P. Wobrauschek, G. Halmetschlager, S. Zamini, C. Jakubonis, G. Trnka, M. Karwowski;...

Volume 29 Number 1 January-February 2000

ectrometr AN INTERNATIONAL JOURNAL

Special Millennium Issue on Cultural Heritage

11 0049-8246(200001 /02)29: 1; 1-4

@Q Guest Editor

Eva Selin Lindgren WILEY

X-RAY SPECTROMETRYX-Ray Spectrom. 29, 25–33 (2000)

Energy-Dispersive X-Ray Fluorescence Analysisof Celtic Glasses

P. Wobrauschek,1* G. Halmetschlager,1 S. Zamini,1 C. Jokubonis,1 G. Trnka 2 andM. Karwowski 2

1 Atomic Institute of the Austrian Universities, Stadionallee 2, 1020 Vienna, Austria2 Institute of Prehistory, University of Vienna, Franz-Kleingasse 1, 1190 Vienna, Austria

Celtic glasses of different colors, opaque or transparent, in the form of beautiful ornaments were charac-terized according to elemental composition by energy-dispersive x-ray fluorescence analysis. This analyticaltechnique is nearly non-destructive and was the method of choice for this project. The Celtic glasses wereloaned for the time of analysis by museums and private collections under the guarantee that no harm wouldbe done to the objects. The main goal was to establish a reliable analytical method to achieve qualita-tive and quantitative results within acceptable errors. The historical and physical background is given anddescribes the motivation for this task. Preliminary tests carried out before the series measurements startedare described and show that within a certain error range good results can be expected from samples of differ-ent shapes, with uneven or rough surfaces. Experimental conditions for the determination of light elementsand subsequent extension to medium- and high-Z elements are described. The results from the test studiesand for selected samples with various colors and shapes are presented. Copyright 2000 John Wiley &Sons, Ltd.

HISTORICAL BACKGROUND

The origin of the people called by the Greeks Kε��o�and by the RomansCeltæor Gallii is still not knownand remains very controversial. The oldest ancient writtensources mention the Celts somewhere in the region ofthe Upper Danube. However, it is certain that the Celtsbelonged to the large family of peoples speaking Indo-European languages.

Corresponding to the historical Celts, the La TeneCulture (ca 500–0BC) was composed of a distinct andhomogeneous cultural complex covering large areas ofEurope (Fig. 1). However, many Celtic tribes settled indifferent lands out of the range of the La Tene Culture.

Through their close connections with the Mediterraneancivilization, the Celts adopted many of the technologicaladvances of the other cultures. The Celtic artists createda characteristic decorative style, making works of art inalmost every class of manufactured article. During the lateLa Tene period, under the influences of Roman civiliza-tion, Celtic society displayed a strong social differentiationand probably was on the verge of transformation into aformal class structure and statehood.

Celtic domination in Europe ended soon after Cae-sar’s conquest. Concurrent with the defeat by the RomanEmpire, the Celts lost power not only to the Dacians in theLower Danube region but also to the Germans in the areaof Middle and Upper Danube and east of the Rhine. How-ever, the total annihilation of the Celts never occurred. TheGauls (Celts) lasted longer in Britain and Ireland. Those

* Correspondence to: P. Wobrauschek, Atomic Institute of the Aus-trian Universities, Stadionallee 2, 1020 Vienna, Austria.

regions have never been Romanized enough to destroy theCeltic culture, which still exists to this day.

CELTIC GLASS

The Celtic La Tene Culture, as the first one in Europe outof the Mediterranean region, achieved highly specializedglass manufacture. The specific Celtic forms of braceletsand ring-beads made of glass came into common use at thebeginning of the Middle La Tene (LT C) period. The art ofglass working in easternCeltica was certainly one of thecomponents of a general trend in which the eastern Celtictribes adopted new technologies during the second halfof the third centuryBC. This cultural development was anaftermath of the retreat movements of the historical Celticraids on Macedonia, Thrace and Greece.

Glass jewelry was a common product in the Celticworld and it seems that it was manufactured in manyworkshops in different places in the large area of theLa Tene Culture. Identification of glass workshops is notan easy matter for archeologists, and scholars making athorough study of the subject still do not agree as towhether the Celts produced the basic glass or whetherthey imported raw glass from Mediterranean sources.The Celtic craftsmen, who possessed a high degree oftechnical knowledge in metal-working, were certainly ableto produce glass.

