Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: Precious objects are often made from a thin layer of metal (gold or silver) superimposed on another metal or
Thickness estimation alloy. Non-invasive XRF analysis is widely used on these objects to determine the composition and thickness of
Ag layers the surface layer. Here, silver surface thickness of two-layered metal samples using an ED-XRF handheld spec
Handheld XRF
trometer is estimated by different methods whose results are compared. These methods involved the relation
PLS regression
X-ray characteristic lines
ships between the intensities of the characteristic lines of Kα, Kβ, Lα and Lβ of Ag and of elements present in the
substrate (Cu, Fe, Pb) and some of their combinations, as well as the PLS regression. Reference layers were used
to derive the calibration curves of each methodology. PLS method was the most accurate, but the other methods
were also found suitable for the purpose. Applications have been successfully made on an ancient Roman coin
and a modern silver-plated object.
* Corresponding author.
E-mail address: carlo.sabbarese@unicampania.it (C. Sabbarese).
https://doi.org/10.1016/j.nimb.2020.11.006
Received 4 August 2020; Received in revised form 31 October 2020; Accepted 8 November 2020
0168-583X/© 2020 Elsevier B.V. All rights reserved.
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
74
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
2.2. Theoretical background Fig. 3. Layout of the measurement configuration for a double layer: b is Ag
layer covering another metal a.
2.2.1. Self-attenuation X-ray and attenuation by a second layer
To investigate homogeneity and two-layered structure of a sample, it characterization of Kα and Lα to provide further information on the na
is essential to determine the thickness of coating layer. For this reason, ture of the surface thickness.
the detailed analysis of the fluorescence peaks and in particular the ratio In another case, when a layer of a sample a of any thickness is
values between the characteristic lines of an element present, is very covered by a second layer b (Fig. 3), the ratio between the lines of the
useful because ratios depend on the abundance of respective elements element behind the first layer is attenuated by the covering according to
deriving from a certain depth. the following theoretical equation [25,26]:
When an incident x-ray beam interacts with the atoms of the element ( ) ( )
{ }
present in a mono-elemental sample of any thickness, the excited shells Kα
=
Kα
* e− (μ1 − μ2 )d (2)
generate secondary X-rays from a point of the material at specific en Kβ Kβ ∞
ergies. These characteristic X-ray lines (Kα, Kβ, Lα, Lβ) are partially
where:
absorbed into the layer itself before reaching the detector. The attenu
ation effect depends on the position of the element, that is the thickness, ( )
and the efficiency of the detector at the energies involved. The x-rays Kα
is the intensity ratio of the characteristic lines of the sample a
Kβ
emitted undergo two attenuation processes: self-attenuation of the ∞
element in its own layer and attenuation by the overlapping layer. Self- with infinite thickness;
attenuation can be expressed by the following equation [24]: μ1 is the linear attenuation coefficient (cm− 1) of the element at en
ergy of Kα emission line of sample a
} {( ’ )} { }
μ2 is the linear attenuation coefficient (cm− 1) of the element at en
( ) ( ) {
1 − e− (μ0 +μ1 )x
’ ’
Kα Kα ε(Kα ) μ0 + μ’2
= * * * (1)
Kβ Kβ 0 ε(Kβ ) (μ’0 + μ’1 ) 1 − e− (μ0 +μ2 )x
’ ’ ergy of Kβ emission line of sample a;
x
d is path length of the characteristic radiation (in cm) and d = cosθ2 ,
where: where x is the thickness of the layer and θ2 is the angle between the
detector and the normal to the sample.
( )
Kα
is the intensity ratio for an infinitely thin sample (that is when
Kβ
0
The Eq. (2) also applies to elements whose characteristic lines are Lα
the thickness of the sample is negligible compared to the depth of and Lβ .
