Analytica Chimica Acta 813 (2014) 15–24

Contents lists available at ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Laser-induced breakdown spectroscopy of archaeological findings with calibration-free inverse method: Comparison with classical laser-induced breakdown spectroscopy and conventional techniques R. Gaudiuso a,b,∗ , M. Dell’Aglio b , O. De Pascale b , S. Loperfido a,c , A. Mangone a , A. De Giacomo a,b a

Department of Chemistry, University of Bari, via Orabona 4, 70126 Bari, Italy CNR-IMIP, Section of Bari, via Amendola 122/D, 70126 Bari, Italy c National Archaeological Museum of Egnatia, Str. Pr. Savelletri, Capitolo, 72015 Fasano (BR), Italy b

h i g h l i g h t s

g r a p h i c a l

a b s t r a c t

• A novel calibration-free LIBS method (inverse method) was developed and tested. • The inverse method was applied to the analysis of a set of archaeological findings. • The inverse method was validated by comparison with LA-ICP-MS and classical LIBS.

a r t i c l e

i n f o

Article history: Received 8 October 2013 Received in revised form 30 December 2013 Accepted 9 January 2014 Available online 15 January 2014 Keywords: Calibration-free laser-induced breakdown spectroscopy Inverse method Archaeological findings Copper-based alloys

a b s t r a c t A modified version of the calibration-free (CF) method was applied to the analysis of a set of archaeological brooches made of various copper-based alloys and coming from the archaeological site of Egnatia (Apulia, Southern Italy). The developed methodology consists in determining the plasma temperature by reversing the set of equations employed in the usual CF algorithm, and it is thus referred to as “inverse method”. The plasma temperature is determined for one certified standard, by using its known elemental composition as an input data, and then applied to the set of unknown samples to evaluate their composition in a CF mode. The feasibility of such an approach is demonstrated by comparing the results obtained with classical LIBS (drawing calibration lines with a series of matrix-matched certified standards) and with independent measurements performed with a conventional technique (LA-ICP-MS). © 2014 Elsevier B.V. All rights reserved.

1. Introduction Laser-induced breakdown spectroscopy (LIBS) has been continuously attracting attention from many diverse fields of materials analysis, due to several well-known practical advantages, which

∗ Corresponding author at: CNR-IMIP, Section of Bari, via Amendola 122/D, 70126 Bari, Italy. Tel.: +390805929513. E-mail address: [email protected] (R. Gaudiuso). 0003-2670/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.aca.2014.01.020

include: possibility of fast multi-elemental analysis of virtually any kind of sample, irrespective of its state of aggregation and chemical nature; no need of sampling or important sample preparation procedures; possibility of in situ and of remote analysis, even in extreme conditions of pressure and temperature, as well as under water [1–6]. One of the LIBS applications that benefits most from such features, as well as from its micro-destructivity, is that of the analysis of cultural heritage samples [4,7]. In the last decade LIBS has been successfully employed to the analysis of historical buildings [8,9], paintings and pigments [10–12], metals [13,14], even in

16

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24

Table 1 Composition and kind of the employed certified samples. Standard sample

wt% Sn

wt% Pb

wt% Zn

wt% Cu

wt% Ni

Alloy

Manufacturer

B3 B4 B21 B22 HPb LPb CSM LH11 L3 UZS SN1 SN2

12.96 11.05 5.13 3.85 5.30 9.80 7.90 0.44 1.50 0.40 11.75 13.54

1.65 2.50 3.79 6.12 5.60 0.77 1.80 1.26 1.02 0.57 5.17 1.97

2.27 1.21 6.17 4.40 5.90 0.47 – 26.20 32.70 15.30 0.804 1.28

80.25 84.00 83.05 82.75 82.40 88.20 90.10 66.80 62.35 78.98 79.93 82.80

1.53 0.57 1.21 2.56 – – – 2.91 0.90 0.49 2.17 0.104

Bronze Bronze Bronze Bronze Bronze Bronze Bronze Brass Brass Brass Bronze Bronze

TechLab, Metz, France TechLab, Metz, France TechLab, Metz, France TechLab, Metz, France Homemade Homemade Homemade TechLab, Metz, France TechLab, Metz, France TechLab, Metz, France MBH, England MBH, England

submerged environment [15–18], ceramics [19–21], also coupled with Raman spectroscopy [22], and to monitor the laser cleaning procedure of a variety of different samples from dirt and protective coatings [23–25]. Another inherent advantage of LIBS is the possibility of performing semi-quantitative and quantitative analyses in calibration-free (CF) mode. This methodology, first proposed in [26] and extensively studied since ([27] and references therein), does not require the use of calibration standards and thus appears particularly well suited for the analysis of samples with complicated and highly inhomogeneous matrices, such as meteorites [28,29], minerals [30–32] and archaeological objects [33], for which suitable matrix-matched standards may be not available. In particular, CFLIBS appears as a suitable candidate for routine museum analyses, e.g. for samples pre-screening to establish their matrix and as a fast micro-destructive method for quantitative determination of major and minor elements. CF-LIBS relies on the assumption of local thermodynamic equilibrium (LTE) and on the plasma parameters (i.e. excitation temperature and electron density), that are determined experimentally. Several variants of the classical CF method have appeared in the literature (see [27] and references therein). The aim of this work was to test an alternative approach to CF-LIBS (in the following referred to as “inverse method”), that was first presented in [30] and discussed and validated in [34]. The inverse method is based on the usual set of assumptions of analytical LIBS, as well as on the LTE assumption, but it introduces a further practical assumption, i.e., that if different samples with similar matrices are ablated in the same conditions, the excitation temperature of the produced plasmas is the same. In the present paper this method is applied to the determination of the composition of a set of ancient brooches. The obtained results are compared with those of classical LIBS (i.e. obtained by drawing calibration lines for the investigated elements with a series of matrix-matched standards) and of a conventional technique (laser ablation-inductively coupled plasma-mass spectrometry, LA-ICP-MS) that were used for analysing these brooches in a series of independent experiments [35].

entrance slit of the spectrograph by means of a mirror placed at 45◦ with respect to the plasma propagation axis and a 7.5 cm collection lens. The time parameters were 1 ␮s delay time (td ) and 5 ␮s gate width (tg ). Calibration lines were drawn for the main elements with certified samples of copper-based alloys, whose composition is reported in Table 1. The standard samples were rotated constantly during the ablation, so to offer a fresh surface to the incoming laser beam. The archaeological samples were kept fixed to limit their damage, but two or more spots were analyzed for each object, so as to detect and account for possible composition inhomogeneity. Moreover, in order to ablate a sample layer as representative as possible of the actual bulk, about 100 laser shots were focused on the surface prior to each acquisition, so to locally remove the external layers of the corrosion patina. The estimated thickness of the layer removed by each laser shot was around 5 ␮m in the employed experimental conditions. LIBS spectra of sample 4207 in the spectral windows used for the analysis are shown as examples in Fig. S1(a)–(d), Supplementary material.

