WP1/236 with the coarse pyroclastic. Axe WP3/2 was defined as a microdiorite using petrography, but it appears on Figures 1 and 2 to differ from the other microdiorites (the WP1/25 group), having higher Sr and Ba (though similar Fe). Examination of Table 1 shows that it also has lower K and Rb concentrations. Petrographically this axe-head was seen to have a carbonate-epidote-rich surface layer and this is believed to be responsible for the unusual geochemistry recorded by the XRF. High field strength (generally immobile) elements, including Zr, Y and Nb, are similar in WP3/2 to the other microdiorites (Table 1).

The axe compositions in Table 1 were compared with published data on the IPC axe-groups reported in Clough and Cummins (1988). Many of the Groups, such as the more basic/ultrabasic Cornish Groups, or the Irish Group IX porcellanites, are not feasible sources because they are very different rock types from the Midlands axes discussed here. Amongst the remaining Groups, Group VII was identified purely on the basis of its geochemistry as being a possible source for the WP1/25 grouping. Several analyses of outcrop material collected at the axe-manufacturing Graig Lwyd area are available for Group VII (unpublished analyses from John Durham, Open University), and these are plotted on Figures 1 and 2, with one example given in Table 1. The compositional range for the outcrop samples is smaller than for the axes, as would be expected because of the larger instrumental and sampling errors using the PXRF on the axes. But overall, the comparison between axes and outcrop support an allocation to this Group. The position of WP3/2 is ambiguous on the geochemistry but is very securely tied to Group VII by petrography. This implies that for some axes, including those most mineralogically (and therefore chemically) altered, geochemistry on its own is inadequate for characterization.

Source comparison for the WP1/187 grouping is more difficult. Initial examination of the geochemistry did not suggest any close source comparison, until PXRF analyses of other Group VI (thin sectioned) axes became available (unpublished data (OWT)). These axes, in the Guildhall Museum of Northampton and in the Ashmolean Museum at Oxford, appeared to be similar geochemically to the WP1/187 grouping. This led to a consideration of the published chemical data for the Borrowdale Volcanic Group (BVG), of which the source of Group VI forms a part. Unfortunately, there appear to be no analytical data specifically for the outcrops of Group VI (part of the Seathwaite Fells tuffs), so that this comparison had to be based on a much broader published database for the BVG, which includes a very large variation of petrographic and chemical types (Millward et al. 1978). Average analyses for intermediate BVG rocks are plotted on Figures 1 and 2, and are given in Table 1. These rocks do indeed show reasonable similarity with the WP1/187 group of axes, with higher Fe and Sr concentrations than the Graig Lwyd axes. However, they do not include the Borrowdale pyroclastic deposits that are also known to be very variable (Millward et al. 1978, 113). For example the average ignimbrite given by Millward et al. (1978, 107) is a dacite with 67 % SiO2 and 4.4 % Fe2O3, and so is not comparable with the Group VI tuffs which are basic-intermediate in composition. Therefore, in this case the geochemistry indicates only that the general characteristics of the BVG are not inconsistent with having been the source of the WP1/187 group of axes. A geochemical survey of the Group VI outcrops would enable a more certain interpretation of these axe data. Until such data are available, however, this comparison remains unsatisfactory.

WP3/3, one of the two porphyritic lavas identified petrographically, is a high Zr ?intermediate rock also characterized by high Ba (Table 1). The axe could not be likened to any of the IPC Groups, but could be from the Lake District or the Ordovician volcanics of north Wales (Millward et al. 1978 and Thorpe et al. 1993 respectively).

WP1/236 as noted above has gross petrographic similarities with Group XX, although it differs in detail. The geochemical analysis of WP1/236 was compared with unpublished analyses of four Group XX (IPC allocated) axes in the Northampton Guildhall Museum. WP1/236 has important differences, particularly in concentrations of the immobile elements Zr and Nb, and in Ba. Comparison with other Midlands, Charnian tuffs (Thorpe 1972) does not reveal a good match with WP1/236. Webb and Brown (1989, 103, table 7.3, Upwood) report a basic agglomerate that has some chemical similarities with, but is not exactly the same as, WP1/236; however, this agglomerate has so far been identified only in drill core and so is an unlikely source.


Both 'total petrography' and geochemistry have been successful in identifying groupings within the studied axes, and, except for WP3/2 that has an ambiguous geochemistry, these groupings, reached independently, are exactly the same. This is in line with previous studies that have reported high levels of agreement between provenancing using 'total petrography' and lithogeochemistry (Ixer 1994; 1996; 1997) In terms of allocating axes to IPC Groups or regional sources both methods broadly reached the same conclusions except for WP1/236.

