A Critique on Thermal NAA Estimation of Coinage Metals in Ancient Myanmar Coins David Tin Win Faculty of Science and Technology, Assumption University Bangkok, Thailand Abstract High activation cross sections of coinage metals, especially silver and copper, for thermal neutrons make it possible to use Non-destructive Thermal Neutron Activation Analysis in estimation of coinage metals. A study at Yangon University involved Thermal NAA determination of silver and copper contents in sixteen ancient Myanmar coins of five different historical periods. Silver targets were bombarded with thermal neutrons from Am/Be neutron source to induce 107 Ag(n,γ) 108 Ag reaction, and the emitted gamma radiation was counted, halflives estimated, and silver and copper contents were calculated. A careful observation of the work revealed that the working curve based on instant activity A o vs mass of silver standard, called working curve A, was not linear. Also a sharp decline in gamma counts was seen at masses higher than 10 g. Hence curve A was limited to masses of 10 g or less. A more appropriate working curve was obtained by excluding the highest mass (standard number 6), was called working curve B. However a working curve based on log A o vs. mass of standard, called working curve C, suggested by the present author, gave a better fit of the linear regression line. A six-fold enhanced accuracy in C values as compared to B values was observed from comparing the mean relative errors of 9.52% (B) and 1.48% (C). Stacking of standards approximated the masses of test coins but did not approximate the coin shapes. Thus uniform self-shielding between test coins and standards was not achieved. High self shielding at high masses and high mass pressures (1 g cm -2 or more) led to very low gamma counts and accounted for the mass limitation of 10 g. Keywords: 107 Ag(n,γ) 108 Ag, Am/Be neutron source, ancient Myanmar coins, coinage metals, copper, silver, self shielding, Thermal NAA Introduction Neutron Activation Analysis (NAA), a simple yet highly efficient chemical analysis method is one of the established analytical tools in chemistry (Win 2004). Determination of silver content in ancient coins using a high neutron flux of a nuclear reactor is well established (Wyltenbach 1966) and is still popular in recent times (Gordus 1995). However, the possibility of using a low neutron flux, such as those from Am/Be source was demonstrated much later. One such successful experiment was at Yangon University in Myanmar (Thiele, et al.1970). A high capture cross section of silver for thermal neutrons induces sufficient activity in silver coins to enable determination of its silver content, even at lower fluxes. Also the short half-life of 108 Ag allows conveniently short irradiation times. Based on this, thermal NAA determination of silver and copper contents in sixteen ancient Myanmar coins of five different historical periods was done at Yangon University (Aung 1989). A description of the work and a critical comment mainly on the working curve used follows. 55
Equipment Experimental A voltage stabilized and regulated (for spectrum stability) Super scalar Aloka Counter, model TDC-6-PS9, 6115 from Japan Radiation and Medicinal Electronics INC, Japan Radio Company, was used for detection of gamma radiation. The operating voltage, determined by using a sealed 60 Co radioisotope from Nuclear Chicago, was 850 v and a gain of ¼ was used. Materials Pure silver and copper made into thin uniform sheets of about 0.025 cm thick and cut to average coin sizes were used as standards. As checked by Atomic Absorption Spectroscopy (AAS), the silver standards contained a minimum of 99.69% silver with 0.28% copper; and the copper standards contained 100% copper, with negligible amounts of other metals. The test coins were ancient Myanmar coins, the oldest being over 2000 years and the most recent was minted in the Yadanabon period during the reign of King Thibaw (1878-1885), the last Myanmar monarch. The major experimental difficulty was using identical exposures on standards and test coins. This demanded a uniform irradiation field and similar self shielding in both standards and coins. A more or less uniform exposure field was achieved by rotating the standards and coins. Equal self-shielding in both standards and coins required identical shapes and masses for standards and coins. Five types of pure silver standards were shaped to approximate the various coin shapes, and coin masses were approximated by stacking the standards. Procedure The operating voltage of the Aloka Counter was determined by using a sealed 60Co radioisotope from Nuclear Chicago. The