- Email: [email protected]
PIXE in experimental archaeometry
399
Nuclear Instruments and Methods in Physics Research B3 (1984) 399-403 North-Holland, Amsterdam
IV B. Analysis of archaeologicalsamples
PIXE IN EXPERIMENTAL Guy DEMORTIER,
Bruno
ARCHAEOMETRY Van OYSTAEYEN
and Annick
BOULLAR
L.ARN, Faculths Universitaires de Namur, 22 Rue Muzet, B-5000 Namur, Belgium A new step in the understanding of the degree of expertise in soldering in antiquity is discussed. It takes into account results from analyses of noncommercial alloys manufactured with techniques known in early times. The procedure involves only rudimentary means: mixture in a charcoal furnace of gold (or natural gold alloys such as electrum) with a natural cadmium mineral whose colour is close to that of gold and to that of gold solders described in the oldest metallurgical handbooks. PIXE both in nonvacuum and in microprobe arrangements is used to analyse ancient artifacts and soldering alloys of commercial origin and those manufactured by this “ancient procedure.”
1. Introduction The analysis of ancient jewelry items requires the use of a quantitative, accurate, reliable, topographical, but mainly nondestructive method which is then suitable for giving the chemical composition of different parts of a jewel without any sampling, even at the microscopic level. PIXE and PIGE (sometimes called PIGME), especially used with the objects under investigation at atmospheric pressure, possess most of these qualities. When used simultaneously, both techniques may solve most of the problems of interference often present in prompt physical methods of analysis. When narrower regions are to be investigated, especially the regions of solders on an artifact, the proton microprobe (with the sample in the vacuum) is used. Its topographical resolution offers the possibility of investigating regions whose dimensions are well below those of the smallest hand-made solders attainable and of testing the homogeneity of the material melted during the soldering procedure. Analyses on noncommercial soldering alloys obtained by dissolving natural minerals in melted gold are used to explain several differences between solders observed on artifacts purported to be ancient on the modern soldering alloys (both containing cadmium) that we have recently reported [1,2].
2. Experimental arrangement for PIXE analysis In the non-vacuum “ milliprobe” arrangement, the incident proton beam crosses a thin (12.5 pm thick) foil of Al before reaching the sample situated at a distance of 1 cm in air. Aluminum is chosen as the material for the exit foil because its characteristic X rays are out of the range of detection of the Si(Li) assembly. When the diameter of the exit collimator supporting the Al foil is 0168-583X/84/$03.00 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)
B.V.
reduced to 0.5 mm, the diameter of the proton beam is about 0.7 mm at the target position, 1 cm after the exit foil. The LARN proton microprobe with the sample mounted on an X-Y manipulation frame in the vacuum may also be used to scan the samples (mechanically as well as electrostatically). Various beam sizes have been used to give a representative composition of materials. These sizes are sometimes higher than the best attainable spatial resolution (= 2 pm). In both experimental arrangements (milliprobe and microprobe) the X-ray analysis is performed using a foil of zinc (10 to 30 pm thick) inserted between the target and the detector. Zinc is indeed a selective absorber for Au L X rays. This procedure allows the reduction of the total counting rate by reducing mostly the contribution of the superabundant Au X rays, and enhances the sensitivity to detect characteristic K X rays from Ag and Cd. No beam monitor is used during the experiments. It is assumed for the quantitative determination of the total concentration of elements that the material contains only au, Ag, Cu, Zn and Cd. Other elements sometimes present at trace levels (e.g. Cr, Fe) are of no importance for the purpose of our study of the composition of the objects in noble metals. If their concentrations may affect the actual composition of the analysed material, their presence is taken into account by introducing the intensity of their characteristic X rays in the calculation. Absolute concentrations are obtained by comparison with a reference material. As no linearity may be expected between the actual concentration of elements and the counting rate in each characteristic X-ray peak, calculation involves physical parameters as cross-sections, stopping powers, mass attenuation coefficients and geometry of the irradiation procedure [3]. Ka lines were used for the analysis of Cu, Ag, Cd; Lb for the analysis of Au. Other lines were also investigated to test the accuracy. The statistical errors on IV B. ARCHAEOLOGICAL
ANALYSIS
G. Demortier et al. / P1X.E in experimental ~r~haearne~~
400
the measured values were always less than 1% for Au, and less than 5% (relative) for Cu, Ag and Cd, if they were present at concentrations higher than 3% but always better than 10% in all less favourable cases.
