Radiological hazard indices and elemental composition of Brazilian and Swiss ornamental rocks

The objective of this paper was to evaluate the radiological risk index of ornamental rocks sold both in Brazil and Europe and to correlate their radioactive content with their chemical composition. The U, Th and K mean values were 62 ± 65, 122 ± 111, 1126 ± 516 Bq kg for Brazilian and 93 ± 59, 70 ± 67 and 1005 ± 780 Bq kg for Swiss samples, respectively. The radiological index: radium equivalent, external hazard index, absorbed dose rate in air, annual gonadal equivalent dose, annual effective dose equivalent, and excess lifetime cancer risk for Brazilian and Swiss samples were calculated. The main contribution for the radiological indices observed was the radionuclide Th, which is associated with REE, Br, Hf, Na, Rb, Sb and Zr in the rock matrix.


INTRODUCTION
In Brazil, approximately 80 percent of the exported dimension stones are commercially named granites [1].This category includes not only the materials meeting the concept of felsic rocks, but also any mafic fine-grained rocks, or other igneous or metamorphic rocks that possess qualities similar to granite's grainy and interlocking texture [2].In general, rocks generated in the Earth's crust are more enriched in radioelements, e.g.uranium, thorium and radium, relative to those formed in the mantle.This phenomenon is a consequence of magma partial fusion and fractioned crystallization processes that concentrated them in the silica-enhanced liquid phase [3].
The natural radioactivity of the rocks used as building materials has been increasingly studied due to the rising number of homes coated with ornamental rocks.Such aspect has been causing health concerns as an elevation of toxicity risk in these places may occur [3].The European Union, for example, has a proper legislation regarding natural radioactivity in building materials, which establishes the gamma activity concentration index as an assessment tool [4].Thus, this paper aims to evaluate the radiological risk index of dimension stones collected both in Brazil and in Switzerland as well as to correlate their radioactive content with their chemical composition based on the concentration of the stable elements.

Sampling and sample preparation
The analyzed samples include magmatic and metamorphic rocks of different lithologic types used for decorative purposes that are best sellers in Brazil (20 samples) and Switzerland (14 samples), i.e. diorite, syenite, monzogranite, granitic pegmatites, gneisses, etc.The selected rocks list from Brazil can be found in Fig. 1, and were named according to the Brazilian state acronym where they were collected.Switzerland samples (named CH-sample) are not necessarily mined in this country, but they are easily found for sale in the market and their real provenance was not known for this study.The geographic diversity of the samples is expected to influence the results on the natural radioactivity.The samples preparation included comminution with two different types of crushers, a jaw crusher and a roll crusher, and pulverization in a pan to attain a product finer than 0.1 mm.The mass reduction stages were carefully conducted so the final 1 g sample was not biased.

Neutron activation analysis measurement
For the determination of the elemental concentrations, approximately 150 mg of each granite sample were weighed and packed in plastic polyethylene bags.Each batch of samples was irradiated together with two reference materials (RM), USGS STM-2 and NIST SRM 1646a, and a paper filter pipetted with a standard solution of the elements of interest.Each sample was calculated related to each reference material and the final report of the results is the mean value related to each RM.All samples and RM were irradiated for 8 hours in the IEA-R1 research reactor, at IPEN (Instituto de Pesquisas Energéticas e Nucleares) under a thermal neutron flux of 10 12 cm -2 s -1 .The cooling time for the elements Na, K, As, Br, Sb, La, Nd, Sm, Tb and U counting varied from 5 to 7 days and for the elements Ca, Sc, Cr, Fe, Co, Zn, Se, Rb, Zr, Cs, Ba, Ce, Eu, Yb, Lu, Hf, Ta and Th was about 15 days.
The counting time was 1 h for each sample and for the reference material.Gamma spectrometry was performed by using an EG&G Ortec Ge high pure Gamma Spectrometer detector (AMETEK Inc., USA) and associated electronics, with a resolution of 0.88 and 1.90 keV for 57 Co (122 keV) and 60 Co (1332 keV), respectively [5].An analysis of the data was carried out by using an in-house gamma ray software, VISPECT program, to identify the gamma-ray peaks.The methodology evaluation was done by cross-checking the reference materials and synthetic standards.