stable voltage of 850 v at ¼ gain, the mid-point of the horizontal portion in the net count vs applied voltage curve, was taken as the operating voltage. The stability of the counter was checked by running a series of counts and computing the standard deviation, which was found to be 3.11. The statistically acceptable range was 2 4.2 (Thiele, et al. 1970). Hence the counter was deemed stable at the operating voltage. Thus all experiments were done at 850 v using ¼ gain. Neutron irradiation was done with an Am/Be neutron source, which was placed in the center of a container filled with de-ionized water. The target (test coin or standard) was attached via a screw lock onto a motor shaft that can be rotated at a uniform steady speed of 135 rpm. The target was 3 cm away from the neutron source. Silver Determinations: Irradiation of coins and standards were done under identical environments. Normal irradiation time was 20 minutes. Silver standards were used singly and also as multi-stacked units consisting of up to six standards. Stacking was done to approximate the masses of the various coins studied. Gamma activities of 110 Ag and 108 Ag were monitored in a lead shielded Aloka Geiger counter, at various decay times (time interval after the end of the exposure). Decay times of 1.00, 1.25, 1.50, 1.75, and 2.00 minutes were normally used. The irradiated coins were placed just above the surface of the detector crystal. The counting time was ten seconds with five second intervals between successive counts. Count rates were corrected by excluding the background radiation. Decay curves were obtained by plotting corrected count rates against decay times. The experimental half-life was obtained from the decay curve of silver standard and compared with that of 108 Ag to positively identify that the gamma counts were due to decay of 108 Ag radionuclide. The instant activity A o was obtained by interpolation of the decay curve to zero decay time. A working curve was drawn using instant activities of silver standards against their silver contents. The instant activity of each coin was then determined and the silver content read off from the working curve. Copper Determinations: Here the gamma radiations were due to 64 Cu. A slightly modified 56
procedure was used for copper estimation. Standard copper samples were sealed in plastic before irradiation. The exposure time was prolonged to five days. Also the counting time was ten minutes with one hour intervals between successive counts. Results and Discussion The instant activities of 108 Ag radionuclides produced in silver standards were shown in table 1. The silver standard masses in gram unit of 2.243, 4.350, 6.586, 8.382, 10.468 and 12.928 reflected zero, two, three, four, five and six stacking. The standards were numbered from 1 to 6. The data was used to construct an A o vs silver standard working curve, called working curve A, and was shown in figure 1. Gamma counts declined sharply at masses above 10 g. This was probably due to enhanced self shielding caused by bulk interference. Hence studies should have been limited to coins with a maximum mass of 10 g. The NAA based silver contents of heavier coins would just be apparent values that were much less than the true values. Results for such heavy coins would be highly unreliable. A more reasonable working curve should exclude standard number 6, since it was outside the limit. The suggested curve, called working curve B, was shown in Fig.2. Working curve C consisting of log A o vs silver standard curve (Fig. 3) would be an even better working curve. Although the error bars in C are apparently wider, both B and C have identical error bars of 3%. As shown by the two linear regression trend-lines in B and C, the latter had a better fit (trend-line passed through all the experimental points) and consequently a better accuracy than those in the original work (Aung 1989). For convenience, the masses read off from the trend-line of working curve B were referred to as B masses, and the corresponding masses obtained from working curve C trend-line were called C masses. From Tables 1 and 2 it was seen that double-stacked standards (4.35 g) had an instant activity A o of 220 cps and gave estimated silver masses of 5.24 g (B) and 4.48 g (C). The latter was 14.5% lower and closer to the true value of 4.35 g. A conversion factor of 0.85 was required to convert the silver content A into the corresponding more accurate B value. A similar examination of triple stacked standards (6.586 g) with instant activity A o of cps gave silver estimations of 7.12 g (B) and 6.68 g (C), with a 6.2% difference. The required conversion factor was 