However, each method has its own advantages and disadvantages. 'Total petrography' is certainly the better of the two methods for identifying rock types but requires destructive sampling. Thin section petrography shows that the axes comprise fine- to medium-grained, intermediate igneous rocks that have suffered low-grade metamorphism. It is possible to use igneous petrology (rather than just petrography) combined with detailed descriptions of low-grade mineral assemblages to discriminate between similar looking axes from different areas. In particular, oscillatory zoning in plagioclase and the presence of both clino- and orthoyroxenes in the porphyritic andesites and coarse-grained lithic tuff suggest calc-alkaline rather than tholeiitic volcanism. This, together with the presence of abundant epidote in the epidotized tuffs, the porphyritic lavas and the coarse-grained pyroclastic rock; secondary actinolitic to hornblendic amphiboles in the porphyritic andesites and coarse grained pyroclastic; and the absence of prehnite and pumpellyite in all these axes, suggests a Lake District origin. In a similar manner, the presence of prehnite and pumpellyite in the microdiorites suggests they are from the Welsh Basin and the pseudomorphed orthopyroxenes provide a good match with Graig Lwyd from North Wales.

The non-destructive nature of the PXRF geochemistry should be stressed. While this carries the corollary that the method is essentially a surface analysis technique except where an inside (broken) axe surface can be analysed, it does mean that data can be obtained without in any way adversely affecting the axe or indeed removal of the axe from its home. The method does, however, have limitations; these include the need for a fairly fresh (and flat) surface to analyse and the limited number of major elements that can be analysed by PXRF. A weathered surface is unlikely to give an analysis representative of bulk composition for all elements analysed, and gross differences from the bulk may sometimes be observed in such surfaces (Williams-Thorpe et al. 1999, 224-225). This problem is intractable because weathering effects are often rock- and environment-specific and very difficult to quantify or to compensate. It should be noted here that some archaeological sites associated with lithic tools constitute extreme geochemical environments where superficial alteration effects can be intense, the most notable examples being the 'supra-gossan' zones associated with ancient, metalliferous mines (Ixer, 1999).

Both methods depend on having a data source with which to compare the axes and, for choice, a dedicated one. For petrography, a large number of axe thin sections, previously allocated to Groups, are available for comparison with new sections, but there appear to be fewer slides from the source material/factory sites in archaeological collections. The problems of subjective interpretation and of petrographic 'drift' of Groups have already been noted, but the forthcoming Atlas of Stone Axe Petrography (Davis and Ixer eds. in prep.) will go a long way to redress these problems by providing a readily available visual and written description for, and definition of the petrographical limits to, each Group. For the PXRF, a pre-requisite for source allocation is the existence of a source geochemical database. While there is a wealth of published data on igneous provinces and sedimentary rocks in the UK, many of these data are not specifically on the axe factory sites or the exact putative source areas, as has been shown by the problems in finding Lake District Group VI comparisons. Thus, what is needed is an increase in the number of geochemical analyses of Group outcrops where these are known and, importantly, on in-situ outcrop material not on loose, waste or worked material.


Total petrography divided the twelve axes into four groupings that are in very broad agreement with the classification obtained by non-destructive PXRF geochemistry.

Group one, WP1/187, WL90/123, W91/80, WF91/36, comprises a number of epidotized tuffs from the English Lake District. Petrographically and geochemically they are assigned to an established axe-group, Group VI.

Group two, WP1/25, WP1/44, WP2/18, WP2/28, WP3/2, comprises microdiorites from North Wales. Petrographically and geochemically they are assigned to an established axe-group, Group VII. The assignment of WP3/2 into group two is less certain geochemically than petrographically.

Group three, WP3/3, WP3/1, comprises two different porphyritic andesites from the English Lake District. The axes come from different rocks and are only superficially similar. Both petrography and geochemistry (WP3/1 was too small to be analysed by PXRF) show them not to belong to any established axe-group.

Group four, WP1/236, is a coarse-grained andesitic tuff. Both petrography and geochemistry showed that the axe did not belong to any established axe-group and in particular is not a member of Group XX. Petrography strongly suggests an English Lake District source.

It is clear that whilst 'total petrography' perhaps remains the best single method for axe provenancing, geochemical provenancing using a portable XRF, although more limited in the lithologies it can investigate, gives comparable results. It may well be that its limitations are mitigated by the advantages of being faster and non-destructive and for axes and other archaeological lithics "if the technique becomes widespread …geochemical characterization may become the norm" (Stillman 1996).