3. The ancient soldering of gold Two recipes for the preparation of gold solders are known from antiquity, one derived from “natural” chrysocolla (from greek upuoos and rcoXXa), the other obtained by mixing Cu, Ag and Au. As reported by Pliny [4] “natural chrysocolla is an exudation found in the shafts.. . . . .The colour of this natural product is yellow.” The preparation of artificial chrysocolla, as described in the Leyden Papyrus X (another old metallurgical handbook), includes “copper of Cyprus: 4 parts; asem: 2 parts; gold: 1 part; The nature of asem used in this preparation is not yet certified. Amongst others, the term asem has been used in old manuscripts to describe any substance whose properties (colour, brightness, weight, . . .) counterfeit those of precious metals [5]. The second gold solder known from the earliest times contains: “gold: 2 parts; copper: 1 part.” This last proce-
dure gives an excellent brazing alloy comparable with those used today [6]. Since the second half of the last century, brazing alloys containing cadmium have been extensively used. A small amount of cadmium in a gold matrix is able to lower the melting point of the brazing alloy below that obtained by introducing equivalent concentrations of copper, zinc or silver in a goid matrix. An extended discussion of different techniques of soldering in ancient jewelry has been published by Thouvenin [7]. In addition to these discussions and analyses, Thouvenin himself reproduced several pieces of jewelry of ancient workmanship. He concluded that goldsmiths of antiquity had skillfully realized very fine and tiny solders, not by using brazing alloys (a technique only useful for “macrosolders”) but by diffusion at about 900°C of very fine powder of deoxidized copper giving rise, at “microsites,” to in situ binary or ternary alloys. 4. Cadmium and forgery The presence of cadmium in a jewel is often interpreted as of modem origin: cadmium was indeed identified
Table 1
Composition of basic gofd-silver-copper aftoy and soldering alloy obtained after melting this basic alfoy with CdS Confidence factor a)
Remarks
71.7
1.01
mean value for 5 impacts
Concentrations (%) CU
Ag
Cd
Au
Antique gold (Au + Ag + Cu) C copper 12.0
16.3
(Antique gold +copper)+CdS
4.8 5.0 4.9 7.2 7.5 9.2
14.5 14.2 14.0 15.3 14.70 12.5 15.6
8.5 8.90 9.1 8.5 10.7 13.0 14.2
71.8 72.1 71.9 71.3 67.4 67.0 61.0
1.01 0.99 0.99 1.03 0.96 0.90 0.98
16.1 16.4 19.4 26.8 25.0 29.3 29.5
14.1 12.3 14.8 16.7 12.2 11.4 10.5
26.9 43.2 28.5 21.4 45.4 46.8 50.5
42.9 28.1 37.3 35.1 17.4 12.5 9.6
0.88 0.89 0.92 1.03 N.S. N.S. N.S.
32.8
14.0
48.7
4.5
N.S.
5.2
a) The confidence factor measures the variation of the ratio I,/lP significant). For discussion of this confidence factor see ref. 1.
gold rich regions
gold poor regions
black earthy coating
of XL lines of gold when this ratio is significant. (N.S. = not
G. Demortier et al. / PIXE in experimenial
by Strohmeyer as a new chemical element only around 1815. During our analyses of more than 150 pieces of ancient jewelry, cadmium was sometimes observed at concentrations up to several per cent but often at lower concentration, mainly at sites where soldering is necessary to bind elements together. We have demonstrated that the amount of cadmium found in these jewels (often originating from Iran, Syria, or south Italy) was related to the amount of copper and silver in proportions completely different from those observed in modern soldering alloys [l]. For jewelry of ancient origin (or purported to us to be ancient) one observes a correlation in cadmium and copper contents: both chemical elements increase simultaneously at sites where some soldering is perceptible or expected. From observation of modern soldering alloys anticorrelation of cadmium and copper contents is always observed, one reason being that for a given degree of fineness (say 7505&o)the copper or/and silver contents of the alloys must be decreased when the cadmium content is increased. Measurements on monetary medallions recently discovered at La Rochelle (France), and analysed in the Laboratoires des Mu&es de France (Louvre) and also by our PIXE technique, indicate the presence of cadmium at trace level only (81. We have also observed that antique gold jewels made in one single block and for which no soldering was necessary sometimes contain cadmium at low concentration, and generally inhomogeneously distributed in the bulk of the material [9]. This observation may be explained by assuming that the jewel had been made with a material obtained by melting an older gold jewel containing cadmium as a soldering component. This cadmium content was then diluted and unevenly distributed in the metal when it was remelted. This last interpretation of cadmium content is also allowable for the jewels from La Rochelle. Cadmium is a rare element in natural minerals, often present as traces in zinc and lead ores. Sometimes it appears as an orange-yellow coating (cadmium sulfide or greenockite) on zinc minerals and particularly on blende. As greenockite and zinc ores are different as far as their crystallographic properties are concerned, the crystals of greenockite may be isolated easily from their mineral backing of blende by only a gentle brushing. The powder that we have collected by this means is nearly pure cadmium sulfide. We think that it is not improbable that the colour of such natural crystals intrigued the jewelry crafstmen of antiquity when searching for materials, not only to be sold, but also to be substituted for gold or gold alloys.