penetration); From formula (2) the thickness of the coating layer d as a function of
ε(Kα), ε(Kβ) are the detector efficiency values at the energies Kα and the ratio between the Kα and Kβ lines of the substrate element is [27,28]:
Kβ; ( )
cosθ2 R1
μ0 = μ0 /cosθ1 , μ0 is the linear attenuation coefficient (cm− 1) of the
′
d = − ′′ ′′
*ln (3)
μ1 − μ2 R2
element at incident energy E0 and θ1 is the angle between the x-ray
tube and the normal to the sample surface. where:
μ1 = μ1 /cosθ2 , μ1 is the linear attenuation coefficient (cm− 1) of the
′
element at energy of Lα line and θ2 is the angle between the detector μ′′ 1 is the linear attenuation coefficient (cm− 1) of the covering layer
and the normal to the sample surface; at energy of Kα (or Lα )line;
μ2 = μ2 /cosθ2 , μ2 is the linear attenuation coefficient (cm− 1) of the
′
μ′′ 2 is the linear attenuation coefficient (cm− 1) of the covering layer
element at energy of Lβ line; at energy of Kβ (or Lβ )line;
x is the thickness of the layer (cm). R1 is the intensity ratio between characteristic lines of the element
{ } { }
′ ′ ′ ′
− μ +μ x
present in the substrate;
(μ0 +μ2 ) 1− e ( 0 0 )
The product * is self-attenuation factor. R2 is the intensity ratio between characteristic lines of the element
(μ′0 +μ′1 ) ′ ′
− μ +μ x
1− e ( 0 2 )
present in the substrate considering an infinite thickness.
If Lα and Lβ are the X-ray lines involved in the ratio, a similar The thickness of a thin layer of element b in presence of a substrate
equation is obtained. layer a (Fig. 1) can be also estimated by the ratio between the principal
This method can be further combined with the peak intensity
75
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 4. 16 XRF spectra vs Ag thickness as used for the calibration curves in the four configurations: a) Ag b) Cu covered with sheets of Ag c) Pb covered with sheets of
Ag d) Fe covered with sheets of Ag.
lines of the two elements characterizing the two layers (for example, in multivariate extensions of the multiple linear regression model because
the case of silver-plated copper, the ratio Nb/Na = AgKα /CuKα ). it can be used also in the presence of inaccurate data. This flexibility
The following theoretical equation expresses its dependence on the makes it usable in situations where other traditional multivariate
thickness of the layer b [25]: methods are limited, such as when there are many variables but not
( ) {( ’ )} ( ) necessarily many observations.
Nb μa0 + μ’aa
=P * 1 − e− (μb0 +μbb )ρb xb *(e(μb0 +μba )ρb xb
’ ’ ’ ’
(4) As multivariate calibration method, PLS regression allow relating
Na (μ’b0 + μ’bb ) thickness of the samples directly to measured XRF spectra (or portions of
them). The procedure comprises of two steps: calibration and validation.
where:
Calibration step establishes the relation between the data matrices X and
Y, while validation is an essential step because it proves the ability of the
P is the probability of detection of the line of interest. It includes the
calibration model to predict the desired thickness [29].
probability of line emission, the fluorescence yield of the element
The study of model performance consists of the evaluation of the root
and the detector efficiency at the energy of the interest line;
mean square error of calibration (RMSEC) and the root mean square
μ’a0 =μa0 /cosθ1 and μ’b0 =μb0 /cosθ1 , μa0 and μb0 are the total mass error of validation (RMSEP) [7].
attenuation coefficients (cm2/g) of the elements a and b at incident The RMSEC value represents to what extent the PLS procedure is
energy; suited for modeling the dependent variable (i.e. coat thickness) and the
μ’ba = μba /cosθ2 , μba is the total mass attenuation coefficient of the uncertainty of the reference method (spectra XRF) [30].
element b to the energy of the affected line of element a (cm2/g); The RMSEP value is the mean deviation of the prediction from the
μ’aa = μaa /cosθ2 , μaa is the total mass attenuation coefficient of the respective reference value calculated with the given model. It is an in
element a to the energy of the affected line of element a (cm2/g); dicator of the average error for coat thickness prediction on unknown
samples. Good models are characterized by low RMSEP values, close to
μ’bb = μbb /cosθ2 , μbb is the total mass attenuation coefficient of the
the error of the reference method [29].
element b at the energy of the involved line of element b (cm2/g);
Its value is a measure of the good fit between known thickness and
ρb is the density of the element b (g/cm3); predicted thickness [31].
xb is the thickness of the layer b (cm). RMSEP is expressed in the same units as the predicted parameter.