The artifacts studied here represent Celtic glass-workfrom the Middle and Late La Tene periods (LT C–D, i.e.ca 250–50BC). This collection resulted from a stay ofCeltic tribes in the present area of Austria. The glasseshave been collected from many archeological sites of theLa Tene Culture in NE Austria. The tested collection

CCC 0049–8246/2000/010025–09 $17.50 Received 20 July 1999Copyright 2000 John Wiley & Sons, Ltd. Accepted 14 October 1999

26 P. WOBRAUSCHEKET AL.

Figure 1. Map of Europe with Celtic territory in 500–0 BC and the location of the research site.

consists already of more than 100 pieces, mostly braceletsand ring-beads, represented in the greater part only bysmall fragments.

The knowledge about the presence of specific chemicalelements which suggest the use of specific combinationsof raw materials in ancient glass helps answer archeo-logical and technological questions in different areas ofresearch: identification and location of the raw mate-rials; investigation of the ways in which glass quality(coloration, transparency, etc.) can be modified; chemi-cal characterization of glass products at presumed cen-ters of glass production; reconstruction of the paths ofdistribution of finished glass products from the glassproduction centers to the areas of glass ‘consumption;’comparison of the presence of the components in Celticglass with the composition of other ancient glass, tobe able to place them into an overall technologicalcontext.

The fact that it is possible to carry out non-destructiveanalyses opens up new possibilities for the study of objectsof archeological importance.

PHYSICAL BACKGROUND

The unique beauty and value of Celtic glasses requirean absolutely non-destructive analytical method for theirelemental analysis, and several analytical techniqueshave already been applied.1 Energy-dispersive x-rayfluorescence (EDXRF) analysis is the method of choicein this work.2 – 4 The properties of this technique, viz. alarge dynamic range in concentration, detectable elementsfrom Na to U, simultaneous multi-element capacity andshort measuring time from 100 to 1000 s, are well suitedto solve the analytical task. The elemental compositionof the colored glasses, the patterns on it and the opaqueor transparent parts can be correlated with the elementspresent. This has been done for glasses from other culturesalso.5 – 8 The homogeneity of the samples, the surfacequality and the geometric shape of the individual artifactswere considered and methods were tested to establishwhether the results from the measurements would besufficiently accurate to be interpreted. After a serious

Copyright 2000 John Wiley & Sons, Ltd. X-Ray Spectrom. 29, 25–33 (2000)

EDXRF ANALYSIS OF CELTIC GLASSES 27

testing period of the method using standard referenceglasses which were investigated in a variety of positionsand with slight geometric changes9 so as to resemble trueartifacts, the observed error was accepted to give dataready for further interpretation by statistical proceduresand historical understanding.

The experiments were separated into two sections,mainly using the optimized excitation conditions on theone hand for the light elements using the Rh L-line seriesto excite elements from Na to S and the continuum upto 15 keV to excite with sufficient accuracy elements upto Ti. The other set was subjected to a 35 kV excitationvoltage to quantify metals, earth metals, plus Cd, Sn andSb. The samples were measured in conventional geometrywith a 45° incidence angle to the sample surface and a 45°

emission angle to the detector at four positions across thesurfaces with a narrow 1 mm collimated beam to avoidthe influence of the geometric form of the surface. Withthe software package QXAS10 developed under the aus-pices of the IAEA the net peak areas were calculated andusing the fundamental parameter program the conversionof measured intensities into concentrations of the elementswas performed. All the data were used to obtain averagesand standard deviations. Most of these data were used asthe primary information on composition and correlationwith the colors of the bulk glass or ornaments. Statisti-cal interpretation will be necessary to obtain answers tothe many historical questions mentioned in the historicalintroduction.

EXPERIMENTAL AND RESULTS

The analyzer used was a Spectrace 5000 instrument man-ufactured by Tracor Northern. This spectrometer is anautomated EDXRF analyzer used for non-destructive ele-mental analysis of solids and liquids. The system includesthe features and flexibilities necessary for solving a widerange of analytical problems, and can be configured toaccommodate a variety of sample geometries.

