Spectrochimica Acta Part B 58 (2003) 627–633
Results of quantitative analysis of Celtic glass artefacts by energy dispersive X-ray fluorescence spectrometry夞 C. Jokubonisa, P. Wobrauscheka,*, S. Zaminia, M. Karwowskib, G. Trnkab, P. Stadlerc a
Atominstitut of the Austrian Universities, Vienna University of Technology, A-1020 Vienna, Austria b Institute of Prehistory, University of Vienna, A-1190 Vienna, Austria c Museum of Natural History, A-1010 Vienna, Austria Received 22 October 2001; accepted 14 May 2002
Abstract Energy dispersive X-ray fluorescence spectrometry measurements were applied on more than 200 Celtic glass ` Culture (f250–50 BC) which were found in the area that artifacts from the middle and late periods of the La Tene is presently Austria. This analysis yielded approximately 14 000 concentration data on 25 elements which could give analytical results for major, minor, and trace elements. Both bulk-material and ornamentation were analyzed. Coloring and opacifying elements were found in accordance with literature. In particular the concentration data of Zr and Sr show strong clustering and in comparison with archaeological chronology of the analyzed artifacts an excellent agreement of present historic knowledge was found. 䊚 2003 Elsevier Science B.V. All rights reserved. Keywords: Energy dispersive X-ray fluorescence spectrometry; Standard reference material (glass); Celtic glass analysis; Trace element analysis; Cluster analysis; Chronology of archaeological artifacts
1. Introduction This paper presents a project with a duration time of more than 3 years, dealing with an interdisciplinary co-operation of the analysis and statistical data evaluation of the chemical composition of Celtic glass artifacts. More than 200 handcrafted glass objects of the middle and late periods of the 夞 This paper was presented at the 16th International Conference on X-Ray Optics and Microanalysis (ICXOM-XVI), held in Vienna, Austria, July 2001, and is published in the Special Issue of Spectrochimica Acta Part B, dedicated to that conference. *Corresponding author. Fax: q43-1-58801-14199. E-mail address:
[email protected] (P. Wobrauschek).
` Culture (LT C1–LT D1, f250–50 BC) La Tene w1x, mainly bracelets and ring beads, were analyzed by means of energy dispersive X-ray fluorescence spectrometry (EDXRS). These artifacts result from the presence of Celtic tribes in the actual area of Austria. The glassfindings have been collected ` from many archaeological sites of the La Tene Culture in Austria. Most are in small pieces (f1– 5 cm=1–2 cm diameter), they are in good condition for measuring with EDXRS. Although there are well known problems of surface-sensitivity w2x, this method was chosen due to its non-destructiveness, an important factor for archaeology and demanded by some of the private collections and museums. Pre-considerations to the calculation of
0584-8547/03/$ - see front matter 䊚 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0584-8547(02)00289-6
628
C. Jokubonis et al. / Spectrochimica Acta Part B 58 (2003) 627–633
Fig. 1. EDXRS–geometrical arrangement.
quantitative results, including studies dealing with the influence of the geometry of the measured surfaces on possible concentrations, were published w3x. The geometry study indicated that sample positioning is very important for yielding reliable results, especially for the light elements, but also that it is possible to work within acceptable errors. Concluding it can be assured, that keeping the above described geometrical conditions and the experimental set-up, within the discussed limits, the results will be reliable. The calculated elemental-composition for each of the measured objects includes concentration data of 25 elements, varying from mgyg for mostly decolorants, colorants and trace elements to %, mostly main glass making substances. The quantitative results were interpreted with respect to glasscoloring and de-coloring substances and other minor trace elements. First attempts on this evaluation resulted in an allocation between the glasscoloring and decoloring substances by their elemental-composition and the color appearance of the artifacts analyzed. One of the most important results found is a relation between the elementalcomposition of Sr and Zr and the archaeological period of the samples. An understanding of raw materials used, trading connections among people or artists, and production procedures used, is expected from further archaeological interpretation of the available analytical data. 2. Experimental description X-ray fluorescence spectrometry (XRS) is a non-destructive, multi-element analytical tool for
qualitative and quantitative determination of the chemical composition of a sample. The working principle of XRS is based on irradiation of the artifact with ionizing radiation which causes excited states in the atoms of the object. The atoms’ reaction is the emission of element-characteristic X-rays, which can be detected and therefore used to determine the elemental-composition of the sample. The energy or wavelength of this characteristic X-rays corresponds to the atomic number of the element w4x, whilst the intensity (number of counted characteristic photons in a certain time) is in relation to the amount of the element of interest. The emitted characteristic X-rays are detected simultaneously by means of an energy dispersive spectrometer. EDXRS is based on spectra containing intensity vs. energy as the fundamental information for further calculations. Computer-based and software-oriented preparation of the yielded data will provide the analyst with net-intensities of collected photons, characteristic for certain elements. The quantitative composition of the artifact can be calculated via conversion of intensities into concentrations, taking into account all information about the condition of the sample and the used measuring procedure. One of the major advantages of EDXRS is the mechanical stability of the device, as no moveable parts are used during the measuring process. The number of necessary components is limited to the X-ray source, the sample, and the detection devices (Fig. 1). EDXRS sets up the possibility of non-destructive analysis of objects with almost no sample
C. Jokubonis et al. / Spectrochimica Acta Part B 58 (2003) 627–633
Fig. 2. Fragment of Celtic glass bracelet; Artifact 噛 104; Spot shows the size and geometry of measuring point ‘A’ on the break (photograph by N. Sautner, IUFG).