2. Experimental

A set of brooches (fibulae), dating from VI century BC to VI–VII century AD and coming from Egnatia (BR, Italy), were analyzed. Egnatia, one of the most important archaeological sites in Southern Italy, was inhabited from the XVI Century BC up to the Middle Ages and had a central role in Mediterranean commercial trades by land, because of the via Traiana, and by sea, because of its harbour. Moreover, during Late Roman and Mediaeval Age, being Episcopal see, it became a place of transit towards the Holy Land [36–38]. Detailed archaeological information about analyzed fibulae, previously studied with typical archaeometric analyses, are reported in [35]. Images of the samples and of the spots where the laser was focused for the LIBS analysis are reported in Section 4 and subsections, along with the relative discussion of the obtained results. The measurements here presented are focused on the main elements, i.e., Cu, Sn, Pb, and, when present, Zn and Ag, while trace

2.1. LIBS setup and acquisition conditions The employed apparatus is a standard LIBS setup which comprises: a 7 ns Nd:YAG laser (Giant, Quanta System) operated at its second harmonic (532 nm) with repetition frequency of 10 Hz and energy of 13 mJ; a spectrograph (TRIAX 550, JobinYvon, 1800 gr mm−1 grating, 0.011 nm nominal spectral resolution at 312.57 nm) and an ICCD (i3000 JobinYvon) for the detection of the radiation emitted by the plasma; a pulse generator (Stanford Inc. DG 535) to synchronize the plasma emission and the spectra acquisition. The laser beam was focused on the sample surface with a 10 cm focusing lens, which provided an ablation spot of ∼200 ␮m. The radiation emitted by the plasma was coupled directly with the

2.2. LA-ICP-MS setup and acquisition conditions The LA-ICP-MS experiments were performed with a New Wave Research UP-213AI Nd:YAG 4-ns laser, operated at the fifth harmonic (213 nm) and with a 55 ␮m spot diameter for the sampling and a Thermo Electron X series I ICP-MS. LA-ICP-MS measurements were carried out on sections (of few mm3 ) sampled from the specimens so as to be representative of the bulk of the objects, not on the whole object. A large number of spectra was acquired (lined at least 55 ␮m wide and 400 ␮m long) and averaged so to account for the inhomogeneity of the sample and of the dimensions of the laser spot. Further details about the employed experimental procedures and setup may be found in [35]. 2.3. Samples

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24

elements were not investigated in details, but only qualitatively detected.

3. Results and discussion 3.1. Calibration-free LIBS methods for quantitative analysis In addition to the classical assumptions of analytical LIBS (i.e. stoichiometric ablation and homogeneous plasma), CF methods for quantitative analysis rely on the validity of the LTE assumption. A detailed discussion about the most appropriate criteria to assess the validity of this assumption is out of the scope of the present paper, and the interested reader may refer to [39,40], where this subject is extensively addressed. The experiments here reported are based on some of the conclusions drawn in the mentioned papers, i.e., that laser-induced plasmas are recombining plasmas whose partial departure from LTE can be ascribed to deviations in ionization/three-body recombination equilibrium. Thus, as suggested in [30], an appropriate time window was selected for the acquisition of spectra, such that the ionization reaction quotient approached the thermodynamic equilibrium constant of the ionization/recombination equilibrium, providing at the same time the best compromise in terms of signal-to-noise ratio. The usual equations to determine the elemental composition of the ablated sample with CF methods are reported below: N0,at G =

Iul Z(T ) exp gu Aul hvul

N0,ion G = 2

Nat Zion Ne Zat

E  u

(1)

kT

 2m kT 3/2 e h2

 E  I

exp −

kT

(2)

N0,tot G = N0,at G + N0,ion G

(3)

N0,i Mi ˙i N0,i Mi

(4)

wt%i =

The meanings of the terms in Eqs. (1)–(4) are specified as follows: N0,at and N0,ion are the relative number densities of atoms and ions of a given element; T is the plasma temperature; Iul is the emission intensity of a given transition from the upper level “u” to the lower level “l”; gu , Aul , ul and Eu are, respectively, the degeneracy of the upper level, the Einstein coefficient of spontaneous emission for the “ul” transition, the emitted photon frequency and the energy of the upper level; Zat and Zion are the atomic and ionic partition functions; Ne is the electron density; EI is the energy of first ionization; Mi is the molar mass of species “i”; G is a not determined experimental factor that is eliminated through the normalization procedure (Eq. (4), for further details see for example [28]). Electron density can be determined from the Stark broadening of emission peaks [41], and in this work Pb I lines were used for this purpose, using reference values taken from [42]. As for plasma temperature determination, the most straightforward method is the well-known Boltzmann plot, which mainly consists in plotting the logarithm of the emission intensity of an appropriate number of transitions of a given species, each normalized by their spectroscopic constants, as a function of excitation energy. A line is obtained from whose slope the plasma temperature is easily determined. A variant of this method employs both atomic and ionic lines, by also taking into account the plasma electron density. This allows extending the range of energy of the abscissa to values beyond the ionization limit, which can help improving the accuracy of the temperature determination, and the obtained linear plot is usually referred to as Saha–Boltzmann plot [43].

17

3.2. Plasma temperature determination and quantitative analysis with the calibration-free inverse method Several experimental issues can prevent the use of the described for plasma temperature determination. First of all, the number of detectable transitions may be insufficient to draw a line, or they may span too narrow an energy range to provide a truly reliable estimation of the whole energy distribution function. Moreover, different apparent excitation temperatures may be observed for the different elements. All these issues are often observed when analysing copper-based alloys, which can prevent CF quantitative analyses of this kind of samples. In [34] we proposed an alternative method for the determination of plasma temperature of samples of known concentration, which was based as well on the LTE assumption and on the usual closure equation of the CF algorithm. This was referred to as “inverse method”, because it employs the composition of a known target to obtain temperature, in contrast to what is normally done in CF for quantitative analysis of unknown targets (that starts from the temperature to retrieve the weight percentage). The inverse method basically consists in applying Eqs. (1)–(4) and repeating the calculation at different temperatures within a plausible range for LIP, in order to obtain, for every element, a set of weight percentages at the different temperatures. At the end of the simulation, the actual plasma temperature is assumed as the one providing the lowest discrepancy between the determined values and the certified ones, wt% = abs(wt% cert − wt%(T )/(wt% cert)), averaged over all the elements included in the calculation. (A similar quantity was used in [44], a modelling study about the stoichiometry of laser ablation of copper-based alloys with different lasers, and referred to as fractionation ratio.) Temperature values obtained with the inverse method in different experimental conditions (i.e. different laser pulse duration and laser wavelength) were demonstrated to be consistent with literature, i.e. lower for plasmas induced by lasers with shorter wavelengths and pulse amplitude [34]. In this work, the effect of acquisition time parameters was also studied, so to check if the inverse method could be employed for optimizing experimental conditions prior to the analysis of actual unknown samples. In Fig. 1, curves of the discrepancy wt% (T) as a function of temperature are reported for two series of experiments, which were performed at increasing delay time from the laser pulse and with two different gate widths, 500 ns (a)) and 5 ␮s (b)). Plasma temperature at each delay time is determined as the value minimizing the discrepancy of the calculated concentrations with the certified ones, and it can be graphically evaluated in a very straightforward way, as the minimum of the wt% (T) curve. The obtained values are plotted in the inset of each figure as functions of time, so to highlight the temporal trend of the plasma temperature itself. Also in this case, physically sound results are obtained, because temperature decreases, as expected, with increasing delay and when using longer gate. Two further facts are worth noting for what concerns the effect of time parameters on the average discrepancy: first, when gate width is short, the average discrepancy and the temperature spread (i.e. the width of the curves of discrepancy vs temperature) are in general much higher than those obtained, at the same delay time, with longer gate; second, higher discrepancy and temperature spread are associated to acquisitions with short delay time, as is particularly evident in Fig. 1(a) (gate 500 ns). The temperature spread can indicate qualitatively how appropriate it is the assumption of imposing a single excitation temperature to all the elements present in the plasma (and thus, it may be regarded to as a qualitative check also of the LTE assumption itself). Therefore, the results shown in Fig. 1(a) and (b) may be considered as a pre-analysis aimed at selecting the most appropriate experimental conditions, not only from the point of view of the signal-to-noise ratio, but also from that of the validity of the employed theoretical and practical assumptions. Based on these and on previous

18

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24 Table 2 List of transitions employed for plasma temperature determination with the inverse method.