A schematic arrangement of an EDXRF measurementis shown in Fig. 2. Sample excitation is effected with thedirect or direct filtered primary beam from a low-powerx-ray tube (17.5 W maximum rating). This technique pro-vides efficient excitation across the full analytical range ofthe instrument, from sodium to uranium. The simplicityof the x-ray optics provides a high degree of sensitivity,reproducibility and reliability.

Figure 2. Scheme of a typical EDXRF experimental set-up usedfor taking the analytical data.

The direct mode is selected in cases where a collimatedbeam is required or the sample is low in mass or foranalysis of low atomic number elements.

Filtering of the primary beam provides optimum peak-to-background ratios for a particular element range. Filter-ing effectively eliminates the background by transmittingonly x-ray energies above the spectral range of inter-est. Sensitivity is optimized by selecting a filter thatpasses only radiation near, but above, the element absorp-tion edge.

X-rays emitted by the sample are detected by a cryo-genically cooled, lithium-drifted silicon [Si(Li)] detector.This high resolution (155 eV/5.9 keV) solid-state deviceprovides the technology that is the basis for the wide appli-cability of EDXRF.

The Spectrace 5000 has a 10-position tray for 31 mmdiameter samples or sample cups. Atmosphere selectionallows three choices: air, helium or vacuum. Air is themost common mode of operation, allowing analyses for allelements having emission lines above 4.5 keV (or atomicnumber above 22, Ti).

Excitation is effected by an air-cooled x-ray tube withan Rh anode and 125µm Be window. This extremelythin tube exit window allows the emission of low-energycontinuum and also the intense Rh L-line series efficientlyexcites the light elements. The x-ray generator providesvoltages from 5 to 50 kV in 1 kV increments and a currentrange from 0 to 0.35 mA in 0.01 mA increments. A six-position filter wheel facilitates optimum filter adjustment.It includes Al, cellulose, Cu, two Pd positions and aposition with no filter. Collimators can vary in diameterbetween 1 and 10 mm. During the experiments with Celticglasses a 1 mm collimator was used to minimize samplegeometry effects.

The x-ray detector system includes an Si(Li) detector(30 mm2 area, 155 eV resolution for 5.9 keV) with a Bewindow of 12µm thickness, well suited for the detectionof light elements.

The experimental parameters for the measurementswere chosen as follows:

Voltage: 15 kV 35 kVCurrent: 0.35 mA 0.35 mAAtmosphere: Vacuum AirLivetime: 1000 s 1000 sFilter: No filter Al filter

The first set-up with 15 kV provides excellent K-linedetection and analysis of elements from sodium to tita-nium. The second opens up the possibility, because of thegood excitation conditions, of examining K-lines up to Sn.Elements with higher excitation energies for K-lines areanalyzed with the help of L- and M-lines.

Sample holder

In order to be able to analyze the Celtic artifacts, a tech-nical problem had to be solved, namely how to mountspecimens of different shapes and sizes. A further com-plication was the fact that for the measurement of thebreaking surface the small area represents an unstable baseon which stand the samples. A spectrocup made of plastichad the best structure. The bottom side of the spectrocupwas removed, so that a cylindrical plastic tube remained.



The bottom side was replaced with a number of piecesof fishing line (diameter 0.2 mm), which formed a gridstructure. To hold the samples in a fixed position, fourthreads were cut through the wall of the container, whichwere perpendicular to each other. With screws it was pos-sible to fix each sample from four sides in such a waythat the desired surface was irradiated by the x-rays. Toestablish the best position, several test measurements for10 s were made.

The plastic of the spectrocup also met expectations andsupplied an excellent blank spectrum. A set of eight ofthese special holders were produced to allow automaticmeasurements with the sample changer.

Homogeneity

In order to have a better overview for the homogeneityof the Celtic glasses, a series of measurements of 21randomly chosen samples from a museum were made.Some of these samples were partly very corroded or

generally in a state rather unfavorable for the EDXRFmeasuring method, so test measurements were performed.

The technical parameters were always the same: volt-age, 35 kV; current, 0.35 mA; atmosphere, air; filter,aluminum; and livetime, 100 s.