The results of future predictions can then be presented as predicted
2.2.2. The PLS regression values ± RMSEP. This measure is valid provided that the new samples
The Ordinary Partial Least Squares regression (PLS) is a quantitative are similar to the ones used for the calibration otherwise the prediction
multivariate analysis method that has been widely used for spectra data error might be much higher [32].
evaluation and it is well-known for several applications in archaeometry
[7].
PLS regression is probably the least restrictive method of the various
76
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 5. Experimental data and fit curves vs Ag layer thickness of: (a) AgKKαβ ; (b) CuKKαβ ; (c) FeKKαβ ; (d) PbLLαβ .
77
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 6. Experimental data of Ag KLαα and fit curve vs Ag thickness layer in the four configurations analyzed using reference samples. a) Ag, b) Cu covered by Ag, c) Fe
covered by Ag and d) Pb covered by Ag.
The trends of the net area of the Kα lines as a function of the thickness
Table 3
of the Ag layer and of the element behind the silver are investigated
Fit data obtained from the linear approximation curve (ax +b) of the AgKα/Lα
(Fig. 7). The differences in trends indicate how much and how the un
ratio and Ag thickness in the different materials used as a reference.
derlying metal affects the fluorescence of silver. Only Kα trends of the
a b (µm− 1) silver foils follow a growth curve. The curve for the Ag-only sample rises
Ag 0.30 ± 0.01 0.68 ± 0.01 faster than the other three. However, Ag on Cu and Fe approaches the
Cu 0.32 ± 0.01 0.5 ± 0.1 Ag-only curve towards the end.
Fe 0.33 ± 0.01 0.5 ± 0.1
The influence of the underlying material on the behavior of the
Pb 0.31 ± 0.01 0.6 ± 0.1
characteristic lines of Kα is shown in Fig. 7. Note that up to thicknesses of
4 µm the characteristic lines of the element behind are much less
decreasing exponential curve. Hence, ratio values of the characteristic attenuated by silver.
lines of copper, iron and lead (i.e. all the materials covered by the silver Another method to determine the thickness of a multi-layer sample is
layers) are fitted by following equation: obtained by the calibration curve of the ratio between the characteristic
( ) lines of two elements characterizing the layers. This is described by Eq.
Kα { }
= a* e− b*d (5) (4) and has been applied to (Ag-Kα /Cu-Kα ), (Ag-Kα /Fe-Kα ), (Ag-Kα /Pb-
Kβ
Lα ) ratios vs Ag thickness, assuming Cu, Fe and Pb of infinite thickness
( ) respectively.
Kα
with a = Kβ , b = (μ1 − μ2 ) and d = thickness of silver layers. Experimental data and fit curves are shown in Fig. 8.
∞.
Fit curves of the cross ratios data in presence of copper and lead are
In this case, the results of fit parameters (Eq. (5)) are in good
represented at all the Ag thicknesses analyzed (Fig. 8); for iron, it was
agreement with theoretical data (Table 2).
possible to use data of silver thicknesses up to 5.5 μm because the
Despite the fact that the ratio of AgKα/Lα vs Ag thickness is described
characteristic lines of iron after a such thickness are completely atten
by a growth curve, its experimental and theoretical values are well
uated (as shown in Fig. 4d)).
approximated by a linear curve for sub-micrometers thicknesses (Fig. 6).
The term P, in the Eq. (4), indicates the probability of detection of the
The fit parameters derived from the linear curve (Table 3) allow tracing
line and the element of interest, which includes the probability of
the silver thickness values more easily and effectively.
emission of a given line, the fluorescence yield of the element and the
78
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 7. Trend of Ag Kα peak area of silver (black) and Ag Kα peak area of copper (red), iron (magenta) and lead (green) that covered by silver foils vs thickness of
silver foils. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
energy efficiency of the detector to the energy of the line [11]. Due to the In the analysis of the silver thicknesses only the model with five
fact that the determination of the P value is difficult, it was derived from latent variables explained 99.67% of the variance. The determination
experimental measurements. However, it increases the uncertainty in coefficient for the calibration curve was 0.9967 and for the prediction
the theoretical trend. Table 4 shows the results obtained from the fit (Eq. data was 0.9867 (Fig. 9a). The RMSEC and RMSEP values were 0.27 and
(4)) of the experimental and theoretical values. 0.33 respectively.