Fig. 1. (a) and (b): Average wt% vs T curves for sample CSM at increasing delay time and at different gate widths (a) 500 ns and (b) 5 ␮s. In the inset are reported the temperatures corresponding to the minima of the curves as functions of delay time, so to show the temporal trend of temperatures determined with the inverse method.

results, it appears that the value obtained with the inverse method can be considered as the actual plasma temperature and that it can be employed for CF-quantitative analysis. The described procedure allows optimizing the available data and, to a certain extent, minimizing the effect of moderate experimental issues on the determined temperature (such as, for example, instrumental calibration problems, inaccuracies in the determination of the emission intensity, near-LTE conditions, limited self-absorption). It is anyway important to underline that the inverse method is an LTE-based approach, thus it cannot make up for strong deviations from equilibrium or for important self-absorption effects, which would alter the energy distribution function of the plasma species and make the Boltzmann distribution inappropriate to describe it. Therefore, in the conditions employed for the quantitative analysis, the validity of the LTE assumption was checked with the usual McWhirter criterion (though it is only a necessary, not sufficient, condition for LTE [39]). The McWhirter critical value for electron density was 1.3 × 1016 cm−3 , while the measured electron density was higher than 5 × 1016 cm−3 for all the experiments. Moreover, the degree of self-absorption was determined with the duplicating mirror technique [45], in order to rule out any significantly self-absorbed transition lines from the calculation. The wavelengths of the transitions employed for temperature determination are listed in Table 2, and in the experimental conditions of this work their estimated self-absorption was not higher than 5%. After the plasma temperature determined with the inverse method is proved meaningful, a further stepo on is necessary to use it for CF quantitive analysis. This step consists in assuming that plasmas of different samples of similar typology and ablated in the same conditions have the same temperature. The validity of this reasonable assumption was discussed in [34] and further checked in this work by

Species

 (nm)

Aul × 108 (s−1 )

gu

Eu (eV)

Reference

Sn I

300.91 303.41 326.23 333.06

0.371 1.51 3.309 0.1584

3 1 3 5

4.33 4.30 4.86 4.79

[46] [46] [46] [46]

Pb I

280.20 282.32 287.33 367.15

1.6 0.3044 0.4151 1.114

7 5 5 3

5.74 5.71 5.64 6.04

[47] [46] [46] [46]

Cu I

301.08 303.61 306.34 312.61 312.86

0.01298 0.02428 0.015 0.7617 0.2213

6 4 4 4 6

5.51 5.72 5.69 8.80 8.94

[46] [46] [46] [46] [46]

Zn I

307.21 330.26

0.1702 1.07

3 5

8.11 7.78

[46] [46]

Ni I

300.36 303.79 305.43 305.76

0.69 0.28 0.4 1

5 7 5 3

4.24 4.11 4.17 4.27

[47] [47] [47] [47]

determining the plasma temperature for the various standards with the described method. Results are shown in Fig. 2, which confirms that even if the target composition differs significantly (i.e. Zn 0–30%, Cu 62–90%, Sn 0.4–13%), if the ablation conditions are the same, the temperatures determined with the inverse method differ at most of about 10%. Table 3 shows the results obtained imposing the temperature determined for CSM to the other certified samples. The good agreement with certified values consequently implies that the temperature determined for a reference sample can also be assumed for an unknown one with similar (not necessarily identical) matrix, provided that the ablation conditions are the same. The main practical advantages of this approach are that all elements can be determined, provided that suitable emission peaks for CF analysis can be detected; that only one standard is required, and that it does not have to be perfectly matrix-matched (e.g., a bronze standard can be used for analysis of brass samples; standard samples which do not contain Ag can be used for samples which do). For the latter reason, the CF inverse method can be regarded to as halfway between the classical calibration methods and the pure CF ones. 3.3. Archaeometric applications of the inverse method The focus of the present work is applying the inverse method to the systematic analysis of a set of ancient brooches and comparing

Fig. 2. Average wt% vs T curves for certified bronze and brass samples. The plasma temperature of each sample is shown in the figure legend and was obtained as the temperature corresponding to the minimum of the relevant wt% (T) curve.

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24 Table 3 Comparison between the certified composition of certified samples and the one determined with the CF inverse method (the employed plasma temperature was determined with sample CSM). Sample

Element

Certified wt%

B22

Sn Pb Cu Zn Ni

3.85 6.12 82.75 4.4 2.56

wt% CF 3.6 6 84 4.1 2.0

CSM

Sn Pb Cu

7.9 1.8 90.1

7±1 1.4 ± 0.3 92 ± 14

B3

Sn Pb Cu Zn Ni

12.96 1.65 80.25 2.27 1.53

10 2 82 3.6 2.0

± ± ± ± ±

2 0.4 12 0.7 0.3

B4

Sn Pb Cu Zn Ni

11.05 2.5 84 1.22 0.57

10 3.2 83 3.1 1.00

± ± ± ± ±

2 0.6 12 0.6 0.15

UZS

Sn Pb Cu Zn Ni

0.4 0.57 78.98 15.3 0.49

0.36 0.24 78 17 0.35

± ± ± ± ±

0.05 0.05 12 3 0.05

L3

Sn Pb Cu Zn Ni

1.5 1.02 62.35 32.7 0.9

1.9 1.6 57 38 1.2

± ± ± ± ±

0.3 0.3 9 8 0.2

± ± ± ± ±

0.5 1 13 0.8 0.3

19

the results not only with classical LIBS, but also with completely independent measurements carried out with a conventional technique, i.e. LA-ICP-MS [35], following the procedure described in [48]. As it could be expected, most samples were in a comparatively bad state of conservation, with surfaces altered by encrustations, patinas and corrosion layers due to the burial conditions. For this reason, in [35] the LA-ICP-MS investigation was conducted on small sections of the samples, so to analyze the unaltered part of the metal. For the LIBS experiments performed in this work, a different protocol was followed, namely, the laser beam was focused directly on the surface of the samples, without any sample preparation. The purpose of this operation mode was to simulate in situ procedures (i.e. in a museum or even directly on the archaeological site, by using portable LIBS instrumentation, see for example [49]), as well as to check if reasonable results could be obtained with the CF inverse method, for a correct classification of the kind of alloy. The LIBS investigation was thus narrowed down to the main elements characterizing the analyzed alloys, i.e. Sn, Pb, Cu, Zn, Ag. Two LIBS methods were employed to carry out the analyses, i.e., calibration line (normalizing the analytic signal to Cu or to the background) and CF with the inverse method, and the correlation between the two was checked. Ag and Cu were not quantified with calibration lines, the first because its amount in the standards was not certified, the second because in this work, consistently with analogous observations reported in the literature [50], its calibration curves in copper-based alloys were found not to be linear (instead, Cu was used for normalizing the signals of the other elements). Thus, for these elements a sort of reference value to compare with the CF data was obtained by using an intermediate approach, i.e., in a CF mode and a modified closure equation, which is reported below: wt%j (T ) =