Generally the samples were pieces of broken glasseswith a slightly curved shape and some of the mostbeautiful ones are shown in Fig. 3. On the exteriorwere sometimes ornaments, partially or full colored. Thecolored ornamentations were from a yellow substance.The lower surfaces were smooth, although one could dis-cover unevenness with the help of a microscope.

Points on the upper and lower surface and on thebreaking surfaces were measured in order to obtain fullinformation about the state of the sample. First was themeasurement of points on the side without ornamentation,the second the ornament itself and third at a location onthe breaking surface.

The comparison of the different sides showed someinhomogeneities, probably produced due to environmental

Figure 3. Fragments of Celtic glass bracelets. Top row left, blue fragment with ornament, No. 18; top row middle and right, front andside views of green–blue fragment, No. 19; bottom row left and right, blue fragment with white ornament, No. 39; bottom row middle,knobbed bead, presumed to be imported from the Black Sea area, No. 72.



influences. Analyzing the two breaking points gave muchbetter results since (1) these parts were changed muchless by the environmental influences and (2) they show asmoother surface.

Geometry study

In order to establish the effects of geometry influences onmeasurements, a series of experiments were performed,divided into four sections: (1) 4ð 90° rotation test in thereference plane; (2) angle dependence vs reference plane;(3) height dependence vs reference plane; and (4) surfaceroughness dependence.

A possible loss of intensity is expected in all four cases.To study all cases, a flat glass reference standard withdimensions 35ð 35ð 3 mm was selected, in this caseNIST SEM 1412, Multielement Glass (National Instituteof Standards and Technology, Gaithersburg, MD USA).This standard was used throughout the entire measuringprocedure, including the partial surface destruction. Themeasurements were made with two different excitationenergies, (i) 15 kV, 0.35 mA, no filter, in vacuum and(ii) 35 kV, 0.1 mA, aluminum filter, in air.

Rotation test. First the square-shaped SRM was placedon the sample tray and the fluorescence intensity wasmeasured in the four positions achieved by a rotationthrough 90° in the reference plane. The reference planeis defined by the plane of the sample tray. Table 1 showsthe results obtained from measurements at 15 kV. Thedata shown are the mean value of the net counts collectedwith operating conditions 15 kV, 0.35 mA, 1000 s, of thefour positions for all detected elements. The maximumdeviation in one of the positions from the mean value isgiven in counts and percentages of the mean value.

Angle dependence of fluorescence intensities.The angledependence of the fluorescence intensities was measured,the sample being inclined with respect to the referenceplane, as shown in Fig. 4. This was performed by support-ing one edge of the selected standard with small Al platesof 0.5 mm thickness on one side to bring the sample up to2 mm inclined above the sample tray. The inclined samplewas rotated four times through 90° degrees as describedin the rotation test.

Table 1. Rotation test on SRM in the reference plane (4×90◦) with operating conditions 15 kV, 0.35 mA(counts in 1000 s)

Element Mean value Max. total deviation Max. deviation (%)

Na 2161 153 7Mg 6739 284 4Al 34688 448 1Si 429379 4013 <1K 105494 778 <1Ca 126365 937 <1Fe 3050 177 6Zn 386053 2779 <1Cd 34087 657 2Ba 68984 787 1Pb 35596 257 <1

Figure 4. Sketch of the arrangement to measure the inclined35 ð 35 mm square sample of an SRM in the four differentpositions for the rotation study.

Table 2 shows results from 15 and 35 kV measurements.The maximum deviation calculated is given in column 2for 15 kV and in column 3 for 35 kV. The results wereachieved as the absolute value of the difference betweena measured point and an in-plane point taken for themaximum of the deviation and divided by the in-planepoint value times 100 to give a percentage.