5. Cadmium soldering alloys in “ancient fashion”. By using “antique” methods, gold of different degrees of fineness, cadmium sulfide and fire, we have
archaeometty
401
obtained in a carbon crucible, soldering alloys quite suitable for use. A few milligrams of gold were melted to form a little sphere in the bottom of the crucible. Powder of cadmium sulfide was then poured into the melted gold. It dissolves instantaneously, the result being a small sphere of alloy which was microscopically analysed. Each sphere of alloy appears as sufficiently homogeneous. Its center contains less cadmium than the surface but the difference in the concentrations between the center and the surface is less than 20%. Depending on several conditions (shape of the crucible, addition of copper and iron minerals, heating just above the melting temperature of the basic gold or far above, etc.), we have observed that the solubility of the cadmium sulfide reaches saturation when the cadmium content lies between 1% and 10%. Addition of copper minerals or copper enhances the cadmium solubility. Thermal analysis of such mixtures indicates that the melting point of the alloy may be lowered down to 720°C even if its degree of fineness is greater than 850%. Spark source spectrometry, used for the analysis of trace elements, indicates that at least 95% and some-
Fig. 1. Photography of the region of a gold granulation 1.2 mm in diameter soldered on a pure gold sheet by diffusion bonding using CdS. Total scan from C to D. Regions A and B contain significant amounts of cadmium. IV B. ARCHAEOLOGICAL
ANALYSIS
402
G. Demortier ei al. / PIXE in experimental
archaeomerry
:.
p-1
11
*i
B
4
. i
r. 2 000
I
io
Ag % 1500
6b
i'0
1000 JJ'm
Fig. 2. Cadmium distribution across the join obtained by PIXE microprobe. times all the sulfur (introduced with the cadmium sulfide) is eliminated during the alloying procedure. When powder of cadmium sulfide is mixed with turnings of gold and simultaneously heated to be melted, the dissolution takes a longer time than when greenockite is poured into liquid gold. Furthermore, the alloy obtained by this last procedure is less homogeneous. The sphere of gold-cadmium alloy sometimes appears to be formed in two different parts: one a rich gold alloy appearing at the top of the crucible, the second a black coating appearing at the bottom. When such a
sphere is gently compressed the black coating is easily separated. When irradiated by the proton beam this black coating gives rise to emission of visible light similar to that reported during the study of ancient artifacts [l]. The concentrations of elements at several sites on soldering alloys obtained by dissolution of CdS in a melted alloy containing 12% Cu and 16.3% Ag in 71.7% Au are given in table 1. The correlation between Cu and Cd is easily observed. The ternary diagram of copper-silver-cadmium is very similar to that observed on artifacts [l]. Copper contents in regions where cadmium is in low concentration are lower than the
b
‘0
Ag% -
So
6'0
Ag % c
sb
6b
LO
20
Fig. 3. Ternary diagrams of Cu-Ag-Cd, (a) Commerical brazing alloy [1,2]: @ before melting, * after melting (rapidly), * used for soldering, 0 after melting for 5 min. (b) Region of solder on an imperial necklace from the 1st century [1,2]; regions where the cadmium content is: A less than 0.5%. 0 between 0.5 and 1%. n between 1 and 2% * between 2 and 4% A more than 4%. (c) Brazing alloys obtained by dissolution of CdS in melted gold: 0 basic gold with 1% Cu, n basic gold with 4% Cu. A basic gold with 12% Cu, n%black coatings.
