Archaeometry ÂŒ Analyzing the Cultural Heritage

ART AND CHEMICAL SCIENCES 915

CHIMIA 2001,55, No. 11

Chimia 55 (2001) 915-922 © Schweizerische Chemische Gesellschaft ISSN 0009-4293

Archaeometry - Analyzing the Cultural Heritage

Willem B. Stern*

Abstract: Archaeometry is an interdisciplinary field of research, where the studied object stems from archaeological, cultural or historical domains whereas the approach of investigation is based on (natural) science. In the present overview the main focus is on case studies and non-destructive techniques. X-ray fluorescence and diffraction studies on ceramics, coin alloys, glass, man-made and natural gemstones, and objects of fine art prove that material analysis may contribute to insights in questions of authenticity and origin, but also in ancient technologies e.g. of coinage. Keywords: Ancient glass· Archaeometry . Authenticity in art . Non-destructive analysis' Numismatics

Archaeometry is a relatively recent term. The objects of investigation cover a the analysis itself cause detectable altera- It came into use around 1950 in England, wide range from alloys, over bone, ce- tions or long-lasting radioactivity of the and soon infiltrated many languages. ramics and coins, paper, pigments and studied object. In general, no (sub-) sam- Periodicals bear the term, like 'Archaeo- postage stamps to teeth and textiles. Be- pling is allowed, nor any polishing or metry' since 1958 in the UK, 'Berliner cause many objects represent not only scratching of the object's surface. Beitriige fiir Archiiometrie' since 1976, cultural, but commercial value as well, The analytical techniques in the fol- or 'Revue d'Archeometrie' in France investigations are preferably performed lowing case studies belong to the group since 1977. Nevertheless, it has not been by non-destructive methods. This was of instrumental analysis and deal with included in many encyclopedias or dic- not possible in the early days of analyti- chemical aspects of solids, but also with tionaries. It is missing e.g. in Collins/ cal chemistry, when e.g. Klapproth exe- their physical-mineralogical aspects. x- Pons (1997), Wahrig (1989), Webster's cuted the first analyses of Roman glass ray fluorescence (XRF) spectrometry (1993). and Greek silver coins in the late 18th performs an elemental analysis of rough- Archaeometry is an interdisciplinary century, a 150 years before the term Ar- ly 95% of all elements in the periodic topic, where the studied object stems chaeometry came into use. system, whilst X-ray diffractometry from historical science or the cultural do- For certain material groups like ce- (XRD) detects elemental assemblages main, but the method of investigation is ramic or glass shards, conventional de- provided that they represent physically from (natural) science. Meaning in the structive methods of chemical and physi- ordered compounds or minerals. beginning merely 'measuring/determin- cal analysis are still used in practice, Both aspects of solid matter are of im- ing the age of a historical object by when the mass of available shards allows portance in archaeometry: a Babylonian physical methods like isotope dating' it consumption of a selected quantity. cylinder seal described as 'white stone' adopted later a more general meaning of The following case studies, however, may chemically consist of calcium as 'performing scientific measurements on deal solely with the challenging task of main component besides carbon and oxy- objects of art, culture and history', or non-destructive analysis. Its aim is not gen, which suggests a compound of shortly 'analyzing the cultural heritage'. only the correct labeling of e.g. museum CaC03, and excludes e.g. gypsum (con- exhibits - alabaster is not onyx, brass not taining also sulfur) or bone (containing bronze, silica not necessarily quartz, a phosphorous). But the chemical analysis white pigment not necessarily chalk, does not tell us whether the sea] is made limestone not marble etc. - but also more of chalk (chemically CaC03, minera]ogi- complex targets like questions of authen- cal]y calcite) or of shell (chemically also ticity, or of ancient trade, to mention only CaC03, but mineralogically aragonite). two. A similar case may be a white pigment 'Correspondence: Prof. Dr. W.B. Stern Geochemical Laboratories The present outline is not intended to consisting chemically of titanium (and Institute for Mineralogy and Petrography be comprehensive. It rather reports ar- oxygen), hence being Ti02. But minera- Department of Earth Sciences University of Basel chaeometric fact-finding as it may occur logically different species with the same Bernouillistrasse 30 in a laboratory specialized in instrumen- composition occur, such as anatase, CH-4056 Basel tal analysis and materia] science. brookite and rutile. Natural Ti02 miner- Tel.: +41612673622 Fax: +41 61 2672881 Analytical methods are non-destruc- als are black or brown due to impurities, E-Mail: [email protected] tive when neither preparation steps nor but two of them have been synthesized ART AND CHEMICAL SCIENCES 916