All the described methods are based on an accurate analysis of K or L In the set-up of copper covered with silver the model with six latent
peaks to obtain a valid calibration curve for the thickness estimation of variables explained 99.9% of the variance. The determination coeffi
the surface layer in a multilayer object. A correct evaluation of the net cient for the calibration curve was 0.999 and for the prediction data was
area and subtraction of the background of involved peaks, and an 0.9963 (Fig. 9b). The RMSEC and RMSEP values were 0.13 and 0.30
effective separation of the peak of interest from possible contiguous, respectively.
disturbing peaks are required [6]. In the case of lead covered with silver the model with six latent
In the application of the PLS regression, sixteen XRF spectra were variables explained 99.95% of the variance. The determination coeffi
used for calibration and seven for validation; all obtained with Ag cient for the calibration curve was 0.9995 and for the prediction data
thickness known. The calibration spectra data were arranged in a 16 × was 0.9989 (Fig. 9c). The RMSEC and RMSEP values were 0.10 and 0.17
1300 matrix (X) and the thickness values in a 16 × 1 matrix (Y). The respectively.
validation data were organized in a 7 × 1300 matrix (X). This method In the case of iron covered with silver the model with four latent
was done for the four experimental set up. The whole spectrum, from 0 variables explained 99.7% of the variance. The determination coeffi
keV to 26 keV, containing all the peaks of Ag and the other elements was cient for the calibration curve was 0.997 and for the prediction data was
used (Fig. 4). 0.998 (Fig. 9d). The RMSEC and RMSEP values were 0.25 and 0.21
PLS regression curves obtained for calibration of the four set up (Ag respectively.
and Cu, Fe, Pb covered by Ag) are shown in Fig. 9. The results of the Although the prediction curves are well explained by the model.
validation samples are shown in the same graphs to show their agree The comparison between the various methods used for thickness
ment with the calibration curve. estimation is shown in Table 5.
79
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 8. Experimental data and fit curves of Ag-Kα /Cu-Kα , Ag-Kα /Fe-Kα and (Ag-Kα /Pb-Lα ) versus Ag thickness according to Eq. (4).
80
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 9. PLS regression curves for calibration and validation samples. The error bars indicate the response residuals.
81
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Table 5
Thickness values for the seven silver foils determined by the different methodologies. y_m indicates the thicknesses measured with the micrometer, y_PLS indicates the
thicknesses predicted with the PLS method, y_ AgKα/Kβ, y_AgKα/Lα indicate the thicknesses obtained from the ratio curves, AgKα/Kβ , AgKα/Lα respectively.
Thickness of the surface layer of Ag (μm)
0.4 ± 0.1 0.5 ± 0.3 0.3 ± 0.4 0.2 ± 0.3 0.4 ± 0.3
2.0 ± 0.1 2.0 ± 0.3 1.9 ± 0.4 1.8 ± 0.3 2.1 ± 0.3
3.3 ± 0.1 2.7 ± 0.3 3.0 ± 0.4 2.9 ± 0.3 3.2 ± 0.3
5.0 ± 0.4 5.2 ± 0.3 5.1 ± 0.4 5.3 ± 0.3 5.5 ± 0.3
9.5 ± 0.1 9.9 ± 0.3 9.2 ± 0.4 9.9 ± 0.3 9.8 ± 0.3
12.6 ± 0.1 12.8 ± 0.3 11.5 ± 0.4 12.6 ± 0.3 12.8 ± 0.3
14.5 ± 0.5 14.4 ± 0.3 12.9 ± 0.4 14.1 ± 0.3 15.0 ± 0.3
100 μm in a silver-copper matrix due to its higher energy. By calculating commercial handheld instrument. It is a closed system that does not
the Ag Kα /Ag Lα intensity ratios it is possible to obtain information on allow particular measurement configurations that could improve its
the degree of exhaustion. The value of this ratio can be used as an efficiency and effectiveness, but it was adequate for this investigation. A
indication of whether or not silver is enriched on the surface of a coin series of twenty-three reference silver layers with known thicknesses
[16]. was used to obtain calibration curves for the estimation of the thickness
of silver objects. A systematic determination of the intensity of Ag Kα, Kβ
and Lα and their ratios is performed for different sheet thicknesses and
3.2. Silver-plated object
for the three elements placed under the silver (copper, iron and lead).