2b
G. Demortier et al. / PIXE in experimental archaeometty
initial copper content of the alloy used. The question is: “ is cadmium sulfide useful for the purification of gold?’ A granule (2 mm in diameter) was soldered onto a thin sheet of pure gold using an interposed turning of “our” brazing alloy obtained by dissolution of CdS in pure gold. The whole was heated at a temperature sufficient to melt the turning without melting the other parts by using a rudimentary blow pipe through a Bunsen burner as heat source. The join observed by the PIXE proton microprobe reveals that the alloy contains less than 1.5%Cd. This small amount is sufficient to lower the melting point of the alloy to be suitable for easy brazing. More surprising is the soldering of a small granule (1.2 mm in diameter) of pure gold on a pure gold sheet without brazing alloy: CdS is powdered on the gold sheet supporting the granule. The whole is heated by the blow pipe until the join is formed but without any destruction or deformation of either sheet or granule. The thin sheet has then been folded in order to free the region of the solder, .which was scanned with a protonmicrobeam of 5 pm in diameter (fig. 1). Both ends of the pad (A and B) joining the sheet to the granule contain more cadmium than the center of this pad. the end of the pad close to the sheet (on which cadmium sulfide is spread before heating) contains a high cadmium concentration (up to 95%) covering a region about 50 pm wide. The region where the pad is fixed on the sheet is easily observed due to some difference in the colours of gold. This region has been indicated here by arrows on a black and white reproduction. The end of the pad close to the granule (B) contains much less cadmium (up to 6%) but is wider. Figure 2 shows the variation of the concentration of cadmium in a region of 1 mm around the solder; note that the concentration of cadmium on the sheets lies in the per cent range (C). The regions of solder where the distribution of cadmium is observed lie in a range of 1 mm. Beams less than 50 pm are necessary to scan the whole region of a solder, but milliprobes of 0.5 mm in diameter are very convenient and often sufficient for an elementary and significant investigation of main components in regions of solders on artifacts. Ternary diagrams of relative concentrations of Cu-Cd-Ag indicate a very close similarity between solders on artifacts (fig. 3b) and data from table 1 obtained on soldering alloys obtained by CdS dissolution in melted gold (fig. 3c), but a complete disagreement with modern commercial brazing alloys (fig. 3a).
The dashed arrows cadmium increase.
indicate
403
in each case the trends
of
6. Conclusions (1) PIXE with milliprobe and microprobe beams are powerful and nondestructive analytical techniques allowing a rapid and accurate determination of the elemental composition of narrow parts of items of jewelry. The milliprobe arrangement (very easy to process) is often sufficient for topographical analysis. The microprobe arrangement (less easy to run) must be reserved for special problems. (2) We have demonstrated that the problem of cadmium in the soldering of ancient gold artifacts, not yet resolved, must be reinvestigated by avoiding the too simple criterion: “cadmium = modem.” We cannot definitively conclude that this criterion is not valid but we have demonstrated that a very elementary technology involving means known from antiquity may have been used in ancient workmanship. Another and completely way to solve this problem may comprise a careful examination of solders on artifacts, mainly from Iran and Syria, for which no doubt of the authenticity is suspected.
VI G. Demortier and T. Hackens, Nucl. Instr. and Meth. 197 (1982) 223. 121 G. Demortier, Archeologia (Prthistoire et archeologic) Dijon (France) 176 (March 1983) 41. 131 B. Van Oystaeyen and G. Demortier, Nucl. Instr. and Meth. 215 (1983) 299. [41 Pliny, Natural History, Book 33, §25,29, English transl. H. Rackham (Harvard University Press); K.C. Bailey, The Elder Pliny’s chapters on chemical subjects, Vol. 1 (E. Arnold, London, 1929). 151 M. Berthelot, Introduction a l’ttude de la chimie des anciens et du moyen age (Editions Georges Steinheil, Paris, 1889). [61 L.B. Hunt, Gold Bulletin 9 (1976) 24. [71A. Thouvenin, Rev. archeol. de 1’Est et du Centre-Est. France 24 (1973) 11. PI J. Flouret, G. Nicolini and C. Metzger, Gallia - Fouilles et Monuments Archeologiques en France Metropolitaine 39 (1981) 85. 191 G. Demortier, in Gold Jewelry - Auvitex 5 - eds., T. Hackens and R. Winkes (Institut d’Archeologie et d’Histoire de l’Art, Louvain-la Neuve, Belgium, 1983) p. 215.