CHIMIA 2001, 55, NO.l1 and commercialized in known, but differ- Crystallography-Mlnoralogy Chemical Micro· end Bulk-Analysls of Solida ent time spans, and belong today to the most widespread white pigments. Hence, the chemical proof of titanium white may 2000 represent a first hint in a question of authenticity, but the mineralogical proof of either anatase or rutile gives further insight in the question of dating. 1990 XRF and XRD are based upon the same physical principles, namely Bragg's equation and Moseley's law of 1914. For a deeper insight, the reader 1980 may consult the comprehensive textbook of Potts (1993) [I]. A schematic overview on X-ray applications used during 1970 the past 100 years is given in Fig. ]. Instrumental analysis is in general based upon an often unspoken, but essen- --- tial prerequisite: the object of investiga- 1960 tion has to be chemically and structurally homogeneous. Analytical data from heterogeneous objects will always be problematic when the investigated area, vol- 1950 ume or mass is smaller than the entire object. The analytical quality may be out- standing, the data nevertheless meaning- less because of lacking representativity. 1940 The topic seems simple but is a perma- nent point of discussion between analysts and archaeologists. 1930

Case Studies

1. Terra Sigillata - Is the Red Gloss 1920 a Clue to Production Origin?

Terra sigillata is one example of in- dustrial large-scale pottery production in 1910 Roman time. Terra sigillata was manufactured at different sites in Italy but also in what is today France and elsewhere. 1900 The manufacturers can, in favorable cases, be identified by stamps of origin on the underside of ceramic bowls, dishes, plates. Sometimes not only the produc- Fig. 1. The application of X-rays in material science over the past 100 years. Although X-ray tion place is identified but also the name diffraction (XRD, blue boxes) and X-ray fluorescence (XFA, red boxes) have a relatively long of the craftsman involved. history, their worldwide application only became important after the second world war. XFA with Terra sigillata is a quite spectacular its wave-length dispersive/sequential (WD-XFA) and its energy dispersive/simultaneous options type of pottery, easy to identify by means (ED-XFA) belong today to the most frequently used instrumental techniques for the analysis of inorganic solids in spite of the relatively high investment of 50'000 to 150'000 € [1]. of its characteristic orange-red gloss ('en- gobe', slip), a very fine grained and thin cover on the ceramic body. This cover is of them - may vary according to the ori- termined non-destructively by means of believed to consist of originally a clay- gin of the clay, and therefore from one Energy-Dispersive X-ray Fluorescence mineral enriched slip which turned by production plant to another geographi- Spectrometry (ED-XFA), and the con- oxidative firing to a glaze- or enamel- cally remote site. centrations of Ti02, MnO, Fe203, and like, quite scratch-resistant coating. One From four different production sites K20 found were plotted as elemental ra- may expect that its chemical composition twelve fragments from museum collec- tios in a XY graph (Fig. 2). Although the reflects roughly 'clay' with silica and tions were chosen, each fragment bearing projection points from the surface, and of alumina as the main detectable compo- a stamp of origin, and some also the mark the different shards from each origin do nents, and additional elements in variable of the potter Senti us, known to have scatter, the individual production centers amounts, like potassium, calcium, titani- worked in Italy and France as well. Arezzo (A, Italy), La Graufesenque um, manganese, iron etc. The relative The chemical composition from the (G, France), Lezoux (X, France) and proportions of these elements - or some pottery surface, i.e. the coating, was de- Lyon (L, France) occupy, however, dis- ART AND CHEMICAL SCIENCES 917

CHI MIA 2001.55. No. 11 tinctively different chemical fields. This less than 1 wt.% silver in the ore to 99% an improvement of the aspect. Depend- means that an individual unknown frag- silver which was used for coinage. Silver ing on different factors, the circulation of ment without a stamp of origin could coins from Greece became tender not coins leads to slower or faster wearing- most probably be attributed correctly to only in Greece, but in many places of the off of the noble layer. one of the four production sites, or possi- Mediterranean where Greek traders were Since blanching transforms an initial- bly to an additional not yet studied origin. active, notably in Spain and in the Rhone ly homogeneous alloy into a heterogene- valley then inhabited by Celtic tribes. ous, onion-shaped body, anyinstrumen- They used Greek coins since the 4th cen- tal analysis of such a coin will lead to Terra Sigillala tury BC and since the 3rd century BC possibly doubtful results. But also strike 100 imitated them by creating a 'barbarian' tests, used since antiquity, will possibly