PLS regression methodology was also used on the same spectra in order
XRF measurements are performed on the surface of untreated sam
to determinate silver thickness of the covering layers. This method uses
ples. Two XRF point measurements are performed on both side of reli
directly the spectrum file without any previous processing. Using the
gious object. In particular, the head of child was analyzed.
calibration curves obtained, the thicknesses of the silver are estimated in
Concentrations of the detected elements on both sides are different.
some objects as a result verification test. It gives consistent results within
Therefore, the plated part consists of Ag (97.6 ± 0.1)% instead the back
the expected thickness range. In fact, the surface enrichment of an
part of the samples is composed by Al (93.8 ± 0.3)%, Fe (0.434 ±
ancient silver coin belonging to the Antoninianus series and the plating
0.003)% and Ag (5.748 ± 0.008)%. The methods used for thickness
layer of a religious icon were estimated.
estimation are PLS and ratio AgKKαβ and AgKα /Lα because the Ag thickness
is very thick. The results obtained are shown in Table 8. Using the self- CRediT authorship contribution statement
attenuation calibration curve, the value attenuates arriving at the
plateau after a thickness of 10 µm so after this thickness it is not possible J. Brocchieri: Software, Formal analysis, Visualization, Investiga
to define the thickness accurately. tion, Writing - original draft. E. Scialla: Software, Formal analysis, Data
curation, Investigation. C. Sabbarese: Conceptualization, Investigation,
4. Conclusions Methodology, Validation, Supervision, Writing - review & editing.
82
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
Fig. 10. Comparison of the estimated thicknesses with the different methods of the different reference samples realised.
Coins_O 94.6 ± 3.1 ± 0.10 ± 1.9 ± 0.059 ± 0.024 ± Supplementary data to this article can be found online at https://doi.
0.3 0.2 0.04 0.2 0.008 0.001
org/10.1016/j.nimb.2020.11.006.
Coins_R 95.1 ± 3.1 ± 0.07 ± 1.5 ± 0.059 ± 0.025 ±
0.6 0.3 0.04 0.2 0.006 0.001
References
83
J. Brocchieri et al. Nuclear Inst. and Methods in Physics Research, B 486 (2021) 73–84
[13] M. Hrnjić, G.A. Hagen-Peter, T. Birch, G.H. Barfod, S.M. Sindbæk, C.E. Lesher, Non- [25] R. Cesareo, A. Brunetti, S. Ridolfi, Pigment layers and precious metal sheets by
destructive identification of surface enrichment and trace element fractionation in energy-dispersive x-ray fluorescence analysis, X-Ray Spectrom. Int. J. 37 (4) (2008)
ancient silver coins, Nucl. Instrum. Methods Phys. Res. Sect. B 478 (2020) 11–20. 309–316.
[14] A.N. Shugar. Portable X-ray Fluorescence and Archaeology: Limitations of the [26] S. Pessanha, I. Queralt, M.L. Carvalho, J.M. Sampaio, Determination of gold leaf
Instrument and Suggested Methods To Achieve Desired Results, in Archaeological thickness using X-ray fluorescence spectrometry: accuracy comparison using
Chemistry VIII, 1147, edited by R.A. Armitage and J.H. Burton (2013), chapter 10, analytical methodology and Monte Carlo simulations, Appl. Radiat. Isot. 152
173-193. (2019) 6–10.
[15] M.T. Doménech-Carbó, F. Di Turo, N. Montoya, F. Catalli, A. Doménech-Carbó, [27] R.C. Nardes, M.S. Silva, A.N.S. Rezier, F.A.C.R.A. Sanches, H.S. Gama Filho, R.
C. De Vito, FIB-FESEM and EMPA results on Antoninianus silver coins for S. Santos, E.M. Zanatta, Study on Brazilian 18th century imperial carriage using x-
manufacturing and corrosion processes, Sci. Rep. 8 (1) (2018) 1–12. ray nondestructive techniques, Radiat. Phys. Chem. 154 (2019) 74–78.