IV B. ARCHAEOLOGICAL
ANALYSIS
Nuclear Instruments and Methods in Physics Research B3 (1984) 399-403 North-Holland, Amsterdam
IV B. Analysis of archaeologicalsamples
PIXE IN EXPERIMENTAL Guy DEMORTIER,
Bruno
ARCHAEOMETRY Van OYSTAEYEN
and Annick
BOULLAR
L.ARN, Faculths Universitaires de Namur, 22 Rue Muzet, B-5000 Namur, Belgium A new step in the understanding of the degree of expertise in soldering in antiquity is discussed. It takes into account results from analyses of noncommercial alloys manufactured with techniques known in early times. The procedure involves only rudimentary means: mixture in a charcoal furnace of gold (or natural gold alloys such as electrum) with a natural cadmium mineral whose colour is close to that of gold and to that of gold solders described in the oldest metallurgical handbooks. PIXE both in nonvacuum and in microprobe arrangements is used to analyse ancient artifacts and soldering alloys of commercial origin and those manufactured by this “ancient procedure.”
1. Introduction The analysis of ancient jewelry items requires the use of a quantitative, accurate, reliable, topographical, but mainly nondestructive method which is then suitable for giving the chemical composition of different parts of a jewel without any sampling, even at the microscopic level. PIXE and PIGE (sometimes called PIGME), especially used with the objects under investigation at atmospheric pressure, possess most of these qualities. When used simultaneously, both techniques may solve most of the problems of interference often present in prompt physical methods of analysis. When narrower regions are to be investigated, especially the regions of solders on an artifact, the proton microprobe (with the sample in the vacuum) is used. Its topographical resolution offers the possibility of investigating regions whose dimensions are well below those of the smallest hand-made solders attainable and of testing the homogeneity of the material melted during the soldering procedure. Analyses on noncommercial soldering alloys obtained by dissolving natural minerals in melted gold are used to explain several differences between solders observed on artifacts purported to be ancient on the modern soldering alloys (both containing cadmium) that we have recently reported [1,2].
2. Experimental arrangement for PIXE analysis In the non-vacuum “ milliprobe” arrangement, the incident proton beam crosses a thin (12.5 pm thick) foil of Al before reaching the sample situated at a distance of 1 cm in air. Aluminum is chosen as the material for the exit foil because its characteristic X rays are out of the range of detection of the Si(Li) assembly. When the diameter of the exit collimator supporting the Al foil is 0168-583X/84/$03.00 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)
B.V.
reduced to 0.5 mm, the diameter of the proton beam is about 0.7 mm at the target position, 1 cm after the exit foil. The LARN proton microprobe with the sample mounted on an X-Y manipulation frame in the vacuum may also be used to scan the samples (mechanically as well as electrostatically). Various beam sizes have been used to give a representative composition of materials. These sizes are sometimes higher than the best attainable spatial resolution (= 2 pm). In both experimental arrangements (milliprobe and microprobe) the X-ray analysis is performed using a foil of zinc (10 to 30 pm thick) inserted between the target and the detector. Zinc is indeed a selective absorber for Au L X rays. This procedure allows the reduction of the total counting rate by reducing mostly the contribution of the superabundant Au X rays, and enhances the sensitivity to detect characteristic K X rays from Ag and Cd. No beam monitor is used during the experiments. It is assumed for the quantitative determination of the total concentration of elements that the material contains only au, Ag, Cu, Zn and Cd. Other elements sometimes present at trace levels (e.g. Cr, Fe) are of no importance for the purpose of our study of the composition of the objects in noble metals. If their concentrations may affect the actual composition of the analysed material, their presence is taken into account by introducing the intensity of their characteristic X rays in the calculation. Absolute concentrations are obtained by comparison with a reference material. As no linearity may be expected between the actual concentration of elements and the counting rate in each characteristic X-ray peak, calculation involves physical parameters as cross-sections, stopping powers, mass attenuation coefficients and geometry of the irradiation procedure [3]. Ka lines were used for the analysis of Cu, Ag, Cd; Lb for the analysis of Au. Other lines were also investigated to test the accuracy. The statistical errors on IV B. ARCHAEOLOGICAL
ANALYSIS
G. Demortier et al. / P1X.E in experimental ~r~haearne~~
400
the measured values were always less than 1% for Au, and less than 5% (relative) for Cu, Ag and Cd, if they were present at concentrations higher than 3% but always better than 10% in all less favourable cases.