Height dependence.In this test the offset from the refer-ence plane and its influence on the fluorescence intensi-ties of each element in the sample were observed. Themeasurements started at the plane of the sample tray(H D 0 mm), then the sample was lifted parallel to the ref-erence plane in steps of 0.5 mm with Al plates up toH D2 mm. The results are presented in Fig. 5 from measure-ments with an applied voltage of 15 kV and in Fig. 6 for35 kV. In particular in Fig. 6 ‘El’ means elastic scattering(Rayleigh) and ‘Inel’ means inelastic scattering (Comp-ton). Thex-axis (H) represents the distance between thereference plane and the sample. The percentage value on

Table 2. Rotation test on SRM inclined at about 3◦ to thereference plane (4×90◦) with operating conditions15 kV, 0.35 mA (counts in 1000 s)

Max. deviation between reference planeposition and inclined sample position (%)

Element 15 kV, 0.35 mA, 1000 s 35 kV, 0.1 mA, 1000 s

Na 3Mg 18Al 9Si 9K 8 15Ca 9 14Ti 9Fe 18Cu 18Zn 9 11Sr 10Zr 7Inelastic scatter 7peak (35 kV)Elastic scatter 10peak (35 kV)Cd 10 10Sb 17Ba 10Pb 9 11



Figure 5. Results from a study of SRM height offset from the reference plane in steps of 0.5 mm from 0 to 2 mm. The percentagedata given are the ratios [N.H//N.0/]ð 100 where N is counts of the respective elemental fluorescence radiation and H is the positionmeasured with 15 kV excitation.

Figure 6. As Fig. 5 but with 35 kV excitation.

the y-axis is calculated as [N.H//N.0/] ð 100, whereNis the counts of the respective elemental fluorescence lineafter a 1000 s collection time.

Surface roughness dependence.To find out the effect of arough surface on the measured results, the standard glasswas scratched manually on the surface with a sharp steelscrewdriver to obtain a rough surface, which had a statuscomparable to the rough Celtic samples. Results of thesemeasurements are given in Table 3 for the excitation at15 and 35 kV. In the second and third columns the ratiosbetween the rough and smooth surface of the measuredfluorescence counts of the elements at the two differentvoltages are given.

Conclusions of test experiments.Rotation test. As can easily be seen, the average errors

were below 2%. Exceptions were those elements with thesmallest peak areas, which was not unexpected.

Angle dependence of fluorescence intensities. Theinclined sample at a rotation of 180° shows the largest

Table 3. Comparison of ratio of fluorescencecounts of elements from scratched andsmooth surfaces

Ratio [N(scratched)/N(smooth)]ð 100 (%)Element 15 kV, 0.35 mA, 1000 s 35 kV, 0.1 mA, 1000 s

Na 94.8Mg 77.6Al 74.6Si 72.4K 73.4 85.2Ca 77.3 87.1Ti 89.5Cu 94.2Zn 83.3 90.7Sr 94.3Zr 87.4Cd 76.8 93.9Sb 85.1Ba 77.1Pb 88.2 93.9Mean 79.6 90.1



deviation from the in-plane values that are presented inTable 2. This can be understood by the geometry of theincident primary beam and the detector position. The datafor the other three cases(0, 90 and 270°) show a deviationbelow 10%.

From these results, it can be concluded that the analysisof Celtic glasses will be possible even for naturallyinclined samples or those having a breaking surface withsimilar inclination. By the design of the sample holder,the best possible geometry conditions can be adjusted.

Height dependence. The variations of the height withrespect to the reference plane shown in Figs 5 and 6 givesthe errors caused of inexact adjustment. At a maximumof 2 mm above the reference plane a maximum deviationof about 20% was observed. The deviation at a level of1.5 mm shows a loss of 5.3% for this position. If the sam-ple is lifted more, this error will rapidly increase. From thereal measurements on Celtic samples done so far, it couldbe seen that this condition was easily fulfilled. Surfacedeviations were approximately within the range of 1 mm.

Surface roughness dependence. It is shown that themean value of the ratio as a percentage depends on theexcitation energy. The mean value for 15 kV was 79.6%,but for 35 kV it was 90.1%.

Generally, it follows that the deviations which resultfrom bad positioning and rough surface of the SRMsaffect the measured values within the above values.Therefore, it can be assumed that based on these resultsthe quantification of the elements in the Celtic sampleswill be possible despite their shape within a maximumerror of 20%.