OWT is grateful to R Kilworth and O Herbert of Thermo FI, Crawley, Sussex (formerly Thermo Electron and Thermo Unicam) for use of the portable XRF. Thanks are due to John Durham for permission to use his unpublished data on Graig Lwyd rocks. Rosie Ixer is thanked for managing the production of this paper and the Constantine XI Palaeologos Fund is thanked for travelling expenses.


Ball, T.K. & Merriman, R.J. 1989. The petrology and geochemistry of the Ordovician Llewelyn Volcanic Group, Snowdonia, North Wales. British Geological Survey Research Report SG/89/1.

Bevins, R.E. 1998. A review of the petrography of a collection of hammerstones from Alderley Edge. Unpublished report submitted to Manchester Museum, in relation to the Alderley Edge Landscape Project.

Bevins, R.E., Oliver, G.J.H. & Thomas, L.J. 1984. Low-grade metamorphism in the paratectonic Caledonides of the British Isles. In R.A. Gayer (ed), The Tectonic Evolution of the Caledonide-Appalachian Orogen, 57-79. Braunschweig/Wiesbaden: F.Vieweg and Sohn.

Bevins, R.E. & Robinson, D. 1993. Parageneses of Ordovician sub-greenschist to greenschist facies metabasites from Wales. U.K. European Journal of Mineralogy 5, 925-935.

Bridge, D. McC., Carney, J.N., Lawley, R.S. and Rushton, A.W.A. 1998. Geology of the country around Coventry and Nuneaton. Memoir of the British Geological Survey. Sheet 169 (England and Wales).

Claris, P. & Quartermaine, J. 1989. The Neolithic quarries and axe-factory sites at Great Langdale and Scafell Pike: a new field survey. Proceedings of the Prehistoric Society 55, 1-25.

Clough, T.H. & Cummins, W.A. (eds). 1979. Stone Axe Studies. CBA Research Report. 23. London: Council for British Archaeology.

Clough, T.H. & Cummins, W.A. (eds). 1988. Stone Axe Studies. Volume 2. CBA Research Report. 67. London: Council for British Archaeology.

Cooney, G. & Mandal, S. 1998. The Irish Stone Axe Project. Monograph 1. Bray: Wordwell.

Davis, R.V. 1984. Stone implements of Neolithic and Bronze Age date from the north-west English highland zone and their origins: an archaeological assessment of recent petrological and experimental work. Unpublished M.Phil thesis. University of Liverpool.

Davis, R.V. 1997. Implement petrology: 50 years of service to the archaeological community. In A. Sinclair, E. Slater & J. Gowlett (eds), Archaeological Sciences 1995, 1-4. Oxford: Oxbow Monograph, 64.

Davis, R.V. & Ixer, R.A. (in prep.) Petrographic Atlas of British Prehistoric Stone Implements. Council for British Archaeology.

Evans, A.M., Ford, T.D. & Allen, J.R.L. 1968. PreCambrian Rocks. In P.C. Sylvester-Bradley & T.D. Ford (eds), The Geology of the East Midlands, 1-19. Leicester: Leicester University Press.

Fell, C.I. & Davis, R.V. 1988. The petrological identification of stone implement from Cumbria. In T.H. Clough & W.A. Cummins (eds), Stone Axe Studies. Volume 2, 71-77. CBA research report, 67. London: Council for British Archaeology.

Fitton, J. G. 1972, The genetic significance of almandine-pyrope phenocrysts in the calc-alkaline Borrowdale Volcanic Group. Contributions to Mineralogy and Petrology 36, 231-248.

Ixer, R.A. 1990. Atlas of Opaque and Ore Minerals in their Associations. Milton Keynes: Open University Press.

Ixer, R.A. 1994. Does ore petrography have a practical role in the finger-printing of rocks? In N. Ashton & A. David (eds), Stories in Stone, 10-23. London: Lithics Studies Society Occasional Paper 4.

Ixer, R.A. 1996. Ore petrography and archaeological provenance. Mineralogical Society Bulletin 113, 17-19.

Ixer, R.A. 1997. Detailed provenancing of the Stonehenge dolerites using reflected light petrography: a return to the light. In A. Sinclair, E. Slater & J. Gowlett (eds), Archaeological Sciences 1995, 11-17. Oxford: Oxbow Monograph, 64.

Ixer, R.A. 1999. The Role of Ore Geology and Ores in the Archaeological Provenancing of Metals. In S.M.Young, M.Pollard, P.Budd & R.A.Ixer (eds), Metals in Antiquity, 43-52. British Archaeological Reports. 792.

Jones, M. C. & Williams-Thorpe, O. 2001. An illustration of the use of an atypicality index in provenancing British stone axes. Archaeometry 43, 1-16.