N0,j Mj N0,cu Mcu + N0,Ag MAg

× (100 − ˙i wt%i )

(5)

Table 4 Comparison between results of calibration line LIBS, CF inverse method LIBS and LA-ICP-MS for samples 2764, 4671 and 186. Data indicated with an asterisk are affected by high experimental uncertainty and should thus be considered only semi-quantitative. Cu weight percentage reported in the “Calibration line” column was determined with the mixed approach described by Eq. (5), and is reported in italic to distinguish it from the weight percentages determined with proper calibration lines. Sample

Spot

Element

wt% calibration line

2764

1

Sn Pb Cu Sn Pb Cu Sn Pb Cu

7.57 0.5 92 9.57 0.6 90

Sn Pb Cu Sn Pb Cu Sn Pb Cu

18.04 1.60 79 10.94 0.81 87 12.43 1.53 86

0.12 0.25 16 0.06 0.21 18 0.06 0.25 17

21 1.4 78 15 0.7 84 15 0.7 84

± ± ± ± ± ± ± ± ±

4 0.3 16 3 0.1 17 3 0.1 17

Sn Pb Cu Sn Pb Cu Sn Pb Cu Sn Pb Cu

11.6 ± 0.1 1.2 ± 0.3 87 ± 9 13.7 ± 0.2 2.1 ± 0.3 84 ± 8 21.0 ± 0.3 1.969 ± 0.28 77 ± 7

12.7 1.6 86 15.3 3.6 80 23 2.0 74

± ± ± ± ± ± ± ± ±

0.9 0.3 9 0.9 0.7 8 5 0.4 15

2

4671

1

2

3

186

1

2

3

± ± ± ± ± ±

0.06 0.3* 19 0.07 0.3* 16

wt% CF 12 0.2 88 13 0.46 86

± ± ± ± ± ±

wt% LA-ICP-MS 2 0.1* 18 2 0.15 15 6.7 ± 0.2 0.281 ± 0.02 93.0 ± 0.2

± ± ± ± ± ± ± ± ±

7.18 ± 0.05 3 ± 0.1 89.7 ± 0.1

5.0 ± 0.3 0.62 ± 0.04 94.0 ± 0.2

20

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24 Table 5 Comparison between results of calibration line LIBS, CF inverse method LIBS and LA-ICP-MS for sample 15031. Cu weight percentage reported in the “wt% calibration line” column was determined with the mixed approach described by Eq. (5), and is reported in italic to distinguish it from those determined with proper calibration lines. Spot

Element

wt% calibration line

1

Sn Pb Cu

25.9 ± 0.2 1.1 ± 0.3 73 ± 15

wt% CF 26 ± 5 0.7 ± 0.1 73 ± 14

2

Sn Pb Cu

30.0 ± 0.2 1.2 ± 0.3 68 ± 14

27.6 ± 5.5 0.8 ± 0.2 71 ± 14

3

Sn Pb Cu

24.8 ± 0.2 1.1 ± 0.3 75 ± 15

23 ± 6 0.74 ± 0.15 76 ± 15

4

Sn Pb Cu

10.0 ± 0.1 0.6 ± 0.1 90 ± 18

10 ± 2 1.2 ± 0.2 89 ± 18

wt% LA-ICP-MS

10 ± 0.1 0.585 ± 0.003 89 ± 0.8

metallic alloys, because of low-purity raw materials and of nonperfect control of metallurgical processes themselves. Thus, two or more spots were analyzed for each object, in order to detect possible inhomogeneity. Moreover, the samples’ external layers, up to some hundred microns, were removed by focusing 100 preparation laser shots prior to each acquisition. This was consistent with the average thickness of patinas in most of the Egnatia brooches, estimated in [35] with SEM microscopy. Anyway, the amount of removed material can be reasonably expected to depend on the patina compactness, structure and kind, which can be different for the various objects, and may change even within the same object. Moreover, it is well known that in the patina the concentration of some elements can differ significantly from that of the bulk, thus if the corrosion layer is not completely removed in the analyzed spot, the obtained results may be not entirely representative of the bulk concentration of the given element. Such effect has been described in bronze alloys chiefly for Sn, that can be enriched, but also depleted, according to the patina kind [51,52]. As for the bulk of the objects, another possible source of inhomogeneity is that Pb is scarcely soluble in the Cu matrix, thus it may accumulate as globules in the metal alloy [53]. Finally, the manufacturing processes for producing the jewels (hammering, hardening, etc.) can also be responsible for some differential distribution of elements in the matrix, e.g. due to corrosion processes taking place preferentially at mechanically weakened sites on the sample surface [54–57]. Fig. 3. (a)–(c) Correlation plots of CF-LIBS and calibration line LIBS for samples 2764 (a), 4671 (b) and 186 (c).

In Eq. (5), Ag and Cu weight percentages are calculated by subtracting the sum of the weight percentages of all the other elements, as determined with the calibration line method, instead of normalizing to 100%. 3.3.1. Archaeometric issues Prior to the actual discussion of results, we address in the following some of the issues inherent to the analysis of the archaeological samples, which should be taken into account when the analysis is performed with micro-destructive point techniques such as LIBS, all the more so when pursuing a comparison with another point technique, i.e. LA-ICP-MS. First of all, archaeological samples are usually very inhomogeneous, particularly on the surface, because of the above-mentioned chemical and morphological alterations related to the burial conditions. Moreover, the ancient manufacturing techniques would likely produce intrinsically inhomogeneous

Fig. 4. SEM image of a section of sample 833, showing the presence of Pb agglomerates (in white) and patina thickness up to 1 mm.