Quantitative analysis of glass standards

The use of standard reference materials (SRMs) is acommon method to check the results of one’s ownmeasurements. The quantification procedure includes thesame steps as for unknowns and the SRMs are treatedin the same way as the sample. The following steps areundertaken for the quantification procedure in a sample:collection of XRF spectra, spectrum evaluation as peakdeconvolution and net peak counts of the elements, andquantification using the fundamental parameter technique.On the experimental side this requires the determinationof the instrumental geometry factor, a measurement whichis done on one ultra-pure single-element sample. If it canbe proved that the chosen experimental conditions andthe data evaluation procedure for the standards fits thequantification of unknowns is not a problem.

Therefore, two NIST standards: (1) Soda LimeContainer Glass (SRM 621) and (2) MulticomponentGlass (SRM 1412), were measured and analyzed. Thetechnical parameters were the same as intended for theunknown samples: voltage, 15 kV; current, 0.35 mA;filter, none; atmosphere, vacuum; and livetime, 1000 s.With these operating conditions, three Soda Limestandards and eight Multicomponent Glass standardswere analyzed. The calibration of the system, for thedetermination of the instrumental constant required inthe software package AXIL Fundamental Parameters,was executed with a piece of pure aluminum. Owingto the position of aluminum in the Periodic Table, goodquantification results are expected for energy region up to

Table 4. Comparison of certified and measured data fromNIST Soda Lime Container Glass (SRM 621)

Certified Mean value 100 ð (mean�Component value (%) (measured) (%) certified)/certified (%)

Na2O 12.74 7.9 �38MgO 0.27 0.24 �12Al2O3 2.76 2.29 �17SiO2 71.13 60.25 �15SO3 0.13 0.23 C74K2O 2.01 1.56 �23CaO 10.71 9.71 �9TiO2 0.014 0.016 C14BaO 0.12 0.11 �6

Table 5. Comparison of certified and measured data fromNIST Multicomponent Glass (SRM 1412)

Certified Mean value 100 ð (mean�Component value (%) (measured) (%) certified)/certified (%)

Na2O 4.69 5.36 C14MgO 4.69 2.94 �37Al2O3 7.52 6.82 �9SiO2 42.38 47.11 C11K2O 4.14 3.65 �12CaO 4.53 4.27 �6BaO 4.67 6.11 C30

Ti K˛ or for the low-Z elements. It is also suitable for Baand Cd, because their L-lines are in the calibrated region.

Results are presented in Tables 4 and 5 and show themean concentration values as a percentage forn D 3and 8, respectively. The deviations as a percentage frommeasured and certified values are given in the last column.

In conclusion, one can see that for the light elementsthe errors are in the region of 20%, within the expecteduncertainty. The results for elements such as Na and Sare still under investigation. There are several reasonsfor such deviations. The sample is thick, corrections forself-absorption have to be considered based on theoreticalfundamental parameters and the x-ray tube spectrum inthis energy range is strongly dependent on the real tube inuse but was taken from theoretical calculations. Also, thedetector efficiency is calculated based on theoretical dataand technical data on the thickness of the Si dead layerand Au contact layers were taken from manufacturers’information, which has the greatest influence on the Naresults. Some of the elements Mg, Ti and Ba are presentin the higherµg g�1 (ppm) range, so errors up to 20%have to be accepted.

Qualitative and quantitative analysis of Celtic glasses

EDXRF has the capacity to give qualitative andquantitative results for sample analysis. Before thequantification procedure it is worth presenting the resultsof qualitative analysis to establish how often certainelements are found in the examined sample. Especiallyfor archeological pieces, sometimes just the existence ofa certain element shows clearly the origin or location ofmanufacture of a sample. With the help of qualitativeanalysis, normally a trend can be seen.



In the case of Celtic glasses, so far 30 artifacts havebeen measured and the results of qualitative analysis areavailable. In Fig. 7 the rate of occurrence is given whichdefines the existence of individual elements in the samplesconsidered (n D 30).

The existence of some of the elements such as Na,Si and Ca is expected. The presence of Co in allof the samples agrees with other publications,2,4 inwhich analysis of blue Celtic glasses was described. Theconcentration of Pb varies; in some of the samples it islow, whereas in others it is much higher. It is difficult todetect Cl, since during the excitation with an Rh tube theRh L-lines overlap Cl K .