Q N X Greek-based coinage which was pro- give erroneous information as long as the o • X X duced and in use until Augustan times. noble layer is still present, and sampled GO ( xx Gold and silver coins from Greece for quality control. II 0 Xfxx l were soon 'treated', i.e. legal tender was Over the past 15 years some 10000 2 10 • 1= U. ••. l* X II forged by individuals producing plated Celtic, Roman, Greek and Central-Asian l ~l i, "" S L L L coins. On top of bronze (seldom iron or coins have been analyzed by XFA [4-9]. L l L ".,. "A lead) flans a thin foil of gold or silver was Surface treatment, i.e. blanching, was fixed, and struck with a die. This tech- widely applied at all times as can be dem- 1 nique has allegedly never been used by onstrated by comparing the specific 0.5 0.6 0.7 0.8 0.9 1 FelK official mints. In ancient Greece the falsi- weight (p, measured experimentally) with fication was tested by official controllers the density calculated from surface anal- Fig. 2. Non-destructive analysis (ED-XFA) of slicing or cutting suspicious coins at least ysis. The density difference 'delta-p' is in- the red slip on Roman terra sigillata of known since the 4th century BC [3]. Although formative (Fig. 3). The data suggest also origin. Pottery from Arezzo (A, Italy), La Grau- Archimedes' density determination was that a second phenomenon influences the vesenque (G, France), Lezoux (X, France) and known at that time, it was not used in coin overall specific weight of a coin. Many Lyon (L, France) display distinct elemental ra- testing. The large density difference of Roman and Islamic coins display a low tios such that the products of a certain potter 3 Sentius (S) working in Lyon and Arezzo can be gold (19.3 g/cm ) and bronze (around experimental density with reference to 3 traced back by the respective chemical finger- 8.5 g/cm ) reveals plating easily, and is the analyzed alloy. prints [2]. today one of the easiest and fastest non- destructive methods to detect plated coins. Itis obvious also that plating caus- Stamps referring to the potter Sentius es a reduction in weight; if the coin size is (S) are present twice in the Lyon group, kept constant, lower weights could be i a ~ and seven times in the Arezzo 'popula- easily detected. 0 f f f " tion' (Fig. 2).The shards stem from exca- Since early times mint offices tried to " , 0 I vations in Augusta Raurica, and Vindo- debase coins in order to produce more " 0 I 0 '. I : nissa (Switzerland). Hence, by means of pieces out of a given mass of noble metal. 0: pol'l;lutAg·eu.llo ••••···" o"· ..j, I

'. I 'fingerprinting' the glossy surface we not 0 p."~ •••.. Changing the weight or metrics would P: Ag-platad bronze and copper' i I .,{ p p ".[ only conclude that sigillata of different have been noticed. But changing the sil- i s: t1ul'1a(;a-t,.atad Ag-Cu 1l1Joy1 I'" 0 [S"" I origin can indeed be chemically distin- ver or gold content by adding a small per- " p i IZl masstv., unlroatad Ag·Cu alloys I !S s I guished, but it confirms also that a certain centage of copper does not change color manlw, untrulld Cu.Sn alloys i ~ I I potter Sentius did work in Italy and or luster noticeably. If, however, larger .,, . , , France as well [2]. The chemical finger- amounts are added, the aspect of a coin Densily expcrimenl.ll [&km3J print of red gloss is indeed a possible clue will change. In order to reduce this effect, to the provenance of terra sigillata. debased (and even 'good') coins were Fig. 3. Non-destructive analyses (ED-, WD- surface-treated. In a first step the plan- XFA) and experimental density determination chets are annealed at several hundred of Roman silver coins, schematic overview. Binary alloys of silver and copper should 2. Legal Cheating and Criminal Coin centigrade in order to oxidize the copper display an Archimedes density between PAg Forgery present. Metallic copper at the surface is (10.5 g/cm3) and PCu (8.9 g/cm3) in case of transformed to black cuprite (CU20), solid, homogeneous alloys. Plated coins will It is common knowledge that the first with an increase in volume. Cuprite is be specifically lighter when the core consists of coins were minted in western Anatolia, dissolved in a second step by boiling the bronze or iron. The X-analysis reflects mainly today Turkey, in the 7th century Be. planchets in acid. This process is called the chemical composition of the surface and These early coins consisted of electron, a blanching in case of silver alloys. The the layer close to the surface. If the coin is homogeneous and solid, the density calculat- natural compound of gold and silver, oc- surface is now bright-silvery and consists ed from surface analysis and the experimental curring in placer deposits of Asia Minorl of nearly pure, but to a certain extent Archimedes density will be identical, their dif- Lydia. Not much later the first silver spongy, silver; the core remains un- ference Delta Rho equal to zero. Surface treat- coins were minted in continental Greece, changed. Striking the planchet in a next ed ('blanched' ,i.e. copper-depleted), silver plat- in Athens, most probably from ore of the step compacts the thin (thickness of a few ed bronze or iron, and porous coins will, how- famous Attica lead mines. The small sil- micrometers) spongy 'noble' layer. This ever, display a Delta Rho with positive values. ver content in lead ore was enriched by technique was in use in ancient and mod- Hence, the combination of two independent, non-destructive methods like experimental means of mineral dressing and early ern times as long as noble metal was density determination and XFA enables an large-scale cupellation technologies mak- coined, even when the fineness was fixed insight in details of coin production and tech- ing possible an enrichment of originally by law. Blanching acted in such a case as nology [4]. ART AND CHEMICAL SCIENCES 918