preparation. Nevertheless, one must be aware of the surface-quality and condition of the artifact with respect to its roughness, status of corrosion, washout, and enrichment, where all of these parameters are influencing the X-ray analytical results, in particular the light elements. The equipment used for the analysis was a ‘Spectrace TN 5000’ machinery from ‘Tracor.’ It represents a standard 908 EDXRS geometry with a low power (17.5 W), air-cooled rhodium anode X-ray tube and a liquid-nitrogen-cooled Si(Li) detector with an energy resolution of 150 eV at Mn Ka. The Be exit window of the X-ray tube is only 127 mm thick, which gives good conditions for exciting low Z elements with Rah-L lines w5x. To enable various measuring positions for the more or less randomly shaped glass artifacts, a sample holder of a spectroscopic suitable material with four holding-screws for mounting the sample was developed. This self-modified sample carrier consists of a commercial available ‘Spectrocup’ from ‘Spex,’ normally used for fluid and powder measurements. It was modified with four additional plastic screws to enable adjustment of the objects and a 3=3 grid of fishing line (0.2 mm diameter) to define the plane and point of the measurements. A collimator of 1 mm diameter, yielding an elliptic illuminated area by the 458 geometrical distortion, was designed and mounted on the exit window of the X-ray tube to irradiate only small parts of the samples, hence it is possible to irradiate flat portions of an irregular shaped surface. This self-made collimator consists of a carefully chosen material-combination (brass–aluminium–carbon) for suppressing unwanted characteristic X-rays
629
from the device itself. Consequent tests showed that these expectations are fulfilled. In the experimental set-up of the ‘TN 5000’-spectrometer a ten sample tray is provided. Using the mechanicallystable, PC-controlled hardware allows repetitive measurements under different excitation conditions as described below. In order to cope with the limitations of nondestructiveness and to face the problem of a soilcontaminated surface, the artifacts were carefully brushed and then cleaned with acetone. An average of four measurements of different spots were taken for each sample. Both bulk and ornamentation were measured on broken positions and surface parts (Fig. 2). To obtain optimal excitation conditions for all elements of interest, two different experimental Table 1 Comparison of measured and certified data of SRM NIST 621 and NIST 1412 NIST 621
Na2O MgO Al2O3 SiO2 SO3 K 2O CaO TiO2 Cr2O3 MnO Fe2O3 CoO NiO CuO ZnO SeO3 Br2O7 Rb2O SrO ZrO2 CdO SnO2 Sb2O3 BaO PbO
NIST 1412
Measured
Certified
Measured
Certified
71 260 -DL 23 230 756 400 -DL 22 160 124 200 30 -DL 84 433 -DL -DL 42 25 -DL 41 101 203 127 47 40 212 1048 300
127 400 2700 27 600 711 300 1300 20 100 107 100 140
75 980 48 400 80 940 502 200 8662 37 770 38 850 378 125 -DL 354 142 -DL 19 38 710 -DL 65 -DL 41 370 463 37 790 -DL 312 43 370 44 070
46 900 46 900 75 200 423 800
All data in mgyg.
