Application of gamma spectrometry for the characterization and influence of the archeological works of an archaeological site

The purpose of this study was the use of gamma spectrometry to characterize the archaeological site of Molina de Aragón, located in the province of Guadalajara in Spain. To do so, a set of samples with different historical influences have been selected. The samples were analyzed by gamma spectrometry with HPGe detectors, and by X-ray fluorescence, to know their chemical composition. The statistical study of the activity concentration of radionuclides from the natural radioactive series, 40K and 137Cs, was carried out using box-and-whisker plots, cluster analysis and principal component analysis (PCA). Likewise, the in-situ effective dose rates and the ones determined from the activity concentrations obtained by gamma spectrometry were evaluated to verify sampling reproducibility. The results obtained made it possible to classify the areas of Jewish and Christian influence based on the relationships between 232Th and 238U series and the possible influence of the archaeological works carried out in the study area. The least altered areas could also be identified from the activity concentrations of 137Cs and 210Pbex. Activity concentrations and effective dose rates were equivalent to the natural radioactive background of the Iberian Peninsula. PCA showed a correlation between Fe2O3, Al2O3 y ZrO2 and the natural radioactive series of thorium and uranium.


Introduction
Historically, Spain has been a country with a very abundant and varied archaeological wealth. In Spain there are remains of different cultures and civilizations such as: cave paintings (Atapuerca, Altamira …), remains of the Bronze Age (Motilla del Azuer, La Codera, …), Arab culture (Medina Azahara, Córdoba, Granada, …), Romans (Tárraco, Itálica, Baelo Claudia, Mérida, …), Visigothics and Islamic remains (Minateda), and a long etcetera. There are also architectural remains from more recent times, between the twelfth and sixteenth centuries, in which Romanesque and Gothic art had its maximum splendour (Segovia, Salamanca…). The period called the Middle Ages, covered the period of time between the fifth and fifteenth centuries, comprising several of the above-mentioned periods. The archaeological study of remains from these cultures is carried out with chemical techniques (XRF, ICP-MS, etc.), used for their characterization and dating [1]. In this sense, the project called ArcheoMedtal arises, which aims to include other types of techniques such as gamma spectrometry to characterize archaeological remains from the Middle Ages.
Gamma spectrometry is a non-destructive technique that allows the determination of the activity concentration of several gamma-emitting radionuclides in a single measurement [2]. Gamma-emitting radionuclides, both natural and artificial, provide valuable information to carry out a characterization of this type of archaeological remains. The possible imbalances between natural radionuclides belonging to the three natural radioactive series of uranium, actinium and thorium allow us to interpret the weathering phenomena that have been able to alter the materials over time [3,4]. These alterations can be interpreted based on the different chemical and physical properties of these radionuclides [5]. Furthermore, artificial radionuclides such as 137 Cs from the fallout are a tool used to study possible ground alterations [6].
In the present work, we set out to evaluate the use of gamma spectrometry as an alternative technique to the existing ones for the physical-chemical characterization of archaeological remains. Our working hypothesis is that the gamma-emitting radionuclides present in the study area, allow characterizing and interpreting the historical footprint that occurred in the archaeological remains belonging to the measured age period. We test our hypothesis by analyzing different samples taken from the ruins of the historic site of Molina de Aragón (Guadalajara province, Spain) to see the possible anthropic prediction derived from the metallurgical work carried out at this site. Soils with productive and non-productive implications were sampled. The objectives of this study were: (1) to characterize the study area using gamma spectrometry, in-situ measurements of the environmental equivalent dose and X-ray fluorescence, and (2) to statistically evaluate the historical footprint of the site and the influence of archaeological works.