CHIMIA 2001, 55, No. 11

To take an extreme example: a binary Fig. 4. Destructive analysis of a cut and pol- alloy of silver-50 and copper-50 may be ished silver dirham by electron microprobe. surface treated which leads to a measured The upper imagedisplays the outer zone, compacted by striking, with its unmixing structures silver content greater than 50% and a of Ag94/Cu06 (white) and Ag06/Cu94 (gray). copper content accordingly lower than The lower micrograph shows the core with 50%, the density being theoretically roughly 24% of pores, black [7]. around 9.8 g/cm3. The experimentally determined density, however, may be as low as 8.9, which is lower than Pelt and hence impossible for a binary Ag-Cu alloy. One explanation would be to assume some kind of small-scale porosity of the coin body. Since surface treatment was common practice, the core of silver coins is shield- ed by the impact of striking, producing a thin and compact layer of noble metal on top of the debased alloy. Hence no direct microscopical evidence for small-scale porosity is possible. In seldom cases, however, destructive measures are allowed. By cutting the coin body, or by grinding off the surface and polishing the exposed core, analysis is possible by electron microprobe [7] (Fig. 4). The compacted rim zone is nearly free of pores, but the core displays a complex system of pores with a diameter of 2 to 5 gradually, but slightly later, between 350 common as in the case of silver. And in- micrometers and less. The total amount and 500 PH (around 950-1 lOa AD), and deed, improving the surface of ancient of pores is well above 20 vol. %. This po- not in all minting places (Fig. 5, bottom). gold alloys is somewhat more difficult rosity reduces the specific weight accord- Plated gold coins are practically ab- than blanching silver. Although ancient ingly. Porosity, however with channel- sent in the investigated set of 466 coins. gold is very often a binary alloy, as is like structure, was also found in Celtic Ei ther the collectors screened carefully, modern gold, an essential difference ex- gold [8]. The explanation for this porosi- or plating was rare in Central Asia be- ists. Ancient gold coins consist of gold ty is not yet understood, it may be con- tween 250 and 650 PH (around 850-1250 and silver resembling electron, whereas nected with degassing, expulsion of oxy- AD). From density considerations it seems modern gold coins consist of gold and gen during the smelting process. Its effect also that surface manipulations are not as copper. Modern gold can be treated much was probably welcome for it enables large coin volumes with a low amount of noble metal. It is clear, however, that the Silver weight of these porous coins is also reduced. Since the phenomenon has to be studied by destructive means, only a re- stricted number of coins can possibly be investigated, and biased results may ~ .. therefore result. ., ~ 40 Another insight taken from coin analysis is the development of fineness as a 20 function of minting time. Whether debasing at a given time interval took place in a •250 300 350 400 450 500 550 specific minting place only, or in a mere- Yo.r I pH I ly regional or larger scale, is easy to pin- Gold Coins point by surface analysis, provided that 100 sufficiently large, numismatically well- Fig. 5. Non-destructive analyses (ED-XFA) of documented coin series are available [9]. •• Central Asian silver (upper) and gold coins In case of Central Asia we clearly see (lower graph). The fineness in wt.% is plotted versus the minting year (years post Hedjra). It is strong debasing of silver coins in a rela- ~ •• tively short time span of roughly one cen- .,. evident that around 350 PH a debasing of silver ~ at all investigated mint offices begins, whereas tury affecting all investigated minting 4. the debasing of gold starts some 50 years later places between 300 and 400 PH (around 20 and is not observed for all mints. The debasing 900-1000 AD). Most of the investigated of silver coins coincides in time with the disap- silver (N = 2700, Fig. 5, top) is surface .. pearance of Central Asian coin import to North- treated, i.e. blanched, but plated pieces 250 :lOO 350 400 450 500 550 ern Europe, and the onset of silver mining in Year! pH I are rare. Gold coins also lost their fineness medieval Europe [8]. ART AND CHEMICAL SCIENCES 919