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24

21

Table 6 Comparison between results of calibration line LIBS, CF inverse method LIBS and LA-ICP-MS for sample 833. Cu weight percentage reported in the “Calibration line” column was determined with the mixed approach described by Eq. (5), and is reported in italic to distinguish it from the weight percentages determined with proper calibration lines. Spot

Element

wt% calibration line

1

Sn Pb Cu Zn

18.04 5.0 76 0.54

± ± ± ±

0.25 0.3 11 0.09

16 ± 3 3.4 ± 0.6 80 ± 12 0.22 ± 0.03

2

Sn Pb Cu Zn

32.00 30.2 37 0.54

± ± ± ±

0.50 2.8 4 0.09

27 30 43 0.07

± ± ± ±

2 3 6 0.01

3

Sn Pb Cu Zn

31.5 16 51 0.54

± ± ± ±

0.5 1 7 0.09

27 11 61 0.14

± ± ± ±

2 1 8 0.05

4

Sn Pb Cu Zn

11.3 ± 0.1 1.1 ± 0.1 88 ± 12 –

3.4. Discussion of archaeometric results In the following sub-sections, the obtained results will be discussed outlining some correlated archaeometric issues that can help significantly to interpret the reported data. The latter are presented in two ways, i.e., tables and plots, so to underline different aspects of the problem at issue. In the tables, LIBS data with the two methods are reported, so to compare their results with each other and with those of LA-ICP-MS obtained for the same samples, as well as to show separately the compositions of the various analyzed portions of the samples. When LIBS and LA-ICP-MS spectra were acquired for the very same portion of sample, the relevant data are compared for the portion itself. Otherwise, LAICP-MS data are reported in the table without a direct comparison with any of the analyzed sample portions. Moreover, the weight percentages obtained with the CF inverse method are plotted as functions of the calibration line ones, in order to display the correlation between the two series of data, and thus between the two LIBS methods. Each correlation plot contains an inset where only the low-concentration elements are reported, in order to highlight the correlation between the two methods on a more narrow range and without the smoothening effect of high-concentration ones (i.e. Cu and Ag). All the correlation plots display high correlation coefficient, very close to unity, thus we have reported them, as examples, only for the first set of samples. Correlation plots of the other samples are reported in the Supplementary material section. 3.4.1. Samples 2764 (VI cent. BC), 4671 (end of VI cent. BC), 186 (V–IV cent. BC) These samples have in common similar composition, (2764 and 186 are bronzes, 4671 is a leaded bronze), state of conservation, thickness and kind of patina. When compared with the results obtained with LA-ICP-MS, all three of them show a substantial agreement, with the main discrepancy arising in samples 4671 and 186, where Sn results are overestimated (see Table 4). This is probably due to their bad state of conservation and type of patina (kind II, with estimated thickness of several hundred microns up to one millimetre, and enriched in Sn, as resulted from SEM-EDS measurements), which did not allow a complete elimination of the patina by the preparation laser shots. The agreement between the two LIBS methods appears good for all three samples, as shown by the high correlation coefficients of the correlation plots, reported in Fig. 3(a)–(c).

wt% CF

wt% LA-ICP-MS

11.9 ± 0.9 4.6 ± 0.4 84 ± 11 –

15.7 0.63 83.5 0.072

± ± ± ±

0.4 0.02 0.4 0.002

3.4.2. Sample 15031 (VI cent. BC) The 15031 sample is a typical bronze. The agreement between the two series of LIBS data is very good, while the main difference with the LA-ICP-MS results is due to a severe overestimation of Sn in spots 1–3 (see correlation plot in Fig. S4, Supplementary material, and Table 5). This is most likely due to the mentioned Sn-enrichment patina effects [51] which can affect mostly LIBS measurements due to the complete lack of sample preparation. This hypothesis is confirmed by the results obtained in spot 4, where the LA-ICP-MS had been performed as well. This spot was taken on the sampled section, thus it corresponds to a deeper horizon of the sample itself and is more representative of the very bulk of the sample. The substantial agreement between LIBS and LA-ICPMS shows clearly that the strong overestimation of Sn found in the other spots is patina-related. 3.4.3. Sample 833 (I cent. BC) Results of the elemental analysis of sample 833, displayed in the correlation plot in Fig. S6, Supplementary material, and Table 6, must necessarily be interpreted by associating them to microstructural investigation by SEM-EDS. The first three examined spots display important overestimation of Sn and, in particular, of Pb. While in the previously examined bronzes, the found percentages

Table 7 Comparison between results of calibration line LIBS, CF inverse method LIBS and LA-ICP-MS for sample 4230. Data indicated with an asterisk are affected by high experimental uncertainty and should thus be considered only semi-quantitative. Cu weight percentage reported in the “Calibration line” column was determined with the mixed approach described by Eq. (5), and is reported in italic to distinguish it from the weight percentages determined with proper calibration lines. Spot

Element

wt% calibration line

wt% CF

1

Sn Pb Cu Zn

0.6 ± 0.1 0.61 ± 0.36* 97 ± 19 1.0 ± 0.8*

0.24 ± 0.1 0.9 ± 0.1 98 ± 19 0.5 ± 0.2

2

Sn Pb Cu Zn

– – Matrix Present

– – Matrix Present

3

Sn Pb Cu Zn

Present Present 100 ± 19 0.3 ± 0.1*

Present Present 100 ± 19 0.33 ± 0.07

wt% LA-ICP-MS

0.02 ± 0.01 0.032 ± 0.001 99.9 ± 0.1 0.021 ± 0.002

22

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24

were anyway compatible with the composition of copper-based alloys employed in jewellery, in the present case Pb percentages as high as 30% led us to conclude that the examined portion of sample could be severely altered with respect to the original alloy. This hypothesis was confirmed by the SEM image of a section of the sample, reported in Fig. 4: this shows that along the section circumference, the thickness of the patina can reach up to 1 mm, and that many and sometimes large Pb agglomerates (visible as white globules in the image) are present in the patina. This can explain the uncommonly high Pb percentage found in spots 2 and 3, as well as the high Sn content, again due to Sn-enriched patina, as confirmed also by EDS analyses of patina. On the other hand, spot 4, which corresponds to the section, thus to the sample bulk, shows fair agreement with LA-ICP-MS, confirming the hypothesis made about the previous spots. For what concerns the correlation between the

two LIBS methods, it can be considered satisfactory, though the CFLIBS series slightly underestimates Sn and overestimated Pb in spot 4. 3.4.4. Sample 4230 (dating unknown) Sample 4230 represents an exception in the set, as it appears to be mainly made of pure Cu. Two of the three analyzed spots contain very low amounts of Sn, Pb and Zn (in some cases below the LOD, and when quantified, accompanied by high experimental uncertainty, thus making the measurement only semi-quantitative). In the spot where also LA-ICP-MS analysis has been performed, signals from other elements were completely absent in the LIBS spectra, with the exception of Zn (below the limit of detection) and other non-quantified impurities (Si, Ag), thus showing a substantial agreement between the two techniques (see Table 7).

Table 8 Comparison between results of calibration line LIBS, CF inverse method LIBS and LA-ICP-MS for samples 4237, 4207 and 4381. Data indicated with an asterisk are affected by high experimental uncertainty and should thus be considered only semi-quantitative. Cu and Ag weight percentages reported in the “Calibration line” column were determined with the mixed approach described by Eq. (5), and are reported in italic to distinguish them from the weight percentages determined with proper calibration lines. Sample

Spot

Element

4237

1

Sn Pb Cu Zn Ag Sn Pb Cu Zn Ag Sn Pb Cu Zn Ag Sn Pb Cu Zn Ag

0.5 1.2 65 2.2 26 0.3 0.8 74 5.4 14 0.6 1.2 76 3.1 18 1.0 0.6 68 6.7 24

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.1 0.1 13 0.3 5 0.1* 0.1 15 0.3 3 0.1 0.1 15 0.3 4 0.1 0.1 13 0.2 5