Radiological hazard indices
For the activity concentration of natural uranium, thorium and potassium, their specific activity of 12.437 Bq kg -1 , 4.057 Bq kg -1 and 262 Bq kg -1 , respectively, was used considering an isotopic abundance of 99.2742% for 238 U, 100% for 232 Th and 0.0117% for 40 K [6].
Radium equivalent activity (Raeq) was defined as the weighted sum of 238 U, 232 Th and 40 K activities, based on the assumption that 370 Bq kg -1 of 238 U, 259 Bq kg -1 of 232 Th and 4810 Bq kg -1 of 40 K gives an effective dose of 1.5 mGy per year, equal to 1 mSv annual effective dose [7].Raeq was calculated from the following relation (Eq. 1) suggested by Beretka and Mathew [8]: where CU, CTh and CK are the activity concentrations of 238 U, 232Th and 40 K, in Bq kg -1 , respectively.
The external hazard index (Hex), proposed by Hewamanna et al. [9], is applied for a house with walls of finite thickness, windows and doors, and can be calculated as Eq. ( 2): The radiation hazard due to Hex will be negligible if its value was less than the unity.Additionally, the internal hazard index (Hin), proposed by Krieger [10], takes into consideration the internal exposure due to radon and its short-lived decay products as a threaten to the respiratory system.It is calculated as Eq. ( 3): The absorbed dose rate (D) is related to the risk due to the amount of radiation deposited in a body per unit of time that arises from terrestrial gamma emitters.D can be derived (nGy h −1 ) from the measured activity concentrations and the following conversion factors, as given by UNSCEAR [11] and shown in Eq. ( 4): The annual gonadal dose equivalent (AGDE) is a measure of the genetic significance of the dose equivalent received by the population's reproductive organs per year [12].Within this context, the activity bone marrow and the bone surface cells were also included by UNSCEAR [13] as organs of interest.The AGDE for the rock samples here analyzed were determined by Eq. ( 5) [14,15].
The annual effective dose equivalent (AEDE) takes into consideration the adsorbed dose rate and also the time spent in contact with the radioactive sourceoccupancy factor, according to the formula given by Eq. ( 6): AEDE (µSv/y) = D (nGy/h) x 8760h x 0.8 x 0.7 Sv/Gy x 10 -3 (6) An occupancy factor of 0.8, calculated for building materials, is used considering that a person spends 80% of his time indoors.The conversion factor of 0.7 Sv/Gy converts absorbed dose rate (nGy/h) to annual effective dose equivalent in mSv/y [16].
Excess life time cancer risk (ELCR) represents the risk of fatal cancer during a life time (DL) of 70 years taking into consideration AEDE and a risk factor (RF in Sv -1 ) established as 0.05 by the ICRP 60 for stochastic effects for the public [17,18] according to Eq. ( 7):

Statistical analysis
Univariate and multivariate statistical analyses were applied to the results for data interpretation.
Pearson Correlation Coefficient, in which the correlation coefficient (r) is used to measure association strengths, was used to verify the relationship between the natural radionuclide activity concentrations and stable elements [19,20].Hierarchical cluster analyses were applied with the purpose of assembling objects based on their similarities.This goal is achieved by sorting cases into groups, or clusters, resulting in a strong association between members of the same cluster, and a weak association between members of different clusters [21].Principal component analysis was used to quantify the significant variation in the data set and to reduce the number of variables to a small number of indices.These indices retain the maximum amount of the variance and result in the retention of only the important characteristics of the original data [22,23].