CHIMIA 2001, 55, No. 11

like silver in order to obtain a copper-free 3. Glass Making Far from the Sea tine due to the presence of this essential surface layer. Ancient gold, however, ingredient. Soda may also be produced cannot be 'improved' in the same way. Natural glass (volcanic glass, obsidi- on a technical scale by calcination of cer- Silver is removed only by a tedious ce- an) has been used for making tools, like tain sodium-incorporating plants like sal- mentation process, where long heating at knives, since prehistoric times. Because icornia, and subsequent washing and re- high temperatures between 850 to 1100 °C not too many sources of obsidian exist, fining of the ash. Because salicornia in presence of rock salt (NaCl) is essen- and the chemical composition of obsidian grows on many sea shores in the tidal tial. may vary, the chemical fingerprint of ob- zone, soda may be produced at many Some additional, fairly exotic facts sidian artifacts can be applied for the places along the coast. But would soda be were also observed. In certain narrow study of trade e.g. in the Mediterranean traded to e.g. central Europe for glass time spans additional elements occur stone age (Fig. 6). Neolithic obsidian ar- making? Or would the glass be transport- which are otherwise rare in silver coins. tifacts from Lipari (Aeolian Islands, Ita- ed to the customers in central Europe ei- Around 190 PH high mercury concentra- ly) do not most probably originate from ther as a final product, or as glass ingots? tions of up to 20 wt. % Hg are common the well known and spectacular, and to- Or would used glass be recycled? Or in Central Asian pieces, as is arsenic day exposed, lava flows (Forgia Vecchia, would soda glass only be produced along around 370 and 600 PH with peak con- Rocche Rosse) but from older, presently the coast, and in the interior potash glass, centrations up to 2 wt. % As. Extraction widely covered flows. The artifacts are since potash can be produced from many of silver from the raw ore by amalgama- also chemically different from examined kinds of wood, which is used for energy tion might have been applied; mercury obsidian ofVulcano, Sardinia and Anato- production in any case? was certainly available at the relevant time ]ia. Volcanic glass is, in genera], potassi- Different study groups are working and place. But it is not easy to understand um-rich and poor in calcium; the K/Ca worldwide on this topic; a comprehen- how mercury can survive the process of ratio is therefore high. The method of sive overview was presented by Bezboro- smelting. On the other hand surface im- plotting elemental ratios is selected be- dov that represents the state of the art of provement by fire silvering is possible, cause of two reasons: more variables can the late 1960s [II]. A later extensive but then it would be applied on debased be presented in a x/y graph and morphol- technological and historical overview alloys with accordingly low specific ogy effects (influence of surface geome- was compiled by Wedepohl in 1993 and weight - but the density of the mentioned tryon measured X-ray intensities) are re- 1998 [12][13]. 190 PH Hg-rich coins do not display con- duced. Early glass analyses tend to be incom- spicuous values. Also, the presence of ar- Man-made glass has been known for plete, since only the expected chemical senic in narrow time spans cannot be ex- at least 3.5 millennia. Glass pearls, vials, elements were searched for, the applied plained so far. flasks, bowls, plates were widely used in methods being sequential. This means The Centra] Asian coins stem from the Mediterranean area and in northern that before starting the analysis, an a pri- a museum collection where a quality Europe. Glass-making workshops, docu- ori assumption was made as to which screening reduced or eliminated obvious- mented today in archaeological excava- main chemical component would be im- ]y forged, i.e. plated, pieces. According- tions, were present all over the Roman portant, and these elements would then ly, published analyses e.g. from Roman empire, although certain ingredients for be analyzed in a sequential order. Since silver coins from British collections [10] glass making are found only at rare sites. the analytical procedure - gravimetric, or display plated coins only rarely. Again, There are three prerequisites for glass spectrometric - was destructive, repeti- silver hoards practically never contain making: a suitable silicate sand, a flux, tions to check the homogeneity of the in- plated coins. and thermal energy. Additional ingredi- vestigated object, and representativity of In contrast the percentage of plated ents may be colorants, bleaching agents, the obtained analysis were seldom per- coins stemming from provincial excava- additi ves for viscosity reduction etc. formed. Certain elements like phospho- tions is relatively high [4]. From 630 sil- Whilst suitable sand and wood are rela- rous and chlorine were often supposed to ver coins from the Augusta RauricaiSwit- tively ubiquitous, flux is not. Natural be absent and hence not analyzed. zerland excavations, 28% were plated. soda (Na2C03) is known from Egypt Ancient glass nowadays often dis- This may imply that plating was not nec- (Wadi Natrun) and it is believed that plays a heterogeneous structure. The sur- essarily considered forgery in imperial glass making started in Egypt or Pa]es- face and the adjacent parts are often Roman provincial money, but the insight of the numismatic special ists is called for this question. As a conclusion one may state that Obsidian surface manipulations were common practice in mints at all documented time 2! -,-- -, periods, and hence 'legal cheating' was usual business. Plating, however, was 20 v Fig. 6. Non-destructive analyses (ED-XFA) of •• considered to be against the law at least at ..•...•. obsidian from different sources. Key: A Anato- 15 LA certain times and in certain places, but lP lia; Iindustrial obsidian-like scories; K Mt. Kibo, lG IT Tanzania; M obsidian arrow, Mexico; S Sardi- possibly not everywhere and at all times. LT 10 Again, the opinion of numismatic spe- LT nia; V Vulcano, Eolian Islands/Sicily; LF Lipari Forgia Vecchia flow; LG Lipari Gabelotto flow; cialists is required. LP Lipari Pumice; LR Lipari Rocche Rosse

A flow. The crosses represent obsidian artifacts o ./-- --~ from Lipari with a chemical fingerprint different o 1D 15 20 from the currentlt exposed obsidian lava flows SilAI of Lipari (unpublished data). ART AND CHEMICAL SCIENCES 920