1.1 1.6 65 2.8 26 0.7 0.9 76 6 15 1.2 1.1 75 3.9 18 1.8 1.0 66 8 23

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.2 0.3 13 0.6 5 0.1 0.2 15 1 3 0.2 0.2 15 0.8 4 0.3 0.2 13 1 5

Sn Pb Cu Zn Ag Sn Pb Cu Zn Ag Sn Pb Cu Zn Ag

1.28 0.53 30 2.9 64 0.99 1.4 57 2.4 37 2.6 0.9 67 9 21

± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.08 0.04 6 0.3 13 0.09 0.1 11 0.3 7 0.2 0.07 13 1 4

0.9 0.35 31 1.9 66 1.0 1.05 58 1.9 38 2.1 1.6 66 9 21

± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.2 0.07 6 0.4 13 0.2 0.21 12 0.4 8 0.4 0.3 13 2 4

Sn Pb Cu Zn Sn Pb Cu Zn Sn Pb Cu Zn Sn Pb Cu Zn

0.84 3.3 77 17.6 1.15 1.1 78 19.3 0.89 5.65 72 19.6

± ± ± ± ± ± ± ± ± ± ± ±

0.09 0.3 15 0.3 0.08 0.3 16 0.3 0.09 0.55 14 0.3

1.5 1.0 78 19 1.4 0.9 74 24 1.3 3.3 69 26

± ± ± ± ± ± ± ± ± ± ± ±

0.3 0.2 16 4 0.3 0.2 15 5 0.2 0.7 14 5

2

3

4

4207

1

2

3

4381

1

2

3

wt% calibration line

wt% CF

wt% LA-ICP-MS

0.253 ± 0.40 ± 64.3 ± 20.3 ± 14.5

0.004 0.01 0.9 0.4

0.88 0.44 65.6 12.1 20.5

± ± ± ± ±

0.01 0.01 0.4 0.3 0.1

0.60 4.25 70.3 24.3

± ± ± ±

0.2 0.07 1.0 0.1

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24

3.4.5. Samples 4237 (VI–VII cent. AD), 4207 (VI–VII cent. AD), 4381 (VI–VII cent. AD) The last group of brooches is the most valuable of the collection and comprises samples of a certain interest for several reasons. From the archaeological point of view they are particularly important, due to the good state of conservation and peculiar shape. This set indeed consists of three zoomorphic brooches, in the shape of birds, much more recent than the bronze ones. Two of them are made of an alloy containing high amount of Ag and Zn, as determined by LA-ICP-MS, while the third is made of a Cu alloy. From the point of view of testing the CF inverse method, therefore, these samples represent a peculiarly interesting case of study, in that their matrix differs significantly from those of the standards employed for the temperature determination. The agreement between the two LIBS methods is rather good (see Table 8 and Fig. S9(a)–(c), Supplementary material). This appears to indicate that compositional differences between the standard used for temperature determination and unknown samples can be even more pronounced than initially supposed, which in turn suggests that further extension of the applicability of the inverse method could be possible. The LIBS data are also in good agreement with LA-ICP-MS results. The main discrepancy with the latter arises in the determination of Zn and Ag in spot 4 of sample 4237, whose percentages appear almost inverted in the LIBS measurements with respect to the LA-ICP-MS ones (on the other hand, for the other two samples this difference is virtually absent). A further interesting aspect of these samples, in particular 4237 and 4207, is that they consist of separate pieces (i.e. pin and body of the brooch), which raises the question whether or not the same alloys were used for the different parts of the samples. This question could not be answered with the sampling protocol employed in [35] by LA-ICP-MS, because only the pin could be sectioned and analyzed, due to the preciousness of these brooches and in order to preserve their integrity. Instead, the LIBS measurements were performed on the various portions of the sample in a virtually non-destructive mode, and gave good agreement with LA-ICP-MS when the same portion was analyzed, enabling at the same time to detect different compositions in the other portions. In particular, in sample 4207 the LIBS analysis showed that a less precious alloy, with a significantly lower Ag content had been actually used for producing the pin, while in the other portions the employed alloy appeared more as a Ag-based alloy, rather than a Cu-based one. Instead, sample 4237 appeared more homogeneous, consistently with its Ag content being similar to that of the pin of 4207, thus overall 4237 may be supposed less precious than 4207. A last remark about this set is about sample 4381, for which the chemical analysis provided an important support to the archaeological one. In spite of its apparent similarity with the others, indeed, this brooch revealed a rather different elemental composition, with virtually no Ag and high amount of Zn. This suggested that it could be an ancient imitation of the more precious ones, made of a Cu-Zn-Pb alloy (lead brass).

4. Conclusions In this work we applied the CF inverse method, that we have developed previously and is based on the use of a single reference standard sample to determine the plasma temperature, to the LIBS analysis of a set of fibulae from the archaeological site of Egnatia (Apulia, Southern Italy). The aim of the present work was dual: first of all, testing the analytical performance of the inverse method for the systematic analysis of archaeological objects, by comparing its results with those of classical LIBS; second, comparing the LIBS data with those of independent measurements performed with LAICP-MS. The agreement between the two series of LIBS data was found to be satisfactory for all the samples, which confirmed the

23

validity of the assumptions on which the inverse method relies. On the other hand, for what concerns the comparison with LA-ICPMS data, the difference between LIBS results appeared to concern mainly Sn, which in the bronze samples was overestimated by LIBS. This circumstance was explained by taking into account the thickness and kind of the corrosion patina of the bronze objects, which for some of them was up to one millimetre thick. This interpretation was confirmed by the acquisition, for two fibulae, of spectra of the actual bulk of the sample (i.e. from the section analyzed with LA-ICP-MS) which gave a very good agreement with the LA-ICPMS data. Thanks to the almost complete non-destructiveness of the employed LIBS protocol, moreover, some interesting aspects were revealed, in particular for the three zoomorphic bird-shaped brooches. The elemental analysis previously performed with LAICP-MS had already revealed that one of the three samples was a less precious imitation of the others, a result that was confirmed in this study with LIBS. In addition, thanks to the almost complete non-destructiveness of the employed LIBS protocol, it was possible to observe compositional inhomogeneity in some of the sample portions, most notably between body and pin. Finally, a very good agreement was obtained, not only between LIBS and LA-ICP-MS, but also between the two LIBS methods for these samples, which had a rather different matrix from that of the standard used for temperature determination. This was a further confirmation of the feasibility of the CF inverse method as an effective tool for the elemental analysis of complex samples, even in the absence of suitable matrix-matched samples of known composition. Acknowledgments This work has been partially supported by the Project P.O. Puglia FESR 2007–2013, Line 1.2, Action 1.2.4, “Restaureo: Diagnostiche laser per la caratterizzazione e recupero delle grandi opere in Puglia”, COD 3Z3VZ46, funded by Regione Puglia (Italy). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.aca.2014.01.020. References [1] F.J. Fortes, J. Moros, P. Lucena, L.M. Cabalín, J.J. Laserna, Laser-induced breakdown spectroscopy, Analytical Chemistry 85 (2013) 640–669. [2] A. De Giacomo, M. Dell’Aglio, R. Gaudiuso, S. Amoruso, O. De Pascale, Effects of the background environment on formation, evolution and emission spectra of laser-induced plasmas, Spectrochimica Acta Part B 78 (2012) 1–19. [3] R. Noll, Laser-Induced Breakdown Spectroscopy, Springer-Verlag, Berlin Heidelberg, 2012. [4] R. Gaudiuso, M. Dell’Aglio, O. De Pascale, G.S. Senesi, A. De Giacomo, Laser induced plasma spectroscopy for elemental analysis in environmental, cultural heritage and space applications: a review of methods and results, Sensors 10 (2010) 7434–7468. [5] J.P. Singh, S.N Thakur (Eds.), Laser-Induced Breakdown Spectroscopy, Elsevier, 2007. [6] A.M. Miziolek, V. Palleschi, I. Schechter (Eds.), Laser-Induced Breakdown Spectroscopy (LIBS): Fundamentals and Applications, Cambridge University Press, 2006, pp. 516–538. [7] A. Giakoumaki, K. Melessanaki, D. Anglos, Laser-induced breakdown spectroscopy (LIBS) in archaeological science—applications and prospects, Analytical and Bioanalytical Chemistry 387 (2007) 749–760. [8] C. Vázquez-Calvo, A. Giakoumaki, D. Anglos, M. Álvarez de Buergo, R. Fort, Classification of patinas found on surfaces of historical buildings by means of laser-induced breakdown spectroscopy, in: Lasers in the Conservation of Artworks, Springer, Proc. Phys. 116 (2007) 415–420. [9] I. Gaona, P. Lucena, J. Moros, F.J. Fortes, J.J. Laserna, Evaluating the use of standoff LIBS in architectural heritage: surveying the Cathedral of Málaga, Journal of Analytical Atomic Spectrometry 28 (2013) 810–820. [10] L. Burgio, K. Melessanaki, M. Doulgeridis, R.J.H. Clark, D. Anglos, Pigment identification in paintings employing laser induced breakdown spectroscopy and Raman microscopy, Spectrochimica Acta Part B 56 (2001) 905–913.