3. The ancient soldering of gold Two recipes for the preparation of gold solders are known from antiquity, one derived from “natural” chrysocolla (from greek upuoos and rcoXXa), the other obtained by mixing Cu, Ag and Au. As reported by Pliny [4] “natural chrysocolla is an exudation found in the shafts.. . . . .The colour of this natural product is yellow.” The preparation of artificial chrysocolla, as described in the Leyden Papyrus X (another old metallurgical handbook), includes “copper of Cyprus: 4 parts; asem: 2 parts; gold: 1 part; The nature of asem used in this preparation is not yet certified. Amongst others, the term asem has been used in old manuscripts to describe any substance whose properties (colour, brightness, weight, . . .) counterfeit those of precious metals [5]. The second gold solder known from the earliest times contains: “gold: 2 parts; copper: 1 part.” This last proce-
dure gives an excellent brazing alloy comparable with those used today [6]. Since the second half of the last century, brazing alloys containing cadmium have been extensively used. A small amount of cadmium in a gold matrix is able to lower the melting point of the brazing alloy below that obtained by introducing equivalent concentrations of copper, zinc or silver in a goid matrix. An extended discussion of different techniques of soldering in ancient jewelry has been published by Thouvenin [7]. In addition to these discussions and analyses, Thouvenin himself reproduced several pieces of jewelry of ancient workmanship. He concluded that goldsmiths of antiquity had skillfully realized very fine and tiny solders, not by using brazing alloys (a technique only useful for “macrosolders”) but by diffusion at about 900°C of very fine powder of deoxidized copper giving rise, at “microsites,” to in situ binary or ternary alloys. 4. Cadmium and forgery The presence of cadmium in a jewel is often interpreted as of modem origin: cadmium was indeed identified
Table 1
Composition of basic gofd-silver-copper aftoy and soldering alloy obtained after melting this basic alfoy with CdS Confidence factor a)
Remarks
71.7
1.01
mean value for 5 impacts
Concentrations (%) CU
Ag
Cd
Au
Antique gold (Au + Ag + Cu) C copper 12.0
16.3
(Antique gold +copper)+CdS
4.8 5.0 4.9 7.2 7.5 9.2
14.5 14.2 14.0 15.3 14.70 12.5 15.6
8.5 8.90 9.1 8.5 10.7 13.0 14.2
71.8 72.1 71.9 71.3 67.4 67.0 61.0
1.01 0.99 0.99 1.03 0.96 0.90 0.98
16.1 16.4 19.4 26.8 25.0 29.3 29.5
14.1 12.3 14.8 16.7 12.2 11.4 10.5
26.9 43.2 28.5 21.4 45.4 46.8 50.5
42.9 28.1 37.3 35.1 17.4 12.5 9.6
0.88 0.89 0.92 1.03 N.S. N.S. N.S.
32.8
14.0
48.7
4.5
N.S.
5.2
a) The confidence factor measures the variation of the ratio I,/lP significant). For discussion of this confidence factor see ref. 1.
gold rich regions
gold poor regions
black earthy coating
of XL lines of gold when this ratio is significant. (N.S. = not
G. Demortier et al. / PIXE in experimenial
by Strohmeyer as a new chemical element only around 1815. During our analyses of more than 150 pieces of ancient jewelry, cadmium was sometimes observed at concentrations up to several per cent but often at lower concentration, mainly at sites where soldering is necessary to bind elements together. We have demonstrated that the amount of cadmium found in these jewels (often originating from Iran, Syria, or south Italy) was related to the amount of copper and silver in proportions completely different from those observed in modern soldering alloys [l]. For jewelry of ancient origin (or purported to us to be ancient) one observes a correlation in cadmium and copper contents: both chemical elements increase simultaneously at sites where some soldering is perceptible or expected. From observation of modern soldering alloys anticorrelation of cadmium and copper contents is always observed, one reason being that for a given degree of fineness (say 7505&o)the copper or/and silver contents of the alloys must be decreased when the cadmium content is increased. Measurements on monetary medallions recently discovered at La Rochelle (France), and analysed in the Laboratoires des Mu&es de France (Louvre) and also by our PIXE technique, indicate the presence of cadmium at trace level only (81. We have also observed that antique gold jewels made in one single block and for which no soldering was necessary sometimes contain cadmium at low concentration, and generally inhomogeneously distributed in the bulk of the material [9]. This observation may be explained by assuming that the jewel had been made with a material obtained by melting an older gold jewel containing cadmium as a soldering component. This cadmium content was then diluted and unevenly distributed in the metal when it was remelted. This last interpretation of cadmium content is also allowable for the jewels from La Rochelle. Cadmium is a rare element in natural minerals, often present as traces in zinc and lead ores. Sometimes it appears as an orange-yellow coating (cadmium sulfide or greenockite) on zinc minerals and particularly on blende. As greenockite and zinc ores are different as far as their crystallographic properties are concerned, the crystals of greenockite may be isolated easily from their mineral backing of blende by only a gentle brushing. The powder that we have collected by this means is nearly pure cadmium sulfide. We think that it is not improbable that the colour of such natural crystals intrigued the jewelry crafstmen of antiquity when searching for materials, not only to be sold, but also to be substituted for gold or gold alloys.