CHIMIA 2001,55, No. 11 leached out; alkalis are replaced by wa- constituent in many glass shards ana- (manufactured with pure alumina), the ter, which induces the well-known moth- lyzed to be of Roman and medieval age. first synthetic 'ruby' with spiked gallium er-of-pearl luster, and finally a putf-pas- Whether bone ash was added as a raw appeared on the market. Hence an impor- try shape of the glass surface when the material is not clear, nor is its influence tant criterion is, in principle, now no process has proceeded far enough. The on glass quality, but it possibly acts as a longer applicable. leached layer may vary from between a stabilizer, making the glass more resist- Great efforts have been made to cor- few micrometers to over 0.2 mm. Non- ant to chemical attack. Bone ash has relate the chemical fingerprint (trace ele- destructive bulk analysis of such leached probably been widely used as a raw mate- ment pattern) of certain gemstones with glass tends to give biased values for the rial for glassmaking, for white coatings their genesis and origin (mining site, alkali content: the surface analysis by x- on ceramic, and also for ceramics ('bone country of production). According to ray fluorescence spectrometry (XFA) china'). It may be added that bone was genesis, this fingerprint may vary consid- will possibly detect no alkalis at all be- most certainly also used for the produc- erably for one and the same mineral spe- cause of the low energies of Na and K tion of glue and of tallow. But not very cies. If the source of a given stone is spectral lines (1.0 and 3.3 KeY, respec- much is known from literature. known, and a statistically representative tively) and hence weak penetrative pow- set of specimens stemming from that er. Other techniques like neutron activa- source has been analyzed, and if, further- tion analysis [NAA] will reflect the entire 4. Decorative- and Gemstones more, as many as possible different glass body, the measured alkali content sources/origins are documented accord- then depends on the ratio between layer- The study of decorative- and gem- ingly, a specimen of unknown origin may and core thickness. stones is normally done by means of mi- be compared chemically, and possibly Tempering the glass prior to analysis croscopical techniques, usually on isolat- identified by means of the reference sets. may have a 'healing' effect by causing a ed crystals or gem stones, and not on set A specimen of unknown or questiona- redistribution of the alkalis, but the pro- stones. The identification is often possible origin may e.g. display an entirely cedure is risky and rarely allowed for ob- ble with considerable certainty, but the different fingerprint, and hence stem jects of value. chemical composition as a clue to origin from an additional, not yet documented Roman glass-making workshops with or authenticity also became important source. It may, however, display a similar kilns are known e.g. from the early impe- when appropriate non-destructive meth- trace element pattern as a documented rial site of A venticum-Avenches/Swit- ods were developed. case, e.g. mine NM, and therefore origi- zerland [14], where not only kilns were Colored minerals, stones, animal nate from NM. It may, in contrast, be dif- excavated, but also large quantities of products (antlers, bone, tuskers, but also ferent from mine NN, and this 'falsifica- broken glass. Shards of different colors coral) have been used for millennia for tion' is a strong criterion in question of were analyzed by ED-XFA, and some, decorative purposes. Their cultural and origin, whereas 'verification' (similarity for comparison, also by destructive WD- commercial importance makes them a with a reference group) is a weaker one, XF A. It was found that colorless glass cherished object of study. As far as ques- because not all mining sites worldwide contains very little iron and was hence tions of authenticity are concerned, not can be known or documented. manufactured with selected, pure raw much is published, however. If genuine From 1979 to 1999 well over a thou- materials, i.e. quartz- and feldspar sand. and falsified objects are clearly discerni- sand gemstones (ruby, sapphire, alexan- The latter is proven by the presence of ble by certain published methods, cun- drite) of known origin have been tested alumina. Red glass contains elevated ning readers will possibly try to avoid for their respecti ve trace element pattern concentrations of manganese, whilst mistakes in future, and get away with un- ('fingerprint') by non-destructive meth- green and blue glass has elevated CuO detected fraud. To mention an example: ods (ED-XFA), and their fingerprints and contents. The conspicuous opaque white Soon after publication [18] of the obser- variation documented. This data base is and blue shards are rich in antimony vation that gallium is an essential trace used for the study of origin and questions which acted as darkener. Phosphorous element in natural corundum (ruby, sap- of genuineness; an example is shown in was not present as a main component, phire), but missing in man-made stones Fig. 7. and hence bone ash was not used. Traces of flux material were not found in the excavated area, but their solubility makes preservation over longer periods of time unlikely. Our own investigations on Celtic, Ro- man, Greek and medieval glass confirm that Roman glass is principally sodium Synthetic Corundum glass and hence produced with soda as a 100 --, flux at all production sites, and never with potash. Roman glass contains in o o general an elevated percentage of chlo- 10 o v rine, suggesting the use of rock salt :> o 0 Do • • v 't (NaCI) as a 'softener', i.e. viscosity re- i= Fig. 7. Natural and man-made gemstones of ••0 ducing agent [15][16]. Medieval glass is different provenance may display distinct chemical fingerprints. V represents synthetic Ver- either potash or soda glass and occasion- neuil, D synthetic Douros, K synthetic Kashan ally contains chlorine [17]. corundum of ruby type, whilst the crosses 0.'1-- __ -.- _ _._-_ •.•••.••--.---1 A problematic observation is the pres- O,QQQ\ 0.001 0,01 0.1 to 100 tooo stand for natural corundum from Viet-Nam ence of phosphorous as a main chemical Cr/F. (sapphire), unpublished data. ART AND CHEMICAL SCIENCES 921