As an example, the results of the quantitative analysisof two samples shown in Fig. 3 are presented in Tables 6and 7. In Fig. 3 the top row (middle and right) showsa green–blue fragment of a Celtic glass bracelet with ayellow ornament (No. 19) in front and side view. Theblue glass part looks blackish in the photograph but withillumination by light becomes a blue color. Two analysispoints were chosen in the blue area, three in the green

area and two on the ornament. The colorants of blue andgreen were subjected to further investigations of higherZelements. In the ornament the main colorant element islead. Sample 72, which is a knobbed bead, can be seen inFig. 3 in the middle of the bottom row. Again the yellowornament has a high lead concentration, whereas the whitering shows no lead.

The overall error varies from element to element andreaches a maximum of 20%.

Archeological interpretation

The studied artifacts resulted from a stay of the Celtictribes in the present area of eastern Austria. The majoritythe finds studied represent Celtic glass-work from theMiddle and Late La Tene periods (glass braceletsand ring-beads) and were found on the open Celticsettlements. The other group of findings comes from thelate Hallstatt and early La Tene graves (small beads ofvarious types), and their Celtic origin is still not clear.The artifacts were collected from different archeological

Figure 7. Rate of occurrence of chemical elements in the set of 30 artifacts.

Table 6. Analytical results for sample 19, in particular three different colors of the artifact (top row, middle and right inFig. 3)

Blue Green Ornament (yellow)Position 1 Position 2 Position 1 Position 2 Position 3 Position 1 Position 2

Component (µg g�1) (µg g�1) (µg g�1) (µg g�1) (µg g�1) (µg g�1) (µg g�1)

Na 27100š2200 18900š1940 8418š1464 11600š1560 14300š1720 <DL 23600š2480Na2O 73061š2970 50954š2610 22692š1970 31270š2100 38548š2310 <DL 63617š3350Mg <DL 1432š383 <DL 1293š334 <DL 3256š597 1570š486MgO <DL 2374š635 <DL 2145š554 <DL 5399š990 2604š807Al 3881š254 6798š278 2985š209 3003š216 2804š231 2804š231 6707š323Al2O3 7333š480 12800š530 5642š395 5676š409 5300š437 5300š437 12700š610Si 182800š400 176000š410 155400š350 163100š370 198500š400 225600š540 221900š500SiO2 391100š860 376500š870 332300š760 349000š780 424500š850 482600š1160 474600š1060S 794š82 949š84 310š68 399š70 256š70 7440š228 2114š131SO3 1982š207 2370š210 774š171 997š176 639š176 639š176 5278š327K 2562š50 3507š54 1782š44 2118š45 1915š45 6729š111 4925š85K2O 3086š60 4224š65 2146š53 2551š55 2306š55 8106š134 5932š102Ca 37800š100 36300š100 28000š90 30700š90 31400š90 54600š180 56600š160CaO 52900š140 50700š140 39100š120 43000š130 43900š130 76300š250 79100š230Ti 565š19 845š20 449š18 513š18 416š17 1329š38 1172š32TiO2 943š32 1409š34 750š30 855š31 695š28 2218š64 1956š53Pb Low Low No No No High High



Table 7. Analytical results for sample 72, in particular three different colors of the artifact (bottom row, middle in Fig. 3)

Ornament (yellow) Ornament (white) BlackComponent Position 1 Position 2 Position 1 Position 2 Position 1 Position 2

Na <DL <DL 6043š1561 5214š1511NaO2 <DL <DL 16290š2104 14055š2037Mg 5934š552 8813š694 1692š391 1497š383 <DL 4809š479MgO 9839š916 14600š1150 2807š648 2483š636 <DL 7973š794Al 6304š358 11300š460 9267š316 11300š330 4588š239 15400š420Al2O3 11900š680 21400š860 17500š600 21300š620 8669š452 29100š800Si 158200š540 236100š660 165800š440 173800š450 80300š340 162400š510SiO2 338500š1150 505000š1420 354600š940 371700š960 171900š730 347300š1100S 14300š560 21900š780 5027š179 5518š183 4804š165 7847š275SO3 35600š1390 54700š1940 12600š450 13800š460 12000š410 19600š690K 5924š121 8045š143 5338š107 6024š116 5653š72 9061š102K2O 7136š146 9691š172 6429š129 7256š140 6809š86 10900š120Ca 47800š200 54000š220 67000š170 66400š180 48700š130 59300š160CaO 66900š270 75600š300 93800š240 92900š260 68200š180 82900š230Ti 780š47 1307š53 955š34 1064š35 1050š31 2090š44TiO2 1301š78 2181š89 1594š57 1775š59 1752š52 3486š73Pb High High No No Low Low