400
70
1200
41 400 45 300
310
44 800
45 500 43 800
46 700 44 000
C. Jokubonis et al. / Spectrochimica Acta Part B 58 (2003) 627–633
630
Table 2 A selection of concentration data from various artifacts Artifact 噛
104A
71A
108A
111A
123A
61A
88A
103A
125A
142A
Artifact 噛
Na2O MgO Al2O3 SiO2 SO3 K 2O CaO TiO2 Cr2O3 MnO Fe2O3 CoO NiO CuO ZnO SeO3 Br2O7 Rb2O SrO ZrO2 CdO SnO2 Sb2O3 BaO PbO Period
58 670 8518 34 190 764 600 3900 7609 91 860 191 -DL 202 8422 35 20 128 53 -DL 33 12 375 279 -DL -DL 20 800 -DL 60 C1
29 290 4576 11 990 837 300 4121 4919 89 430 181 -DL 230 7444 -DL 18 55 43 -DL 62 15 466 377 -DL -DL 9265 -DL 239 C1
91 450 -DL 14 160 768 700 -DL 6006 102 800 146 -DL 225 12 300 777 39 1742 107 -DL 45 28 621 479 -DL 94 118 -DL 142 C1
66 750 4622 18 370 741 500 8260 4856 97 480 128 -DL 142 10 750 58 -DL 53 67 -DL 14 -DL 400 413 -DL -DL 45 630 -DL 230 C1
74 790 -DL 10 320 796 100 -DL 4213 92 340 134 -DL 204 13 550 1882 69 4313 231 -DL 29 11 682 632 -DL 116 165 -DL 185 ?
69 020 9280 31 680 737 100 2546 11 890 110 300 158 -DL 5037 10 760 2633 86 6642 244 -DL 64 27 1598 191 -DL 208 320 -DL 220 D1
73 200 6920 22 700 770 900 4624 9199 81 290 113 -DL 6836 12 090 3973 126 5503 90 -DL 61 51 1298 142 -DL 126 180 481 148 ?
68 730 5610 28 750 767 300 6410 9156 58 540 51 -DL 259 2551 -DL 36 19 030 76 -DL 49 -DL 653 101 76 1065 34 207 31 360 D1
64 770 -DL 19 860 784 000 -DL 10 370 92 240 85 -DL 5639 11 900 3105 113 5344 153 -DL 99 53 1668 175 -DL 62 202 -DL 162 C2
66 090 -DL 21 910 789 000 -DL 10 930 95 360 59 -DL 254 8197 2286 72 3550 126 -DL 79 59 1681 143 -DL 50 58 -DL 54 C2
Na2O MgO Al2O3 SiO2 SO3 K2O CaO TiO2 Cr2O3 MnO Fe2O3 CoO NiO CuO ZnO SeO3 Br2O7 Rb2O SrO ZrO2 CdO SnO2 Sb2O3 BaO PbO period
All data in mgyg. The last line correlates to the archaeological period. Rows ‘Cr2 O3 ’ and ‘SeO3 ’ in these selected artifacts do not show results, whereas in some others they do. Errors of quantification are below 20% and not presented for the clearness of the presentation.
parameters were applied on each of the spots. For low Z elements, ranging from Na to Fe, a tube voltage of 15 kV and a current of 0.35 mA with vacuum inside the chamber and no filter were used. Whilst the high Z elements, ranging from Cr to Pb, were measured with 35 kV, again 0.35 mA, air inside the chamber and an Al filter of 0.127 mm thickness in front of the tube window, respectively. Spectrum deconvolution, the stripping of the background-signal and calculation of net peak areas were done by the software program ‘AXIL’ of the ‘QXAS’ software package w6x, basically by means of iterative least square fitting. Problems of line overlaps concerning S-K and Pb-M lines as well as Ti-K and Ba-L lines are taken into account, whilst the severe overlap of Cl-K and the anode
emitted Rh-L lines was not solved, leading to rejection of the Cl information. The method of fundamental parameters calculation was tested on standard reference materials (SRM) (NIST 1412 Multicomponent Glass; NIST ¨ 621 Soda-Lime Container Glass; Breitlander, Silicate Glasses, Monitor-Samples for XRF, SVG1, SVY1, SVZ1), yielding quantitative results for all of the investigated elements. The applied quantification procedure takes into account the physics of X-ray production and gives a perfect interpretation of the whole origin of the measured spectrum. It calculates the primary radiation from the anode as well as the response function of the detection device, takes absorption correction in the sample into consideration, and inserts the unknown but demanded chemical com-
C. Jokubonis et al. / Spectrochimica Acta Part B 58 (2003) 627–633
631
Fig. 3. Measurements represented by the elemental-concentration of Zr and Sr with respect to the archaeological period of the artifacts. Errorbars are not presented due to the clearness of the presentation.