Radiological and elemental characterization
For quality control of the results, Table 1 shows the certified values of elements present in the reference materials USGS-STM-2, SRM 1642a and the limit of detection.It can be seen that a good agreement was found between the certified values and the measured values for all the determined elements.this sample in consideration, the mean value and standard deviation become 70 ± 67 Bq kg -1 for this radionuclide, lower than the Brazilian mean AC.
The differences observed in the activity concentrations are not just related to the provenance of the rocks, but mainly to their mineralogical composition.Elements with a lithophile behavior such as U, Th and K concentrate in the Earth`s crust during the melting in the mantle and basalt differentiation being accumulated in felsic igneous melts [24].Higher concentrations of U are generally found in acid igneous or granitic rocks while basaltic rocks contain lower concentration [25].Thorium is 3 to 4 times more abundant than uranium in the rocks of the lithosphere [26] and sedimentary rocks generally contain only few µg g -1 of this element.On the other hand, its concentration can be up to 10 times higher in acid igneous rocks, where it can substitute the rare earth elements [27].
Fig. 3 shows the Th/U ratio in the samples.It can be seen that the ornamental rocks from Brazil tend to be enriched in Th compared to the Swiss samples.This ratio is closer to the global average (3 to 4) in the CH samples while most of the Brazilian ones present this ratio in the range of 5 to 15.
The high values of Th/U ratio in the São Paulo samples are probably due to an enrichment in Th as their concentrations are generally higher than mean values found in granites [28].For chemical characterization of the ornamental rock samples, trace elements were also determined and their concentrations are shown in Table 3. Table 4 shows the Pearson correlation coefficients obtained for these results.According to the results, Th showed a good correlation with rare earth elements (REE), Br, Hf, Na, Rb, Sb and Zr; uranium was well correlated only with Cs; and K presented good correlations with Ba, REE, Zr and a negative correlation with Ca.The good correlation among Th, Zr, and REE may indicate that these elements are probably associated with minerals such as zircon and apatite [29,30].The elevated value of Th, REE and Zr found in sample CH7 matches with the nepheline syenite mineralogy, probably due to the presence of apatite-like minerals, monazite and zirconosilicates [31].
Cluster analysis (CA) applied to the variables, Fig. 4, also shows the Th association with REE, Hf and Zr, also indicating a common origin for these elements, e. g., present in the same minerals constituents of the rocks.Potassium and U, on the other hand, are associated with elements such as Ca, Co, Zn and Fe.Fig. 5 shows the loading factors, obtained by factor analysis for the determined elements.The result agreed with the one obtained by CA and shows a high loading factor for Th and REE in the first factor, while U and K, present low loading factors for both, factor 1 and 2, being negative for U.This results indicates that the mineral bearing Th must be distinct from that bearing U and K.In addition to the uranium-bearing minerals, this element can be present in rocks by isomorphic substitution of calcium [32], what could cause the weak inverse loading factor for these elements in factor 2.
The influence of elemental composition of the analyzed rocks is highlighted in Fig. 6 in which clusters formed according to the rock type can be seen.Generally, rocks of the same petrographic classification are grouped together independently of their provenance.