sites: Pottenbrunn-Ratzersdorf, Neukirchen, Niederleis,Altenburg-Umlauf, Bad Fischau, Gumping, Rosenburg,Zwingendorf, Goettlesbrunn, Oberleiserberg and Vicenice(Bohemia). The studied glass ornaments were mademostly of blue (or turquoise) and also colorless, violet,brown–yellow and green glass of different transparency.Some of them were additionally decorated with white oryellow opaque glass. The glass coloration was specifiedaccording to the Pantone color charts.

The results of the analyses obtained so far allowus to draw the following conclusions: the Celtic glassfrom the area of eastern Austria belongs mainly tothe soda-lime type, the most common in the AncientWorld; the glass colorants (e.g. cobalt, iron, copper,manganese), decolorants (antimony, manganese) andopacifying elements (lead, antimony) are strongly markedby the results of the analysis and are strongly dependentof their relative concentrations; the samples of glassstudied show substantial similarities in their chemicalcomposition and a distinctive similarity to the analogouscollections from the La Tene culture; a number ofother elements appear in glass probably as naturalimpurities.

Further conclusions concerning questions of thepresumed production centers and the distribution zonesand also aspects of the chronological significance of the

concentrations of different elements in the studied glasseswill be possible after more data have been obtained.

CONCLUSIONS

Apart from the results of the quantification tests onthe SRMs, it could be shown that fragments of Celticglasses can be analyzed successfully. They are in mostcases sufficiently homogeneous, have a sufficiently goodsurface quality and by using the Rh tube excited EDXRFtechnique a high dynamic range of concentration canbe measured. Applying 15 kV the low-Z elements aremeasured; for medium-Z elements 35 and 50 kV areused particularly to determine Sb and Ba. A correlationamong the elements causing the coloring will be obtainedvia statistical evaluation of all data on the elementsfound. Questions of archeological interest can hopefullybe answered after the full data set has been interpreted.

Acknowledgements

We acknowledge the support of this project by the Austrian NationalScience Foundation, project No. P12526-SPR, and the excellentcooperation with the administrators of museums and collections thatallowed us to analyze their extremely valuable objects.

REFERENCES

1. J. Frana, A. Mastalka and N. Venclova, Archaeometry 29, 69(1987).

2. P. Hoffman, Fresenius’ J. Anal. Chem. 349, 320 (1994).3. S. Bichlmeier, M. Heck and P. Hoffmann, in Proceedings

of the European Conference on EDXRS, edited byJ. E. Fernandez and A. Tartari, p. 311. Bologna UniversityPress, Bologna (1999).

4. J. Braziewicz, M. Karwowski, M. Jaskola and M. Pajek, Adv.X-Ray Anal. 39, 857 (1997).

5. I. C. Freestone, in Catalogue of Mediaeval Enamels inthe British Museum, edited by N. Stratford, p. 37. BritishMuseum Press, London (1992).

6. I. C. Freestone, in Science and The Past, edited byS. Bowman, p. 37. British Museum Press, London (1991).

7. M. Wegstein, H. Urban, P. Rostam-Khani, A. Wittershagenand B. O. Kolbesen, Spectrochim. Acta, Part B 52, 1057(1997).

8. G. A. Cox and A. M. Pollard, Archaeometry 19, 45 (1977).9. G. Halmetschlager, Diplom Thesis, Technische Universitat

Wien, Vienna (1998).10. QXAS Software Package: Quantitative X-Ray Analysis

System. IAEA Headquarters, Vienna (1995).


P. Wobrauschek, G. Halmetschlager, S. Zamini, C. Jakubonis, G. Trnka, M. Karwowski;...

Documents

Transcript of P. Wobrauschek, G. Halmetschlager, S. Zamini, C. Jakubonis, G. Trnka, M. Karwowski;...