5. Cadmium soldering alloys in “ancient fashion”. By using “antique” methods, gold of different degrees of fineness, cadmium sulfide and fire, we have
archaeometty
401
obtained in a carbon crucible, soldering alloys quite suitable for use. A few milligrams of gold were melted to form a little sphere in the bottom of the crucible. Powder of cadmium sulfide was then poured into the melted gold. It dissolves instantaneously, the result being a small sphere of alloy which was microscopically analysed. Each sphere of alloy appears as sufficiently homogeneous. Its center contains less cadmium than the surface but the difference in the concentrations between the center and the surface is less than 20%. Depending on several conditions (shape of the crucible, addition of copper and iron minerals, heating just above the melting temperature of the basic gold or far above, etc.), we have observed that the solubility of the cadmium sulfide reaches saturation when the cadmium content lies between 1% and 10%. Addition of copper minerals or copper enhances the cadmium solubility. Thermal analysis of such mixtures indicates that the melting point of the alloy may be lowered down to 720°C even if its degree of fineness is greater than 850%. Spark source spectrometry, used for the analysis of trace elements, indicates that at least 95% and some-
Fig. 1. Photography of the region of a gold granulation 1.2 mm in diameter soldered on a pure gold sheet by diffusion bonding using CdS. Total scan from C to D. Regions A and B contain significant amounts of cadmium. IV B. ARCHAEOLOGICAL
ANALYSIS
402
G. Demortier ei al. / PIXE in experimental
archaeomerry
:.
p-1
11
*i
B
4
. i
r. 2 000
I
io
Ag % 1500
6b
i'0
1000 JJ'm
Fig. 2. Cadmium distribution across the join obtained by PIXE microprobe. times all the sulfur (introduced with the cadmium sulfide) is eliminated during the alloying procedure. When powder of cadmium sulfide is mixed with turnings of gold and simultaneously heated to be melted, the dissolution takes a longer time than when greenockite is poured into liquid gold. Furthermore, the alloy obtained by this last procedure is less homogeneous. The sphere of gold-cadmium alloy sometimes appears to be formed in two different parts: one a rich gold alloy appearing at the top of the crucible, the second a black coating appearing at the bottom. When such a
sphere is gently compressed the black coating is easily separated. When irradiated by the proton beam this black coating gives rise to emission of visible light similar to that reported during the study of ancient artifacts [l]. The concentrations of elements at several sites on soldering alloys obtained by dissolution of CdS in a melted alloy containing 12% Cu and 16.3% Ag in 71.7% Au are given in table 1. The correlation between Cu and Cd is easily observed. The ternary diagram of copper-silver-cadmium is very similar to that observed on artifacts [l]. Copper contents in regions where cadmium is in low concentration are lower than the
b
‘0
Ag% -
So
6'0
Ag % c
sb
6b
LO
20
Fig. 3. Ternary diagrams of Cu-Ag-Cd, (a) Commerical brazing alloy [1,2]: @ before melting, * after melting (rapidly), * used for soldering, 0 after melting for 5 min. (b) Region of solder on an imperial necklace from the 1st century [1,2]; regions where the cadmium content is: A less than 0.5%. 0 between 0.5 and 1%. n between 1 and 2% * between 2 and 4% A more than 4%. (c) Brazing alloys obtained by dissolution of CdS in melted gold: 0 basic gold with 1% Cu, n basic gold with 4% Cu. A basic gold with 12% Cu, n%black coatings.