Study area
The samples analyzed in this work were taken in the Castle of Molina de Aragón and in "Prao de los Judios", located in the historic complex of Molina de Aragón in the province of Guadalajara (Spain). This historical complex combines archaeological remains of the Jewish and Christian cultures during the Middle Ages [7]. The samples were selected to contemplate remains of these two cultures.
Samples 1, 2 and 3 were taken in the inner courtyard of the castle, sample 10 at the base of the Christian church and sample 4 at the hillside. All these samples were located inside the walled enclosure. The area with purely Jewish influence was represented with samples 5-9, which were located outside the walled enclosure. The samples were collected in different dependencies whose uses were: pottery (sample 5), metal foundry (sample 6), livestock activity (sample 7), housing (sample 8) and synagogue (sample 9). Figure 1 shows the different sampling points within the historic complex of Molina de Aragón.

Sample collection and preparation
The minimum quantity necessary of soil samples were collected superficially since it was not allowed to perform alterations in the studied ruins. Sampling was done with a 30 cm × 30 cm grid, taking 5 fractions in the four corners and in the center of the grid. These samples were then bagged and identified. Subsequently, the samples were dried in a Selecta oven at 105 °C. Once constant weight was obteined, the dried samples were ground in a FRITSCH brand ball mill (Planetary Ball Mill, pulverisette 5) and sieved at 300 µm. Finally, samples were deposited in a cylindrical polypropylene box with a diameter of 76 mm and a height of 30 mm. The boxes were completely sealed using parafilm to prevent loss of 222 Rn, and were left to rest for 21 days to achieve secular equilibrium between 226 Ra and 214 Pb and 214 Bi.

Gamma spectrometry
In this study, 3 types of gamma detectors were used: (1) extended range coaxial, (2) reverse electrode coaxial and (3) broad energy. Detectors were connected to 3 electronic chains consisting of a high voltage source (HV), an amplifier (AMP), an analog digital converter (ADC) and a data transfer module (AIM), all of them from Canberra Industries brand. The electronic chains were connected to a PC and the spectra were acquired and analyzed using Genie 2000 software. Efficiencies were calculated using LabSOCS software as the 3 detectors were characterized by CANBERRA. To calculate the efficiency, the procedure described was the one developed in [2]. The energies and emission probabilities of the radionuclides analyzed in this study were the following Interferences of 235 U with the 186 keV photopeak of 226 Ra, and 228 Ac with the 1460 keV photopeak of 40 K were corrected using the method proposed in [9].Samples were measured for 80,000 s and the backgrounds were measured for 600,000 s [10].The gamma spectrometry laboratory, where the measurements were performed, is accredited based on the UNE-EN ISO/ IEC 17025:2017 standard [11].

Chemical composition of samples
Chemical composition of samples was determined by X-ray fluorescence (WDXRF). An aliquot of each sample was shaped and sieved to 70 µm particle size with a 200 mesh sieve. The analysis was performed in the Chemistry Division of CIEMAT with a Malvern-PAN analytical AXIOS automated spectrometer.

Absorbed and effective dose rates calculation
Effective dose rates were measured in-situ and calculated from the activity concentrations obtained by gamma spectrometry. In-situ H*(10) effective dose rates at 1 m above ground were measured using a Lamse model MS6020 portable multi-probe monitor with an RD2L probe.
Theoretical effective dose rates were calculated from the absorbed dose, determined using the kerma rate per unit activity at 1 m height (nGy h −1 per Bq kg −1 ) for natural radionuclides (uranium, actinium and thorium series along with 40 K) and 137 Cs from the fallout [12,13]. The expression used to calculate the absorbed dose rate at 1 m height ( Ḋ 1m , nGy h −1 ) was the following one: where Cxxx A is the activity concentration in Bq kg −1 of radionuclide xxx A.
The expression used to determine the effective dose rate (E, µSv h 1 ) was as follows: where O is the occupancy factor that was considered 1 to be compared with the average in-situ H* (10), C is the conversion factor from absorbed dose in air to effective dose received by an adult (equal to 0.7 Sv Gy −1 ). The uncertainties associated with expressions (1) and (2) were estimated using the Kargten method [14,15].