0.94. Four stacked standards (8.382 g) with instant activity A o of cps gave 8.382 g (B) and 8.38 g (C) silver estimations, with a difference of 0.02%. The required conversion factor was 0.99. Five stacked standards (10.468 g) with instant activity A o of cps gave silver estimations of 9.84 g (B) and 8.52 g (C), with a 13.4% difference. The required conversion factor was 0.86. The average of the above differences was therefore 8.53% and the mean conversion factor was 0.91. Thus the C series silver contents were lower than the B series values by roughly 8.53% and were closer to the true value. The C series values were more accurate than the B series values. An average correction factor of about 0.91 was needed to convert the B values in the original work (Aung 1989) into the true experimental values given by C. As shown in Table 2, the B values had relative errors of 20.46, 8.11, 0 and 5.99%. The mean relative error was 8.64%. The relative errors of the C values were 2.99, 1.43, 0.02, and 18.61%, with an average relative error of 5.76%. Comparison of the average relative errors of 8.64 (B) and 5.76% (C) showed that working curve C gave more accurate values than those from working curve B. This was more pronounced if the last data was neglected. This data gave an abnormally inflated relative error of 18.61% in C values and magnified the mean relative error. It probably reflected a drastic self-shielding effect at high mass. It this last data was excluded, the mean relative errors were 9.52 (B) and 1.48% (C), indicating that curve C based values were sixfold more accurate than curve B based values. Silver and copper contents of ancient Myanmar coins of the Tagaung, Hanlin, Rakhine, Bodawpaya and Yadanabon periods 57
(Kings Mindon and Thibaw) were shown in Table 3. The Tagaung period, which appeared over 2000 years ago, was usually considered as the start of organized Myanmar history. This was followed by Hanlin period (1 st -10 th Century). Arakan or Rakhine (recent name) period lasted from 16 th 17 th Century). The Bodawpaya (1782 1819) and Yadanabon periods (1853 1885) were part of the Konbaung dynasty, which was founded by King Alaungpaya. King Mindon, the patron and initiator of the fifth Buddhist synod, was the second last King of the Yadanabon period. King Thibaw was the last King of the Yadanabon period, the last King of the Konbaung dynasty and the last Myanmar King. He was exiled to India by the British. The coins studied were from the above historical periods. Apparently the primary objective of the Yangon University study (Aung 1989) was demonstrating the feasibility of using NAA in coinage metal estimation in ancient coins, and not a study of coinage metal comparison of coins from the different Myanmar historical periods. Hence sampling was not exhaustive for any of the periods. Silver and copper averages calculated from table 3 would not represent the real averages for the different historical periods. Hence comparison would be rough guesses. However some observations could still be made. The Tagaung coins were over 2000 years old. Hanlin coins dated to between the first and tenth centuries. Rakhine coins were minted during the 16 th and 17 th centuries. The Bodawpaya and Yadanabon coins were relatively recent. Of the coins studied, the Yadanabon number 5 coin, with a diameter of 1.04 cm and mass of 0.708 g was the smallest and had the highest silver content of 94%. It probably reflected the need to minimize the size in order to use minimum amount of quality grade silver. The Tagaung number 1 coin, with a diameter of 3.6 cm, was the largest. It contained only 8.285% silver with 63.38% copper. The Tagaung number 2 coin also had a high copper content. Hence Tagaung coins were copper coins. Similarly the Rakhine coins with 25.1% silver average and the Bodawpaya coin were copper coins. The copper value of 117.4% in the Bodawpaya coin needed an explanation. Chemical analyses like other measurements, always involve errors that depended on the instrument or procedure used. Some methods could yield large positive errors that result in percentages over 100%. The coins of the Hanlin and Yadanabon periods did not contain copper. The Hanlin coins had an average of 68.36% silver and the Yadanabon coins had average silver contents of 38.92%. Thus Hanlin coins had the highest silver contents. As compared to a silver content of 49.17% minimum (coin number 2) and 83.59% maximum (coin number 3) in Hanlin coins, variation was more pronounced in the Yadanabon coins, which had a minimum of 5.25% (coin number 7) and a maximum of 94% (coin number 5). The Tagaung coins and Rakhine 3 were massive (above 10 g) and