[16] R. Linke, M. Schreiner, G. Demortier, M. Alram, Determination of the provenance [28] S.A. Barcellos Lins, S. Ridolfi, G.E. Gigante, R. Cesareo, M. Albini, C. Riccucci,
of medieval silver coins: potential and limitations of x-ray analysis using photons, L. Tortora, Differential X-Ray attenuation in MA-XRF analysis for a non-invasive
electrons or protons, X-Ray Spectrom. Int. J. 32 (5) (2003) 373–380. determination of gilding thickness, Front. Chem. 8 (2020) 175.
[17] R. Klockenkämper, H. Bubert, K. Hasler, Detection of near-surface silver [29] J. Irudayaraj, C. Reh, Nondestructive testing of food quality, Vol. 18, John Wiley &
enrichment on roman imperial silver coins by x-ray spectral analysis, Archeometry Sons, 2008.
41 (2) (1999) 311–320. [30] F. Westad, F. Marini, Validation of chemometric models–a tutorial, Anal. Chim.
[18] E. Asderaki et al., in Proceedings of the 34th International Symposium on Acta 893 (2015) 14–24.
Archaeometry, Zaragoza Spain, 37 May 2004, edited by J. Perez-Arontegui 2006. [31] P. Lemberge, P.J. Van Espen, Quantitative energy-dispersive x-ray fluorescence
[19] C. Vlachou-Mogire, B. Stern, J.G. McDonnell, The application of LA-ICP-MS in the analysis of liquids using partial least-squares regression, X-Ray Spectrom. Int. J. 28
examination of the thin plating layers found in late Roman coins, Nucl. Instruments (2) (1999) 77–85.
Methods Phys. Res. Sect. B Beam Interact. Mater. Atoms 265 (2007) 558–568. [32] A.M.C. Davies, T. Fearn, Back to basics: calibration statistics, Spectrosc. Eur. 18
[20] C.J. Keturakis, et al., Analysis of corrosion layers in ancient Roman silver coins (2006) 31–32.
with high resolution surface spectroscopic techniques, Appl. Surf. Sci. 376 (2016) [33] R. Cesareo, M.A. Rizzutto, A. Brunetti, D.V. Rao, Metal location and thickness in a
241–251. multilayered sheet by measuring Kα/Kβ, Lα/Lβ and Lα/Lγ X-ray ratios, Nucl.
[21] G.M. Ingo, et al., Roman sophisticated surface modification methods to Instrum. Methods Phys. Res. Sect. B Beam Interact. Mater. Atoms 267 (2009)
manufacture silver counterfeited coins, Appl. Surf. Sci. 421 (2017) 109–119. 2890–2896.
[22] Metcalf, W. (2016). Analyzing the silver coinage from Nero to Trajan - KEVIN [34] R. Vitale, J. Brocchieri, C. Sabbarese, Gli incusi della collezione numismatica del
BUTCHER and MATTHEW PONTING, with contributions by Jane Evans, Vanessa museo campano: presentazione preliminare, Annali Istituto Italiano di
Pashley and Christopher Somerfield, THE METALLURGY OF ROMAN SILVER Numismatica (2019) 33–75.
COINAGE: FROM THE REFORM OF NERO TO THE REFORM OF TRAJAN [35] J. Brocchieri, R. Vitale, C. Sabbarese, Characterization of the incuse coins of the
(Cambridge University Press 2014 [2015] Journal of Roman Archaeology, 29, 742- Museo Campano in Capua (Southern Italy) by X-ray fluorescence and numismatic
746. doi:10.1017/S104775940007269X. analysis, Nucl. Instrum. Methods Phys. Res., Sect. B 479 (2020) 93–101.
[23] V. Solé, E. Papillon, M. Cotte, P. Walter, J. Susini, A multiplatform code for the [36] Vasiliki Kantarelou, et al., X-ray Fluorescence analytical criteria to assess the
analysis of energy-dispersive X-ray fluorescence spectra, Spectrochim. Acta B 62 fineness of ancient silver coins: application on Ptolemaic coinage, Spectrochim.
(2007) 63–68. Acta Part B 66 (9-10) (2011) 681–690.
[24] R. Cesareo. X-ray physics, in: La Rivista del Nuovo Cimento, Publ. Compositori,
Bologna, 2000, 1–231.
84