Statistics
Analyses used were: (1) box-and-whisker plots, (2) cluster analysis with the nearest-neighbour method and squared Euclidean distance and (3) relative kurtosis and skewness. This statistical study of the data was performed using the Statgraphics Centurion XVII software, version 17.0.16. The box-andwhisker plot was used to analyze the possible imbalances in the natural radioactive series and the distribution of activity concentrations obtained for 40 K. Standardized kurtosis and skewness were used to check whether the activity concentrations of a given radionuclide fit a normal distribution. Cluster analysis was used to search for simple groupings based on their chemical composition and radiological content [16].
Relationship between the activity concentrations of natural (natural series together with 40 K) and anthropogenic ( 137 Cs) gamma-emitting radionuclides were analyzed along with chemical composition using principal component analysis. Factors selection was made based on the KMO (Kaiser-Meyer-Olkin) parameter. The analysis was carried out iteratively, removing the factors with the least correlation until a KMO greater than 0.7 was obtained, which indicates an acceptable correlation in the PCA [17]. The PCA analysis for this paper was generated using the Real Statistics Resource Pack software (Release 7.6). Copyright (2013-2021) Charles Zaiontz. www. real-stati stics. com [18]. Lastly, variables and scores of the samples were represented using a HJ-Biplot graph [19].
Comparison of in-situ and theoretical dose rates was performed using three statistical tests: (1) overlap percentage, to compare individual values [20], (2) Student's t-test to check if the means were statistically comparable, and (3) Fisher's F to verify if the variances were statistically comparable. Both tests were done for an α = 0.05 and for the tails test since it was sought whether the means and variances were statistically comparable, or not [21]. Table 1 shows the activity concentrations (Bq kg −1 ) of the samples collected in the historical site of Molina de Aragón. The range of activity concentrations for uranium series had a maximum value for 210 Pb of 63 ± 10 Bq kg −1 and a minimum value of 9.65 ± 0.46 Bq kg −1 for 214 Bi. The range of activity concentrations for thorium series was between 26.7 ± 1.1 Bq kg −1 for 228 Ac and 4.79 ± 0.30 Bq kg −1 for 208 Tl. Activity concentration of 226 Ra was higher than the one of 234 Th in samples 1 and 2, while in the rest they maintained the expected ratio of 1. Activity concentrations of 40 K varied between 189.4 ± 9.2 Bq kg −1 and 428 ± 19 Bq kg −1 . 235 U activity concentrations were below the lower limit of detection in all cases. Furthermore, 137 Cs was detected in 3 samples. 232 Th/ 238 U ratio in the samples from the inner courtyard of the castle (samples 1, 2 and 3) was 1.2, whilst the value obtained for the remaining samples was 0.75. Figure 2 shows the box-and-whisker plot for the natural radioactive series of uranium and thorium. Both series were in secular equilibrium. However, the activity concentration of 210 Pb in samples 10 was higher than that of the other radionuclides of the radioactive uranium series. This sample corresponds to the final part of the church wall. This behavior was also observed in sample 9, corresponding to the synagogue. The activity concentrations of 208 Tl were equivalent to those of 228 Ac and 212 Pb, taking into account the branching ratio of 35.94% for the disintegration of 212 Bi, which is the mother of 208 Tl [22]. Figure 3 shows the box-and-whisker plot of the 40 K activity concentrations of the analyzed samples, whose spread was low, as shown by the width of the 2nd and 3rd quartiles with the mean and median also being close to each other. However, 3 atypical values were found which corresponded to samples 1, 3 and 10. These values do not have a significant statistical difference since relative kurtosis and skewness values of 0.22 and 1.1 were obtained. These values must be taken with caution since the number of samples is only 10; however, they allow verifying that the atypical data would not be considered anomalous but distant from the distribution of the activity concentrations.

Effective dose calculation
The comparison of the effective doses calculated from the absorbed doses (expression 1) was used to verify that the sample measured in the laboratory was representative with respect to the sampled area. The reason was that the amount of sample that could be taken was less than usual in a typical soil sampling and therefore, this way, its reproducibility is ensured. Table 2 presents the calculated effective dose rates and H*(10) dose rates for the studied samples. The % overlap was in all cases greater than 20%. The overlaps indicate that both data sets are statistically the same. The result of Student's t-test and Fisher's F showed that the means were comparable (p = 0.91) and the variances were homoscedastic (p = 0.63).