were outside the limit of the working curves. From table 3, the mass population, defined as the mass per unit area (for one surface - one side of coin), of these massive coins were 1.13 g cm -2 (Tagaung 1), 1 g cm -2 (Tagaung 2), and 1.21 g cm -2 (Rakhine 3). These had respective silver contents of 8.285%, 7.533% and 19.84%. These were deflated results caused by high self shielding. Yadanabon coin 1, another massive coin, had a low mass population of 0.78 g cm -2 and a relatively high silver value of 25.11%, which was apparently free from self shielding interference. Thus it could be concluded that massive coins (10 g up) that have mass population of 1 g cm -2 or more would show drastically reduced apparent silver contents. Conclusion A six-fold enhanced accuracy in C values as compared to B values was observed as shown by the mean relative errors of 9.52% (B) and 1.48% (C). Hence the log instant activity vs silver standard working curve should have been used in the Yangon University work. Since coin sampling was not exhaustive, coinage metal comparison of coins was not appropriate. Observations showed that the coins 58
of the Hanlin and Yadanabon periods did not contain copper. The Hanlin coins had an average of 68.36% silver and the Yadanabon coins had average silver contents of 38.92%. The Arakan average was 25.1%. Apparently Hanlin coins had the highest silver contents. Finally, massive coins (10 g up) with mass population of 1 g cm -2 or more would have drastically reduced apparent silver contents due to enhanced self shielding. Estimation should be limited to coins with less than 10 g mass. References Aung, T. 1989. Determination of Coinage Metal Composition of Ancient Burmese Coins by Thermal Neutron Activation Analysis. M.Sc. Thesis, Yangon University, Myanmar. Gordus, A.A.1995. Neutron Activation Analysis of Microgram Samples of Sasanian Coins and Metallic Art. Material Issues in Art and Archaeology IV. Proc. Materials Research Society Symposium, Pittsburgh, PA, USA. 352: 613-20. Thiele, R.W.; Khin, A.; and Kyaw, M. 1970. Neutron Activation Analysis of Ancient Burmese Coins with a Low Flux Am/Be Neutron Source. Chemistry Document, Yangon University, Yangon, Myanmar. Win, D.T. 2004. Neutron Activation Analysis (NAA). AU J.T. 8: 8-14. Wyltenbach, A. 1966. Coinage Metals in Ancient Coins. Helvetia Cnimica Acta, 49: 2555-64. Table 1. Instant activities of 108 Ag radio-nuclides produced in silver standards Silver standard Mass of silver Instant activity number standard/g Ao/cps Log Ao 1 2.243 160 2.2041 2 4.35 220 2.3424 3 6.586 2.3979 4 8.382 2.4393 5 10.468 2.4771 6 12.928 180 2.2552 Silver standard number Table 2. Estimated Silver Contents of Ag Standards Silver standard/g Estimated silver (B)/g Error (B)/% Estimated silver (C)/g Error (C) /% 1 2.243 NE NE NE NE 2 4.35 5.24 20.46 4.48 2.99 3 6.586 7.12 8.11 6.68 1.43 4 8.382 8.382 0 8.38 0.02 5 10.468 9.84 5.99 8.52 18.61 6 12.928 NE NE NE NE B = working curve (Ao vs Silver Standard). C = working Curve (log Ao vs Silver Standard). NE = not estimated 59
Table 3. Silver and copper contents of ancient Myanmar coins Historical period Coin Diam Area/ Press./ number Mass/g /cm cm 2-2 g cm Silver/% Copper/% Tagaung > 2000 years 1 22.933 3.6 10.18 1.13 8.285 63.38 2 13. 2.9 6.61 1 7.533 65.97 Hanlin (1-10th Century) 1 9.265 3.33 8.77 0.53 72.32 Nil 2 9.965 2.76 5.99 0.84 49.17 Nil 3 9.965 2.03 3.27 1.52 83.59 Nil Rakhine (16-17th Century) 1 9.074 2.27 4.08 1.11 28.65 64 2 9.778 2.56 5.15 0.95 21.48 54 3 11.091 2.44 4.6 1.21 19.84 Nil Bodawpaya* (1782-1819) 1 8.73 2.47 4.83 0.9 9.16 117.4 Yadanabon (1853-1878) Mindon * 1 11.55 3.07 7.45 0.78 25.11 Nil Yadanabon (1878-1885) Thibaw * 2 5.525 2.44 4.68 0.59 45.25 Nil 3 2.999 1.98 3.08 0.49 61.66 Nil 4 2.919 1.5 1.77 0.82 34.27 Nil 5 0.708 1.04 0,85 0.42 94 Nil 6 5.795 2.47 4.83 0.6 6.9 Nil 7 5.711 2.57 5.23 0.55 5.25 Nil * Bodawpaya, Mindon, Thibaw = Myanmese kings ND = not detected Instant Activity vs Standard Ag Mass 325 Instant Activity / cps 225 200 220 175 180 160 150 2.243 4.35 6.586 8.382 10.468 12.928 Standard Ag Mass / g Fig. 1. Working curve A instant activity A o vs silver standard Series1 60
Instant Activity vs Standard Ag Mass 325 Instant Activity / cps 225 200 220 175 160 150 2.243 4.35 6.586 8.382 10.468 Standard Ag Mass / g Series1 Linear (Series1) Fig. 2. Working curve B instant activity A o vs silver standard Log Instant Activity vs Standard Ag Mass 2.6 2.55 2.5 2.4771 Log Instant Activity 2.45 2.4 2.35 2.3 2.3424 2.3979 2.4393 2.25 2.2 2.2041 2.15 2.243 4.35 6.586 8.382 10.468 Standard Ag Mass / g Series1 Linear (Series1) Fig. 3. Working curve C log instant activity vs silver standard 61