2b
G. Demortier et al. / PIXE in experimental archaeometty
initial copper content of the alloy used. The question is: “ is cadmium sulfide useful for the purification of gold?’ A granule (2 mm in diameter) was soldered onto a thin sheet of pure gold using an interposed turning of “our” brazing alloy obtained by dissolution of CdS in pure gold. The whole was heated at a temperature sufficient to melt the turning without melting the other parts by using a rudimentary blow pipe through a Bunsen burner as heat source. The join observed by the PIXE proton microprobe reveals that the alloy contains less than 1.5%Cd. This small amount is sufficient to lower the melting point of the alloy to be suitable for easy brazing. More surprising is the soldering of a small granule (1.2 mm in diameter) of pure gold on a pure gold sheet without brazing alloy: CdS is powdered on the gold sheet supporting the granule. The whole is heated by the blow pipe until the join is formed but without any destruction or deformation of either sheet or granule. The thin sheet has then been folded in order to free the region of the solder, .which was scanned with a protonmicrobeam of 5 pm in diameter (fig. 1). Both ends of the pad (A and B) joining the sheet to the granule contain more cadmium than the center of this pad. the end of the pad close to the sheet (on which cadmium sulfide is spread before heating) contains a high cadmium concentration (up to 95%) covering a region about 50 pm wide. The region where the pad is fixed on the sheet is easily observed due to some difference in the colours of gold. This region has been indicated here by arrows on a black and white reproduction. The end of the pad close to the granule (B) contains much less cadmium (up to 6%) but is wider. Figure 2 shows the variation of the concentration of cadmium in a region of 1 mm around the solder; note that the concentration of cadmium on the sheets lies in the per cent range (C). The regions of solder where the distribution of cadmium is observed lie in a range of 1 mm. Beams less than 50 pm are necessary to scan the whole region of a solder, but milliprobes of 0.5 mm in diameter are very convenient and often sufficient for an elementary and significant investigation of main components in regions of solders on artifacts. Ternary diagrams of relative concentrations of Cu-Cd-Ag indicate a very close similarity between solders on artifacts (fig. 3b) and data from table 1 obtained on soldering alloys obtained by CdS dissolution in melted gold (fig. 3c), but a complete disagreement with modern commercial brazing alloys (fig. 3a).
The dashed arrows cadmium increase.
indicate
403
in each case the trends
of
6. Conclusions (1) PIXE with milliprobe and microprobe beams are powerful and nondestructive analytical techniques allowing a rapid and accurate determination of the elemental composition of narrow parts of items of jewelry. The milliprobe arrangement (very easy to process) is often sufficient for topographical analysis. The microprobe arrangement (less easy to run) must be reserved for special problems. (2) We have demonstrated that the problem of cadmium in the soldering of ancient gold artifacts, not yet resolved, must be reinvestigated by avoiding the too simple criterion: “cadmium = modem.” We cannot definitively conclude that this criterion is not valid but we have demonstrated that a very elementary technology involving means known from antiquity may have been used in ancient workmanship. Another and completely way to solve this problem may comprise a careful examination of solders on artifacts, mainly from Iran and Syria, for which no doubt of the authenticity is suspected.
VI G. Demortier and T. Hackens, Nucl. Instr. and Meth. 197 (1982) 223. 121 G. Demortier, Archeologia (Prthistoire et archeologic) Dijon (France) 176 (March 1983) 41. 131 B. Van Oystaeyen and G. Demortier, Nucl. Instr. and Meth. 215 (1983) 299. [41 Pliny, Natural History, Book 33, §25,29, English transl. H. Rackham (Harvard University Press); K.C. Bailey, The Elder Pliny’s chapters on chemical subjects, Vol. 1 (E. Arnold, London, 1929). 151 M. Berthelot, Introduction a l’ttude de la chimie des anciens et du moyen age (Editions Georges Steinheil, Paris, 1889). [61 L.B. Hunt, Gold Bulletin 9 (1976) 24. [71A. Thouvenin, Rev. archeol. de 1’Est et du Centre-Est. France 24 (1973) 11. PI J. Flouret, G. Nicolini and C. Metzger, Gallia - Fouilles et Monuments Archeologiques en France Metropolitaine 39 (1981) 85. 191 G. Demortier, in Gold Jewelry - Auvitex 5 - eds., T. Hackens and R. Winkes (Institut d’Archeologie et d’Histoire de l’Art, Louvain-la Neuve, Belgium, 1983) p. 215.
IV B. ARCHAEOLOGICAL
ANALYSIS