Statistical grouping of samples
Cluster analysis was used to study the groupings between the experimental values of the different samples. This analysis was applied both for the activity concentrations (Fig. 4a), and for the percentages of oxides obtained by XRF (Fig. 4b). In the Fig. 4, similarities between the grouping obtained for the activity concentration of the natural radionuclides and 137 Cs and the obtained for the oxides determined by XRF are observed. In both cases samples from "Prao de los Judios" (samples 5, 6, 7, 8 and 9) grouped for the radiological and chemical results. Chemical analyses showed more clearly the results of the samples of the Christian ruins, although in both cases the conclusions are equivalent. Analyses show as well that the more affected samples by archaelogical works (1, 4 and 10) would be different from the rest.

Correlation between samples, their activity concentrations and chemical compositions
The correlation between the samples and their activity concentrations and chemical compositions was studied using PCA (Fig. 5). PCA represented the 96.2% of the variance of the values. The variables that allowed obtaining a KMO higher than 0.7 were: 208 Tl, 212 Pb, 214 Bi, 214 Pb, 228 Ac, Fe 2 O 3 , Al 2 O 3 and ZrO 2 . The vectors obtained for the variables show a high correlation between Fe 2 O 3 , Al 2 O 3 and the natural radioactive series of thorium ( 208 Tl, 212 Pb and 228 Ac). Likewise, a correlation between the uranium series ( 214 Pb and 214 Bi) and ZrO 2 was observed. The scores obtained for the previous variables showed three data sets: (a) samples 4-9, (b) sample 10 and (c) samples 1-3. The set (a) would represent the samples from the "Prao de los Judíos" and set (c) the samples from the Christian zone. Sample 10 [set (b)] would be more centered in the graph, which would indicate that it does not follow the same behavior as the previous ones. Finally, the samples of set (c) would have a higher radioactive content and the percentage of oxides would also be higher than those of set (a).