position of the sample in iterative steps until all the values fit. As a result of the analysis on SRM NIST 621 and NIST 1412, a comparison of measured and certified data is given in Table 1. The analytical results of the glass-making substances Na, Mg, Al, Si and K have to be considered carefully for the following reasons. Existing silicic acid gel layers, which are formed during contact of the artifact with water w7x while buried in soil, are actually not affected by the sample preparation. Absorption leads to extremely small information depth of low energy characteristic X-rays from light elements of some micrometers. Washout processes on near surface and surface areas due to humidity of the soil environment result in concentration-gradients from the exterior to interior layers of the object. Only a destructive treatment of the samples by entering deeper layers of the artifacts would give
more accurate results on the light element composition. Nevertheless, the medium to high Z elements from Ca upwards to Pb show reliable results because of increased information depth in the silicon matrix. 3. Results From more than 200 analyzed objects of Celtic glasswork, resulting in more than 14 000 data of elemental-concentration in stoichiometric oxide states, an assorted variety of measurements of ten different artifacts from various periods of time is presented in Table 2 where in particular the data for Zr and Sr are important with respect to results in Fig. 4. A first result of an interpretation of the concentration data is the separation of all measured objects by their elemental-composition of Sr and Zr into two groups and is presented in Fig. 3. Data
632
C. Jokubonis et al. / Spectrochimica Acta Part B 58 (2003) 627–633
Fig. 4. Elemental-concentration of Zr and Sr with reference to the artifact 噛, as presented in Table 2.
include bulk-material and ornamentation. A chart of selected data, dealing with the elemental-concentration of Sr and Zr to show the individual distribution of artifacts and time correlation presented in Table 2, is given in Fig. 4. In Figs. 3 and 4 the artifacts dated to the earliest ` period of Celtic glass-working (La Tene C1, f250–180 BC) are marked with circles (s). The squares (h) and triangles (n) symbolize the artifacts dated to the later periods (LT C2–LT D1, f180–50 BC), when glass ornaments became to be common and widespread products across the ` La Tene Culture and as we believe they were manufactured in numerous workshops. With only few exceptions, separation of those both groups of artifacts (‘early’ and ‘late’) depending to the concentrations of Sr and Zr is clear-cut. As one can easily see there is a set of artifacts with high Zr and low Sr concentration belonging to the group of early glass artifacts, whilst the other set belongs to the late ones, both dated due to archeological information.
The glass artifacts of uncertain chronology (within LT C1 and LT C2, f250–120 BC) are marked on the diagram with the rhombi (e). All artifacts in the upper-left part of the diagram represent the earliest stage of Celtic glass making. We believe it was a period of experimentation and recognition of glass as a new raw material, which soon after was commonly used to produce characteristic Celtic bracelets and many other ornaments. 4. Conclusions Applying EDXRS with a low power X-ray spectrometer and a specially modified collimator of 1 mm diameter—to avoid strong influence of the analyzed surface of the artifact—in a series of measurements, 200 Celtic glass artifacts were analyzed. Two different procedures were applied to analyze light elements (15 kV, vacuum, no filter) and medium—high Z elements (35 kV, Air, Al filter). Conversion of elemental intensities into
C. Jokubonis et al. / Spectrochimica Acta Part B 58 (2003) 627–633
concentration data was done by the fundamental parameters method with the software package ‘QXAS.’ More than 14 000 concentration data are available from the analyzed artifacts. Concentration data of coloring and opacifying elements are in accordance with literature. The clustering of concentration data of Sr and Zr seems to be of major archaeological importance, because it coincides with the present knowledge of archaeological chronology. Acknowledgments The project was supported by the FWF Austrian Science Fund, project 噛P12526-SPR. One of us (M. Karwowski) is grateful for funding of the Austrian BM:BWK. The presentation of results during the ‘16th International Conference on Xray Optics and Microanalysis-ICXOM XVI’ was awarded as best poster by the jury of the conference.
633
References w1x R. Gebhard, Der Glasschmuck aus dem Oppidum von Manching, in: Die Ausgrabungen in Manching, 11 1986, pp. 46–134. w2x G.A. Cox, A.M. Pollard, X-ray fluorescence analysis of ancient glass: the importance of sample preparation, Archaeometry 19 (1) (1977) 45–54. w3x P. Wobrauschek, G. Halmetschlager, S. Zamini, C. Jokubonis, G. Trnka, M. Karwowski, Energy-dispersive X-ray fluorescence analysis of Celtic glasses, in: E. Selin Lindgren (Ed.), X-Ray Spectrometry, 2000, pp. 25–33, Special Millennium Issue on Cultural Heritage. w4x (a) H.G. Moseley, The high frequency spectra of the elements, Philos. Mag. 26 (1913) 1024–1034 (b) H.G. Moseley, The high frequency spectra of the elements, Philos. Mag. 27 (1914) 703–713. w5x Thermo Noran (Spectrace Instruments, Inc.), Operators Manual, Spectrace 5000, (1990). w6x G. Bernasconi, A. Tajani, P. Kregsamer, QXAS Quantitative X-ray Analysis System IAEA, manual of software package, (1995–1996). w7x B. Soldenhoff, Proba okreslenia typow zniszczen szkiel archeologicznych, Acta Universitatis Nicolai Copernici, Archeologia Szkla, 5 (1991) 97–123.