Radiological risk assessment
The radiological hazard index radium equivalent activity (Raeq), external hazard index (Hex), internal hazard index (Hin), absorbed dose rate (D), annual gonadal dose equivalent (AGDE), annual effective dose equivalent (AEDE) and excess life time cancer risk (ELCR) obtained for the samples analyzed in this study are shown in Table 5.As sample CH7 showed the highest values for all radiological hazard indices determined, it will be discussed separately.Thus, citation to all samples means that CH7 is not included.
The range of Raeq varied from 5.5 to 775 Bq kg -1 in the samples CH9 and SP2, respectively, with a mean value of 299 Bq kg -1 .Fig. 7 shows the mean values for samples from Switzerland and Brazil and also for Brazilian states.The mean value (and standard deviation) of Raeq in Swiss samples was 257 ± 187 Bq kg -1 and for the Brazilian samples 325 ± 175 Bq kg -1 .Samples from Espírito Sample CH7 exceeded this value 11.5 times.For a comparison purpose, Fig. 7 shows mean Raeq obtained in samples from Pakistan, PK, [33]; Egypt, EG, [34]; Jordan, JO, [35]; and with two other Brazilian assessments, BR2 [3] and BR3 [36].As can be observed, Brazilian samples tend to present higher Raeq average values than those observed abroad.There was a good agreement in the mean values obtained in this study with the ones reported for other Brazilian samples.[37], whose measures varied from 0.23 to 3.99 for rocks, sediment and building samples in Southwestern Nigeria and by Alharbi et al. [38] who found values varying from 0.14 to 2.12 in granite rocks of Kingdom of Saudi Arabia.Considering the limit of 1 for Hex and Hin, the former was exceeded in two SP samples while the latter was exceeded in all SP samples but SP5, in BA and RJ samples, among the Brazilian ones.
Among the Swiss samples Hex was not exceeded while Hin was higher than 1 in 5 samples.CH7 presented values for external and internal hazard index 5.7 and 11.5 times higher than the recommended value, respectively.
The absorbed dose rate (D) ranged from 2.3 to 344 nGy h -1 , with a mean value of 139 nGy h -1 considering all samples.The lowest value was measured in sample CH9 and the highest one in sample SP2. this study and also values reported in the literature.Among the Brazilian samples, the ones from São Paulo (SP) and Rio de Janeiro (RJ) showed the highest values, and the lowest ones were found in Espírito Santo (ES) samples.Brazilian average was higher than the one observed in Swiss samples and also higher than that reported for PK, JO and EG.The absorbed dose rate measured is quite similar to that found in granites analyzed by Anjos et al. [39].Compared with the world average value of 58 nGy h -1 [11], only three Swiss and three Brazilian samples were below it, indicating that the majority of the analyzed samples are up to six times greater than the world average value.In the case of sample CH7 with a D of 1791 nGy h -1 , the worldwide average value is exceeded by 31 times.with the same values for literature in construction materials from Iraq (IQ) [40] and from Egypt (EG).The mean value of AGDE is higher for the Brazilian samples than that collected in Switzerland and those reported for Iraq and Egypt.Among the Brazilian samples the highest values were found in São Paulo (SP) and Rio de Janeiro (RJ) samples.Considering the average world value of 0.36 mSv y -1 for AGDE [12] (UNSCEAR, 1998) only two samples from Switzerland and three samples from Brazil have values below it.The values found in this study are up to 6.6 times higher than the world average, and sample CH7 exceeded this value by 34 times.Of the total, fifteen samples are greater than the established limit of 1 mSv y -1 [41].The results of AEDE were lower than the average annual indoor effective dose from terrestrial radionuclides of 0.41 mSv y -1 [42] in six samples from Switzerland and in five samples from Brazil.
The measured values were up to 3.6 times higher than the worldwide average value and sample CH7 was 19 times greater than this average value.Six samples exceeded the limit of 1 mSv y -1 .
The results of ELCR were lower than the average global value in only three Swiss samples and in three samples from Brazil.The values above the average were up to five times higher and for sample CH7, 26 times higher.

Multivariate statistics of the radiological risk indices
Cluster analysis was applied to evaluate the radiological indices among the analyzed samples considering the activity concentrations of 232 Th, 238 U and 40 K, and the radiological indices determined in this study.The result is showed in Fig. 11.From the CA result it is obvious that sample CH7 is highlighted, being the one with the highest values for all radiological risk indices determined among the analyzed samples.Cutting the dendrogram at a level of 20% of the 100*Dlink/Dmax, three main groups (excluding sample CH7) were observed (Fig. 11).The basic statistics for each group is summarized in Table 6.

CONCLUSION
The analysis of ornamental rocks from Brazil and Switzerland showed that the activity concentrations of thorium, uranium and potassium covered a wide range of values, varying from 2.5 to 2966 Bq kg -1 , from 3.9 to 214 Bq kg -1 and 25 to 2372 Bq kg -1 , respectively.The highest 232 Th activity concentration was found for sample CH7, from Switzerland; the highest 238 U activity concentration was measured in RJ1 sample from Rio de Janeiro, Brazil; and for 40

Figure 2 :Figure 3 :
Figure 2: Mean activity concentration according to the sample origin: CH: Switzerland, BR: Brazil.SP, ES, BA, MG and RJ stands for the Brazilian states of São Paulo, Espírito Santo, Bahia, Minas Gerais and Rio de Janeiro, respectively

Figure 4 :Figure 5 :
Figure 4: Dendrogram for the variables measured in the analyzed ornamental rocks

Factor
Santo state (ES) were the ones with the lowest Raeq mean value.The recommended value of 370 Bq kg -1 was exceeded in samples from São Paulo (SP), Rio de Janeiro (RJ) and Switzerland (CH).