CHIMIA2001, 55, No. 11

5. The Question of Authenticity in Art One of the latest important cases of art presence of titanium by X-ray spectrome- forgery known to the public (Drew/My- try, but subsequently (on the same speci- One and the same masterpiece of art att) could have been rapidly and easily men) also the mineralogical speciation by may be labeled 'certainly false' by an art solved simply by examining the paint: means of X-ray diffraction (Fig. 8). historian, and 'possibly authentic' by a masterpieces by Braque, Chagall, Dubuf- Because microsampling is an estab- scientist; both opinions put down in a re- fet, and Giacometti were claimed to have lished method in forensic and police pro- port on behalf of a court of justice. The been painted in the 1920s, but were actu- cedures, a so-called 'criminal' adhesive authenticity of a masterpiece of art is by ally painted with acrylic paint, which tape was used to prepare specimen hold- tradition not merely based upon the sig- only came into use in the 1950s. These ers. Plexiglass frames, fitting into conven- nature and/or year of creation, but upon paintings are clear forgeries even if they tional X-ray fluorescence sample chang- its 'pedigree', i.e. its documented line of are signed and dated, and even when cun- ers, were prepared with a central opening owners back to the artist's workshop. If ning forgers have fabricated 'authentic' of 15 mm diameter, over which the tape such a pedigree is missing, the evidence pedigrees, and even when famous experts is placed (Fig. 8a). Plexiglass is used of authenticity has to rely either upon sty- of contemporary art have attested authen- because it is cheap, easy to handle and listic, iconographic and/or comparison ticity, relying on stylistic arguments. The virtually free of all interfering chemical studies, or else upon scientific arguments. Drew/Myatt case ended 1998/99 in Lon- elements. Specimens taken from refer- The stylistic approach is probably the don with a prison sentence for the rela- ence paintings, or standard materials for most common one, but also the most con- tively careless forger and his manager calibration are prepared the same way. troversial, because a high degree of per- [19], but only after a lengthy court case. The necessary sample mass may vary be- sonal 'feeling' enters the final judgment. The chance to find time-inconsistent tween less than 0.5 to 2 mg, spread over a More objective, but time-consuming is arguments is relatively small because surface of less than 2 cm2, and is suffi- the iconographic and comparative ap- many paints and pigments have been in cient to perform the elemental analysis proach, where authentic masterpieces of use for centuries, if not millennia [20]. If, for main components and some trace ele- the artist in question are studied for com- however, a masterpiece of art, supposed ments. The analytical procedure is strict- parison in art galleries, musea and re- to be painted, say, in the mid-19th centu- ly non-destructive, the specimen there- spected collections. The art historian will ry, contains so-called titanium white in fore ready for additional tests, e.g. by finally give a statement as to the authen- the original painting material, the entire XRD (Fig. 8b), for alternative measure- ticity of a questioned masterpiece in piece is a forgery - titanium white, chem- ments, or for storage. terms of 'certainly authentic', or 'certain- ically Ti02, only came into use as an art- The mineralogical, non-destructive ly false'. ist pigment after the first world war. This speciation of micros am pies is much more The scientific approach is relatively commercially produced titania consisted difficult and possible only when the min- seldom used, mainly for two reasons: mineralogically of anatase. Later com- erals to be identified are present as main first, the scientific analysis has to be non- mercial developments led, after the sec- components in the pigment (Fig. 8c). The destructive, which was not easily possi- ond war, to an improved titania, minera- diffraction runs are in any case extremely ble in the past, and second because the logically rutile. It is hence not only the time consuming (exposure time IOta 50 h), evidence for authenticity or non-authen- mere presence of titania of importance, therefore costly and not fit for large serial ticity may not be present in a given but also its mineralogical character when investigation. masterpiece of art. Physical age determi- the question of dating is concerned. This kind of microsampling for chem- nations by isotope analysis, or thermo- The presence of titania, anatase or ical fingerprinting has also become in the luminescence dating are not suitable for rutile, in artist pigments is obviously of meantime a routine procedure for archae- objects created after ca. 1500. The evi- importance. The practical test is relative- ological objects (alloys, bone, ceramics dence has hence to rely upon time-con- ly difficult when it has to be carried etc.), since it is easy to handle, but also to form or time-inconsistent parts of the out without any subsampling or invasive preserve for further additional examina- questioned object, like paint, pigment, means, i.e. non-destructively. Titania white tion and storage. canvas/cardboard/wood, working material. is furthermore extremely fine-grained If a scientist analyzes pigments in objects (particle size below 1 micrometer) and Received: September 19,200 I of art that were created allegedly in the can therefore not easily be studied with 19th century, but which contain pigments conventional optical light microscopy. which were commercialized after the sec- One approach has been to take subsam- ond world war only, his statement will pIes of a microscopical dimension and to be 'certainly false'. If the questioned ob- examine them by high-resolution scan- [1] P.J. Potts, 'A Handbook of Silicate Rock jects, however, do not contain any time- ning electron microscopy (SEM-tech- Analysis', Blackie, Glasgow, London, 1992, ISBN 0-216-93209-2. inconsistent materials whatsoever, he will niques). There exists, however, a prob- [2] P.J. Bailie, W. B. Stern, 'Non-Destructive state 'possibly authentic', since by lack lem of dimensionality: are a few grains of Surface Analysis of Roman Terra Sigilla- of arguments he has to plead in favor of at titania already a proof of an artist's pig- ta: A Possible Tool in Provenance Stud- least possible authenticity. The scientific ment? And how can the presence of ana- ies?', Archaeomelr)' 1984, 26, 62-68. approach is hence by nature able to falsi- tase or rutile be documented? Both min- [3] G. Stumpf, 'Ein Athenisches MUnzgesetz fy, but not to verify with same certainty. erals also occur in nature, although not as des 4.JH.v.Chr.', iB Numismatik lind In the example mentioned above, the white crystals, but as black or brown Geldgeschichle 1986, 26, 23-40. statement of the art historian 'certainly grains. [4] C.N. Zwicky-Sobczyk, W.B. Stern, 'x- Ray Fluorescence and Density Measure- false', and of the scientist 'possible au- Attempts have been made in our labo- ments on Surface-Treated Roman Silver thentic' are not contradictory, but have to ratory to perform a quasi-nondestructive, Coins' , Archaeomelry 1997, 39, 393-405. be explained to the court of justice in- yet representative sampling and to do not [5] A. Burkhardt, W.B. Stern, G. Helmig, volved. only a chemical analysis to examine the 'Keltische Mi.inzen aus Basel. Numisma- ART AND CHEMICAL SCIENCES 922