Discussion
The obtained results support our hypothesis that natural gamma-emitting radionuclides and 137 Cs make it possible to characterize and interpret the historical footprint of the archaeological remains of the historic site of Molina de Aragón.
Activity concentration levels of 226 Ra, 232 Th ( 212 Pb) and 40 K were equivalent to the mean values for Spanish soils [23]. The activity concentration of 137 Cs indicated an alteration of the study area due to the archaeological campaigns that have been carried out. However, activity concentration for point 10 (Christian Church) was 2.81 ± 0.29 Bq kg −1 , which is equivalent to the terrestrial fallout and would indicate that this area has remained unchanged [6]. For points 7 and 9, activity concentrations were equivalent to the lower detection limits of the remaining points, we cannot obtain the same conclusions as in point 10.
The obtained 232 Th/ 238 U ratio of 1.2 for samples 1, 2 and 3, corresponding to the inner courtyard of the castle, indicated the presence of granitic materials in the construction of the castle structure [24]. However, in the remaining samples, a ratio of 0.8 was obtained, which could reflect the use of other construction materials such as brick and plaster [25]. Likewise, the activity concentration of 226 Ra in samples 1 and 2 (visually verified clayey appearance) was higher than the one of the other radionuclides of the  natural radioactive series of uranium. This 226 Ra enrichment is typically given in the presence of clay, as has been observed in previous studies [26]. The hypothesis that could be proposed is that the 226 Ra from the eroded granite of the walls has been retained in the clay of the soil. This behavior is consistent with the results obtained in [27] in which the radiological content was found to be higher in the fine or loose fraction of the granite. Activity concentration of 210 Pb in sample 10 would indicate a value of 210 Pb in excess, or not supported by the radioactive  (Table A1 of the Supporting Information). The red key classifies the samples taken in the "Prao de los Judíos" uranium series ( 210 Pb ex ) of 47 Bq kg −1 . The presence of 210 Pb ex is related to slightly altered soils in which 210 Pb is generated by the disintegration of 222 Rn, and accumulates. 210 Pb ex is also related to the concentration of 137 Cs [28]. The values obtained in sample 10 are consistent with this behavior and the contribution of 210 Pb ex could be due to the presence of granite on the walls of the structure of the Christian church.
The effective dose rates obtained in-situ are equivalent to those deduced from the activity concentration of the gamma emitters present in the analyzed samples from the study area. The absorbed and effective dose rate levels are consistent with those obtained in previous studies in the Iberian Peninsula [29]. This finding shows that the samples analyzed in the laboratory by gamma spectrometry were representative of the areas where they were sampled.
The groupings obtained for the gamma spectrometry results (Fig. 4a) and the chemical composition obtained by XRF (Fig. 4b), were equivalent. In both cases, a clear grouping of the samples from the "Prao de los Judíos" is observed. In the case of the samples from the Christian zone, a grouping of samples 2 and 3 was observed, which was clearer for the clusters obtained for the chemical composition (percentage of oxides). Samples 1, 4 and 10 that would be most affected by the archaeological works were not grouped for these two groups described.
Finally, HJ-Biplot graph allowed to clearly observe the findings described above. The relationships of the variables showed a relationship between the natural radioactive series of thorium ( 208 Tl, 212 Pb and 228 Ac), and to a lesser extent with the uranium one ( 214 Pb and 214 Bi), with Fe 2 O 3 and Al 2 O 3 , which was already found in previous works on granites [27]. This association may be due to the presence of biotites (K(Mg,Fe) 3 (AlSi 3 O 10 )(OH) 2 ), chlorites ((Mg,F e) 3 (Si,Al) 4 O 10 (OH) 2 ·(Mg,Fe) 3 (OH) 6 ) and siderophyllites (KFe 22 + Al(Al 2 Si 2 O 10 )(OH) 2 ) [30][31][32]. On the other hand, another correlation was observed between the radioactive series of uranium ( 214 Pb and 214 Bi), and to a lesser extent with the thorium series ( 208 Tl, 212 Pb and 228 Ac), with Zr 2 O, which reflects the well-known relationship between zirconite and the radioactive series of uranium and thorium [32]. The scores obtained for the samples showed 3 sets of samples: (a) those from the "Prao de los Judíos" area, (b) samples alterated due to archaeological works and (c) samples from the Christian area. In the case of the samples from the "Prao de los Judíos", the results reflected the use of the materials described above (bricks and filling). However, samples 1-3 from the Christian ruins area, would show the use of granitic materials that is reflected by both, their chemical and radiological properties. In the case of sample 10 and to a lesser extent sample 4, the results show that they did not belong to none of the two groups described and, therefore, they would be mixed by the archaeological works in the study area.

Conclusions
The physiochemical characteristics of the natural gammaemitting radionuclides and 137 Cs, as well as chemical composition, have made it possible to classify the different areas of the historic site of Molina de Aragón. The statistical study applied to the analyzed samples distinguished the area with Jewish influence ("Prao de los Judios") from the area with Christian influence (Church and inner courtyard of the Castle). 137 Cs and 210 Pb ex allowed to identify the less altered areas of the historical assemblage. The dose rates obtained were equivalent to the radioactive background of the area.
The relationships found with the HJ-Biplot graph clearly differentiated the samples based on the scores obtained with the factors that provided a KMO > 0.7. On the one hand, a relationship between Fe 2 O 3 and Al 2 O 3 with the radioactive thorium series ( 208 Tl, 212 Pb and 228 Ac) and ZrO 2 with the uranium series ( 214 Pb and 214 Bi) was obtained. Samples with the highest radioactive content and with the greatest granite presence were those from the Christian castle complex. However, samples obtained in the "Prao de los Judíos" showed a lower contribution of these variables. The most mixed samples by the archaeological works were more centered in the HJ-Biplot graph.
This work has shown that it is possible to characterize an archaeological area solely with the information provided by gamma spectrometry. Moreover, it is not necessary to search for high levels of radioactivity to study a site, but to interpret the information provided by natural radionuclides together with 137 Cs. Although man has tried to leave his mark on time with great architectural constructions, something as small as uranium and thorium atoms, disintegration by disintegration, will always teach him and show him the true rhythm of history.