Figure 6 :Figure 7 :
Figure 6: Dendrogram obtained in the cluster analysis according to the rock classification The external (Hex) and internal (Hin) hazard indices presented mean values of 0.38 and 0.995, respectively considering all samples.The values ranged from 0.011 to 1.25 for Hex, in samples CH9 and SP2, respectively, and from 0.015 to 2.45 for Hin, in the same samples.Fig.8shows that the average values of both of these indices were slightly higher for the Brazilian samples than for the Swiss ones.Among the Brazilian states, the samples from São Paulo (SP), Espírito Santo (ES) and Bahia (BA) presented very close average values while Minas Gerais (MG) samples presented the lowest ones.Comparing with values in the literature, only the samples from MG present lower values than those observed in samples from abroad.Good agreement for this parameter was also found among the Brazilian samples analyzed in this study and the values (BR2) reported by Moura et al.[3].Brazilian samples presented Hin average values higher than those from all other countries shown in Fig 8.Nevertheless high values of Hin were also found in building material byBello et al.

Figure 8 :
Figure 8: External (Hex) and internal (Hin) hazard indices, according to the samples origin (this study: CH, BR1, SP, ES, BA, MG and RJ) and literature values (PK, EG, BR2 and JO) Fig 9 shows the comparison among the D values obtained in the samples analyzed in

Figure 10 :
Figure 10: Annual gonadal dose equivalent (AGDE), annual effective dose equivalent (AEDE) and excess life time cancer risk (ELCR) according to the samples origin (this study: CH, BR1, SP, ES, BA, MG and RJ) and literature values (IQ and EG)

Figure 11 :
Figure 11: Dendrogram obtained in cluster analysis considering the activity concentrations of the radionuclides and radiological risk indices

Figure 12 :
Figure 12: Principal component analysis result with the loading factor for the first and second component for the group 1 of the dendrogram shown in Fig 11

Figure 13 :
Figure 13: Principal component analysis result with the loading factor for the first and second component for the group 2 of the dendrogram shown in Fig. 11

Figure 14 :
Figure 14: Principal component analysis result with the loading factor for the first and second component for the group 3 of the dendrogram showed in Fig. 11

Table 1 :
Certified values (CV) and measured values (MV) obtained in reference materials for quality control of the results and limit of detection (LD).Values in mg kg -1 , except where indicated (%).
232Th.The mean AC and standard deviation for all Brazilian and Swiss samples were: 62 ± 55 Bq kg -1 and 93 ± 60 Bq kg -1 for 238 U; 1126 ± 516 Bq kg -1 and 1005 ± 778 Bq kg -1 for 40 K; 122 ± 111 Bq kg -1 and 293 ± 806 Bq kg -1 for 232 Th, respectively.These values indicate that a wide range of radionuclides AC can be found in the analyzed ornamental rocks.The high mean and standard deviation for 232 Th in the Swiss samples were due to the value found in sample CH7.If not taking

Table 2 :
Activity concentration of 232 Th, 238 U and 40 K, in Bq kg -1 , of the analyzed samples, their commercial names and petrographic classification.

Table 3 :
Concentrations and uncertainties of trace elements, in mg kg -1 , except where indicated %, determined in the analyzed samples.

Table 4 :
Pearson correlation coefficient obtained using the whole data set of elemental concentrations.Marked correlations are significant at p < .050.
a RV or WV: recommended value or worldwide average value.

Table 6 :
Basic statistics of the obtained groups using cluster analysis, considering the activity concentrations of the radionuclides and radiological indices.theyare the safest samples among the analyzed ones.The values of Skewness and Kurtosis, near zero and negative, respectively, indicate that these samples form an homogeneous group with normal distribution for the radiological indices.For these samples the radiological hazard indices are mainly related to232Th concentrations as can be seen in Fig.12, with inverse correlation between 238 U and 40 K.Group 2 was formed by the samples with the highest activity concentrations and radiological hazard indices.Thus, these are the less safe samples among the analyzed ones considering gamma exposure, indicating that care must be taken in their use.This group contains foid-monzodiorite, mylonitic gneiss with chlorite and muscovite, monzogranite, syenogranite and basaltite rock types from Switzerland, Bahia, São Paulo and Rio de Janeiro.The radiological hazard indices are mainly related to 232 Th activity concentration, as can be seen in Fig.13, with 238 U and 40 K directly correlated.As group 1, this group is also an homogenous group with normal distribution for the radiological hazard indices.