24

R. Gaudiuso et al. / Analytica Chimica Acta 813 (2014) 15–24

[11] R. Bruder, D. L’Hermite, A. Semerok, L. Salmon, V. Detalle, Near-crater discoloration of white lead in wall paintings during laser induced breakdown spectroscopy analysis, Spectrochimica Acta Part B 62 (2007) 1590–1596. [12] R. Bruder, V. Detalle, C. Coupry, An example of the complementarity of laserinduced breakdown spectroscopy and Raman microscopy for wall painting pigments analysis, Journal of Raman Specroscopy 38 (2007) 909–915. [13] A. De Giacomo, M. Dell’Aglio, O. De Pascale, R. Gaudiuso, A. Santagata, R. Teghil, Laser induced breakdown spectroscopy methodology for the analysis of copper-based-alloys used in ancient artworks, Spectrochimica Acta Part B 63 (2008) 585–590. [14] F.J. Fortes, M. Cortés, M.D. Simón, L.M. Cabalín, J.J. Laserna, Chronocultural sorting of archaeological bronze objects using laser-induced breakdown spectrometry, Analytica Chimica Acta 554 (2005) 136–143. [15] V. Lazic, F. Colao, R. Fantoni, V. Spizzicchino, Recognition of archaeological materials underwater by laser induced breakdown spectroscopy, Spectrochimica Acta Part B 60 (2005) 1015–1024. [16] M. Simileanu, R. Radvan, Remote method and set-up for the characterization of the submerged archaeological remains, Journal of Optoelectronics and Advanced Materials 13 (2011) 528–531. [17] S. Guirado, F.J. Fortes, V. Lazic, J.J. Laserna, Chemical analysis of archeological materials in submarine environments using laser-induced breakdown spectroscopy. On-site trials in the Mediterranean Sea, Spectrochimica Acta Part B 74–75 (2012) 137–143. [18] A. De Giacomo, M. Dell’Aglio, A. Casavola, G. Colonna, O. De Pascale, M. Capitelli, Elemental chemical analysis of submerged targets by double-pulse laserinduced breakdown spectroscopy, Analytical and Bioanalytical Chemistry 385 (2006) 303–311. [19] F. Colao, R. Fantoni, V. Lazic, V. Spizzichino, Laser-induced breakdown spectroscopy for semi-quantitative and quantitative analyses of artworks—application on multi-layered ceramics and copper based alloys, Spectrochimica Acta Part B 57 (2002) 1219–1234. [20] I. Osticioli, J. Agresti, C. Fornacelli, I. Turbanti Memmi, S. Siano, Potential role of LIPS elemental depth profiling in authentication studies of unglazed earthenware artifacts, Journal of Analytical Atomic Spectrometry 27 (2012) 827–833. ˜ [21] A. Ramil, A.J. López, A. Yánez, Application of artificial neural networks for the rapid classification of archaeological ceramics by means of laser induced breakdown spectroscopy (LIBS), Applied Physics A: Materials Science and Processing 92 (2008) 197–202. [22] I. Osticioli, N.F.C. Mendes, S. Porcinai, A. Cagnini, E. Castellucci, Spectroscopic analysis of works of art using a single LIBS and pulsed Raman setup, Analytical and Bioanalytical Chemistry 394 (2009) 1033–1041. [23] S. Klein, J. Hildenhagen, K. Dickmann, T. Stratoudaki, V. Zafiropulos, LIBSspectroscopy for monitoring and control of the laser cleaning process of stone and medieval glass, Journal of Cultural Heritage 1 (2000) S287–S292. [24] M. Oujja, E. Rebollar, M. Castillejo, C. Domingo, C. Cirujano, F. Guerra-Librero, Laser cleaning of terracotta decorations of the portal of Palos of the Cathedral of Seville, Journal of Cultural Heritage 6 (2005) 321–327. [25] P. Maravelaki-Kalaitzaki, V. Zafiropulos, P. Pouli, D. Anglos, C. Balas, R. Salimbeni, S. Siano, R. Pini, Short free running Nd:YAG laser to clean different encrustations on Pentelic marble: procedure and evaluation of the effects, Journal of Cultural Heritage 4 (2003) 77s–82s. [26] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti, E. Tognoni, New procedure for quantitative elemental analysis by laser induced plasma spectroscopy, Applied Spectroscopy 53 (1999) 960–964. [27] E. Tognoni, G. Cristoforetti, S. Legnaioli, V. Palleschi, Calibration-free laserinduced breakdown spectroscopy: state of the art, Spectrochimica Acta Part B 65 (2010) 1–14. [28] A. De Giacomo, M. Dell’Aglio, O. De Pascale, S. Longo, M. Capitelli, Laser induced breakdown spectroscopy on meteorites, Spectrochimica Acta Part B 62 (2007) 1606–1611. [29] M. Dell’Aglio, A. De Giacomo, R. Gaudiuso, O. De Pascale, G.S. Senesi, S. Longo, LIBS applications to meteorites: chemical analysis and composition profiles, Geochimica et Cosmochimica Acta 74 (2010) 7329–7339. [30] A. De Giacomo, M. Dell’Aglio, R. Gaudiuso, A. Santagata, G.S. Senesi, M. Rossi, M.R. Ghiara, F. Capitelli, O. De Pascale, A laser induced breakdown spectroscopy application based on local thermodynamic equilibrium assumption for the elemental analysis of alexandrite gemstone and copper-based alloys, Chemical Physics 398 (1) (2012) 233–238. [31] J.D. Pedarnig, P. Kolmhofer, N. Huber, B. Praher, J. Heitz, R. Rössler, Element analysis of complex materials by calibration-free laser-induced breakdown spectroscopy, Applied Physics A: Materials Science and Processing 112 (1) (2013) 105–111. [32] B. Praher, R. Rössler, E. Arenholz, J. Heitz, J.D. Pedarnig, Quantitative determination of element concentrations in industrial oxide materials by laser-induced breakdown spectroscopy, Analytical and Bioanalytical Chemistry 400 (10) (2011) 3367–3375. [33] M. Corsi, G. Cristoforetti, M. Giuffrida, M. Hidalgo, S. Legnaioli, L. Masotti, V. Palleschi, A. Salvetti, E. Tognoni, C. Vallebona, A. Zanini, Archaeometric analysis of ancient copper artefacts by laser-induced breakdown spectroscopy technique, Microchimica Acta 152 (1–2) (2005) 105–111.