CHIMIA 2001,55, No. 11

a b

•••.•• .Jr.,i _

4 •

Fig. 8. Chemical fingerprinting in questions of authenticity in art: a) A set of microsamples taken from objects of fine art; the right two columns represent specimens as used for chemical investigation on pigments present, and inorganic compounds. The left column displays specimens which were first analyzed by non-destructive ED-XFA, and subsequently by destructive ICP-analysis (punched central part) for control measurements. The sample mass varies between less than 0.5 and 2 mg. Quite small quantities of roughly 100 micrograms can be used for element identification; b) where the element titanium is clearly visible. When mineral speciation is asked for, X-ray diffraction (Fig. 8c) is possible on microsamples on police tape, provided that a suitably large tape surface is covered by the pigment powder, and a long exposure time of 10 to 50 h is applied. The identification of the present compounds is performed by on-line data base research and reveals rutile (red bars), calcite (blue), quartz, whereas anatase (green bars) is proved to be absent.

tische und metallanalytische Untersuchun- Mitteleuropa: Zusammensetzung, Herstel- gen', Antiqua 1994, 25; Schweizerische lung, Rohstoffe', Nachrichten der Akade- Gesellschaft fUr Ur- und FrUhgeschichte mie der Wissenschaften G6ttingen, n. Basel, ISBN 3-908006-17-1. Mathematisch-Physikalische Klasse Nr. 1, [61 A. Burkhardt, 'Quantitative Methoden zur 1998. keltischen Numismatik am Beispiel der [14] H. Amrein, A. Burkhardt, W.B. Stern, MUnzfunde aus latenezeitlichen Siedlun- 'Analysen von Glasern aus der frUhkai- gen der Oberrheinregion', Dissertation serzeitlichen Glaswerkstatt von Aventi- Basel, Paul Haupt Verlag, Bern, 1998, cum (Schweiz)', Bulletin Association Pro ISBN 3-258-05858-X. Aventico 1995,37,189-201. [7] W.B, Stern, A. Burkhardt, S. Schmidt, 1. [15] N.R.U. GmUr-Brianza, 'Chemische Un- Kraut, 'Was leistet die chemische Analyse tersuchungen an r6mischen Glasern, Vin- in der Numismatik?', Festschrift Ludwig donissa und Eretria', Dissertation Basel, Berger, Forschungen in Augs/ 1998, 25, 1990. 235-241. [16] B. RUtti, L. Berger, S. FUnfschilling, W.B. [8] H.G. Bachmann, A. Burkhardt, R, Dehn, Stern, N, Spichtig, 'Die Zirkusbecher der W.B. Stern, 'New Aspects of Celtic Gold Schweiz. Eine Bestandesaufnahme', Jah- Coinage Production in Europe', Gold Bul- resbericht Gesellschaft Pro Vindonissa, le/in 1999,32,24-29. 1987. [9] W.B. Stern et al., 'Zur naturwissen- [17] R. Glatz, 'Hohlglasfunde der Region Biel. schaftlichen Analyse von Edelmetall- Zur Glasproduktion im Jura', Staatlicher mUnzen', VW-Projekt, 2001, in press. Lehrmittelverlag Bern, Paul Haupt, Bern, [10] D.R. Walker, 'The metrology of the Ro- 1991, ISBN 3-258-04562-3. man silver coinage', BAR Supplement Se- [18] H.A. Hanni, W.B. Stern, 'Ober die Bedeu- ries London 1976-1978,5,22, and 40. tung des Gallium-Nachweises in Ko- [II] M.A. Bezborodov, 'Chemie und Techno- runden', Zei/schrift Deutsche Gemmolo- logie der anti ken und mittelalterlichen gische Gesellschaft 1982, 31, 255-260. GHiser', Philip von Zabern, Mainz, 1975. [19] P. Landesman, 'A 20th Century Master [12] K.H. Wedepohl, 'Die Herstellung mittel- Scam', New York Times Magazine 1999, alterlicher und antiker Glaser', Abhandlun- July 16. gen Mathematisch-Naturwissenschaftli- [20] E. West FitzHugh, Ed., 'Artists' Pig- che Klasse Nr. 3, Franz Steiner Verlag, ments. A Handbook of Their History and Stuttgart, 1993. Characteristics', 1997, Vol. 1-3, Oxford [13] K.H. Wedepohl, 'Mittelalterliches Glas in University Press, ISBN 0-89468-256-3.