Keiller, A., Piggott, S. & Wallis, F.S. 1941. First report of the subcommittee of the South-western Federation of Museums and Art Galleries on the petrological identification of stone axes. Proceedings of the Prehistoric Society 7, 50-72.

Mandal, S. 1996. The petrology of the Irish stone axe. Unpublished Ph.D. thesis. Trinity College, Dublin.

Mandal, S. 1997. Striking the balance: the roles of petrology and geochemistry in stone axe studies in Ireland. Archaeometry 39, 289-308.

Mandal, S., Cooney, G., Meighan, I. & Jamison, D. 1997. Using geochemistry to interpret porcellanite stone axe production in Ireland. Journal of Archaeological Science, 24, 757-63.

Millward, D., Moseley, F. & Soper, N. J. 1978. The Eycott and Borrowdale volcanic rocks. In F. Moseley (ed.), The Geology of the Lake District, 99-120. Leeds: TheYorkshire Geological Society

Potts, P. J., Webb, P. C., Williams-Thorpe, O. & Kilworth, R. 1995. Analysis of silicate rocks using field-portable X-ray fluorescence instrumentation incorporating a mercury (II) iodide detector: a preliminary assessment of analytical performance. Analyst 120, 1273-1278.

Potts, P. J., Webb, P. C. & Williams-Thorpe, O. 1997(a). Investigation of a correction procedure for surface irregularity effects based on scatter peak intensities in the field analysis of geological and archaeological rock samples by portable X-ray fluorescence spectrometry. Journal of Analytical Atomic Spectroscopy 12, 769-776.

Potts, P. J., Williams-Thorpe, O. & Webb, P. C. 1997(b). The bulk analysis of silicate rocks by portable X-ray fluorescence: effects of sample mineralogy in relation to the size of the excited volume. Geostandards Newsletter 21, 29-41.

Robinson, D. & Bevins, R.E. 1999. Patterns of regional low-grade metamorphism in metabasites. In M. Frey & D. Robinson (eds), Low-grade Metamorphism, 143-168. Oxford: Blackwell Sciences.

Schiffman, P. & Day, H.W. 1999. Petrological methods for the study of very low-grade metabasites. In M. Frey & D. Robinson (eds), Low-grade metamorphism, 108-142. Oxford: Blackwell Sciences.

Shotton, F.W. 1959. New petrological groups based on axes from the West Midlands. Proceedings of the Prehistoric Society 25, 135-53.

Shotton, F.W. 1988. The petrological identification of stone implements from the west Midlands: third report. In T.H. Clough & W.A. Cummins (eds), Stone Axe Studies Volume 2, 49-51. CBA Research Report 67. London: Council for British Archaeology.

Stevenson, I.P. & Mitchell, G.H. 1955. Geology of the country between Burton on Trent, Rugeley and Uttoxeter. Memoirs of the Geological Survey of Great Britain. London: H.M.S.O.

Stillman, C.J. 1996. Using geochemistry in archaeology. Mineralogical Society Bulletin 113, 11-15.

Sutherland, D.S. (ed.) 1982. Igneous rocks of the British Isles. Chichester: John Wiley and Sons.

Thorpe, R. S. 1972. The geochemistry and correlation of the Warren House, the Uriconian and the Charnian volcanic rocks from the English Precambrian. Proceedings of the Geologists' Association 83, 269-285.

Thorpe, R. S., Leat, P. T., Mann, A. C., Howells, M. F., Reedman, A. J. & Campbell, S. D. G. 1993. Magmatic evolution of the Ordovician Snowdon volcanic centre, North Wales, (UK). Journal of Petrology. 34, 711-734.

Webb, P. C. & Brown, G. C. 1989. Geochemistry of pre-Mesozoic igneous rocks. In J. A. Plant & D. G. Jones (eds), Metallogenetic models and exploration criteria for buried carbonate-hosted ore deposits - a multidisciplinary study in eastern England, 95-121. London and Keyworth: The Institution of Mining and Metallurgy and the British Geological Survey.

Williams-Thorpe, O., Potts, P. J. & Webb, P. C. 1999. Field-portable non-destructive analysis of lithic archaeological samples by X-ray fluorescence instrumentation using a mercury iodide detector: comparison with wavelength-dispersive XRF and a case study in British stone axe provenancing. Journal of Archaeological Science 26, 215-237.

To be published by the Lithic Studies Society in:
Lithics in Action: papers from the Lithic Studies in the Year 2000 Conference:
edited by Elizabeth A. Walker, Francis F. Wenban-Smith and Frances Healy
as a Lithic Studies Monograph.

Return to Homepage