[34] R. Gaudiuso, M. Dell’Aglio, O. De Pascale, A. Santagata, A. De Giacomo, Laserinduced plasma analysis of copper alloys based on local thermodynamic equilibrium: an alternative approach to plasma temperature determination and archeometric applications, Spectrochimica Acta Part B: Atomic Spectroscopy 74–75 (2012) 38–45. [35] L.C. Giannossa, S. Loperfido, M. Caggese, G.E. De Benedetto, R. Laviano, L. Sabbatini, A. Mangone, A systematic characterization of fibulae from Italy: from chemical composition to microstructure and corrosion processes, New Journal of Chemistry 37 (4) (2013) 1238–1251. [36] R. Cassano, C.S. Fioriello, L. Tedeschi, A. Pedone, V. Di Grazia, in: M. Pani (Ed.), Epigrafia e territorio. Politica e società. Temi di antichità romane, vol. VII, Edipuglia, Bari, 2004, pp. 7–98. [37] M.R. Cassano, et al., Ricerche archeologiche nell’area del foro di Egnazia. Scavi 2001-2003: relazione preliminare, in: M. Pani (Ed.), Epigrafia e territorio, Politica e società, temi di antichità romane, vol. VII, Bari, 2004, p. 7. [38] M.R. Cassano, Egnazia tardoantica: nuove indagini e prospettive di ricerca, in: G. Volpe, M. Turchiano (Eds.), Paesaggi e insediamenti urbani tardoantichi in Italia meridionale fra Tardoantico e Altomedioevo, Proceedings of Secondo seminario sul Tardoantico e l’Altomedioevo in Italia meridionale, Lucera, Italy, 24–25 maggio 2005, 2005. [39] G. Cristoforetti, A. De Giacomo, M. Dell’Aglio, S. Legnaioli, E. Tognoni, V. Palleschi, N. Omenetto, Local thermodynamic equilibrium in laser-induced breakdown spectroscopy: beyond the McWhirter criterion, Spectrochimica Acta Part B: Atomic Spectroscopy 65 (1) (2010) 86–95. [40] M. Capitelli, F. Capitelli, A. Eletskii, Non-equilibrium and equilibrium problems in laser-induced plasmas, Spectrochimica Acta Part B 55 (2000) 559–657. ´ Plasma broadening and shifting of non-hydrogenic spectral lines: [41] N. Konjevic, present status and applications, Physics Reports 316 (1999) 339–401. [42] A. Alonso-Medina, Experimental determination of the Stark widths of Pb I spectral lines in a laser-induced plasma, Spectrochimica Acta Part B 63 (2008) 598–602. [43] S¸. Yalc¸in, D.R. Crosley, G.P. Smith, G.W. Faris, Influence of ambient conditions on the laser air spark, Applied Physics B 68 (1999) 121–130. [44] L. Fornarini, R. Fantoni, F. Colao, A. Santagata, R. Teghil, A. Elhassan, M.A. Harith, Theoretical modeling of laser ablation of quaternary bronze alloys: case studies comparing femtosecond and nanosecond LIBS experimental data, Journal of Physical Chemistry A 113 (52) (2009) 14364–14374. [45] H.-Y. Moon, K.K. Herrera, N. Omenetto, B.W. Smith, J.D. Winefordner, On the usefulness of a duplicating mirror to evaluate self-absorption effects in laser induced breakdown spectroscopy, Spectrochimica Acta Part B 64 (2009) 702–713. [46] Kurucz Atomic spectral lines database, http://www.cfa.harvard.edu/ amp/ampdata/kurucz23/sekur.html (accessed August 2013). [47] NIST atomic spectral database, http://www.nist.gov/pml/data/asd.cfm (accessed August 2013). [48] A. Mangone, G.E. De Benedetto, D. Fico, L.C. Giannossa, R. Laviano, L. Sabbatini, I. van der Werf, A. Traini, Multianalytical study of archaeological faience from Vesuvian area as a valid tool to investigate provenance and technological features, New Journal of Chemistry 35 (2011) 2860–2868. ˜ [49] J. Cunat, F.J. Fortes, J.J. Laserna, Real time and in situ determination of lead in road sediments using a man-portable laser-induced breakdown spectroscopy analyzer, Analytica Chimica Acta 633 (2009) 38–42. [50] O.V. Borisov, X.L. Mao, A. Fernandez, M. Caetano, R.E. Russo, Inductively coupled plasma mass spectrometric study of non-linear calibration during laser ablation of binary Cu–Zn alloys, Spectrochimica Acta Part B 54 (1999) 1351–1365. [51] L. Robbiola, J.-M. Blengino, C. Fiaud, Morphology and mechanisms of formation of natural patinas on archaeological Cu–Sn alloys, Corrosion Science 40 (12) (1998) 2080–2111. [52] I. Costantinides, A. Adriaens, F. Adams, Surface characterization of artificial corrosion layers on copper alloy reference materials, Applied Surface Science 189 (2002) 90–101. [53] D.A. Scott, Copper and Bronze in Art. Corrosion, Colorants Conservation, Getty Publication, Los Angeles, 2002. [54] F.J. Sarabia-Herrero, J. Martín-Gil, F.J. Martín-Gil, Metallography of ancient bronzes: study of pre-roman metal technology in the Iberian Peninsula, Materials Characterization 36 (4–5) (1996) 335–347. [55] D.A. Scott, Metallography and Microstructure of Ancient and Historic Metals, Institute of Archaeology, London, 1987. [56] G.M. Ingo, T. De Caro, C. Riccucci, E. Angelini, S. Grassini, S. Balbi, P. Bernardini, D. Salvi, L. Bousselmi, A. C¸ilingiro˘glu, M. Gener, V.K. Gouda, O. Al Jarrah, S. Khosroff, Z. Mahdjoub, Z. Al Saad, W. El-Saddik, P. Vassiliou, Large scale investigation of chemical composition, structure and corrosion mechanism of bronze archeological artefacts from Mediterranean Basin, Applied Physics A 83 (2006) 513–520. [57] L. Robbiola, R. Portier, A global approach to the authentication of ancient bronzes based on the characterization of the alloy–patina–environment system, Journal of Cultural Heritage 7 (1) (2006) 1–12.