Characterization of the elemental composition of polish banknotes by X-ray fluorescence and laser-induced breakdown spectroscopy

In this study, two very promising techniques, micro X-ray fluorescence (μXRF) and laser-induced breakdown spectroscopy (LIBS) were applied to the examination of Polish banknotes – Polish zloty (zl). Several areas on each banknote were selected and analysed. Different elemental compositions were identified after comparing the spectra recorded from various measurement locations. It was possible to identify characteristic atomic emissions from one or several elements such as Ca, Ti, Fe, Ba, Co, Cr, Cu, Mg, Mn, Ni, V, and Zr, depending on the banknote denomination, issue date, and evaluated spot. Potentially good discriminators with unique elemental composition were identified: black serial number (C) and microlettering (A). A comparison of brand-new banknotes with used banknotes (which have been in circulation) was also performed. The middle horizontal section of the banknotes shows higher exposure to contamination and consequently constitutes the most difficult part to analyse. Counterfeit banknotes were also analysed and were clearly distinguished from authentic notes in all cases. It was demonstrated that a comparison of the elemental composition is a useful way to detect counterfeit banknotes (10, 20, 50 and 100 zl) in ‘real-world’ cases. This study shows the potential of LIBS and μXRF as effective and practical techniques to analyse Polish banknotes. Their many advantages provide a good alternative to the analytical methods routinely used for the examination of these objects.

demonstrated that XRF examination of major and trace elemental composition supported by multivariate statistical methods is a useful way to compare and classify counterfeit banknote papers. The potential of portable X-ray fluorescence (pXRF) spectrometry [12] and portable near-infrared spectroscopy (pNIR) [13] was shown in the examination of differently coloured regions of dollar, euro, and real banknotes. These methods were supported by chemometric methods, including principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA). Nondestructive procedures based on the analysis of several areas of euro banknotes using X-ray diffraction (XRD) [14] and micro-ATR-infrared spectroscopy [15] have been proposed to characterize and distinguish between original and counterfeit bills. A forgery detection method that uses a compact Fourier transform infrared spectrometer (FT-IR) has been developed. This method allows not only experienced scientists in laboratories, but also bank employees, police personnel, and shop workers to identify forged banknotes [16]. Mössbauer spectroscopy has been applied to the analysis of pigments in authentic and counterfeit banknotes [17]. The former contain relatively larger amounts of iron, so the detection of counterfeits is possible. The results have allowed the rapid and non-destructive identification of three types of fake banknotes: 100 USD old issue (OI) and new issue (NI) banknotes, and 50 USD NI banknotes. Desorption electrospray ionization (DESI) and easy ambient sonic-spray ionization (EASI) have been employed in the direct analysis by ambient mass spectrometry of Brazilian real, US dollar, and euro banknotes [18]. An instantaneous, reproducible, and non-destructive method for the chemical analysis of banknotes (including so-called 'real-world' counterfeits) has been developed. Electrospray laser desorption ionization mass spectrometry (LDMS), resulting in minimal damage to the banknote, was used to distinguish between authentic and counterfeit US dollars, and authentic and counterfeit new Taiwan dollars [19]. The authentic dollars had a different surface chemical composition, which was used as a marker to distinguish real banknotes from counterfeited ones. In another paper [20], LDMS was presented as an effective analytical tool for the detection and identification of security ink (e.g. Basic Red 1:1 or Basic Violet 11:1) in the presence of other colourants applied to the surface of paper (banknotes) and textiles. Although less common, there are also articles in which the destructive examination of banknotes has been reported. For instance, a rapid mass spectrometry procedure based on direct mixture analysis with conventional electron ionization was used to match pigments present on counterfeit $100 US banknotes with those contained in two suspicious inks [21]. Gas chromatography coupled to mass spectrometry (GC-MS) was used to characterize the volatile components associated with US and Canadian currencies [22].
The present study evaluates the effectiveness and usefulness of laser-induced breakdown spectroscopy (LIBS) since, to the authors' best knowledge, this technique has not been applied yet to the analysis of banknotes. A correlation of LIBS and micro X-ray fluorescence (μXRF) data -leading techniques in forensic trace elemental analysis -was also performed. The combination of these two techniques is very useful in the case of multilayer materials. The penetration depth of an X-ray beam in a sample is dependent on the elemental composition, density and absorption characteristics of that sample. Therefore, X-ray methods are often referred to as "bulk" analysis and the analyst should take precautions to minimize contamination from underneath the sample (e.g. stage or mounting media) during analysis. LIBS, on the other hand, is a surface technique. In this case, the information comes only from the ablated mass. LIBS has a small sample requirement and offers rapid, sensitive, and precise semi-quantitative elemental analysis. Compared to μXRF, LIBS is less complex to operate and has higher sample throughput (a large number of data registered in a short time). One of the main drawbacks of LIBS is its destructiveness -a small crater is visible after ablation: diameter 250-1000 μm, depth < 100 μm [23][24][25] depending on the physical and chemical properties of the analysed object. The second disadvantage of LIBS is the difficulty of performing an overall reliable data analysis -sometimes over 500 spectral lines need to be interpreted. During the last decade, the LIBS technique has undergone significant development [26][27][28][29][30]. Many different identification or classification methods [31][32][33][34], as well as quantification ones, have been applied [35][36][37][38][39] to support the interpretation of LIBS results. LIBS is more and more often used in various forensic examinations [40][41][42][43][44][45][46][47][48]. Previous literature [23][24][25][49][50][51][52][53] shows the potential of LIBS for the examination of colouring matter on paper using this technique. It was selected for the present study in order to choose the most discriminating areas on Polish banknotes based on elemental analysis. To the best of the authors' knowledge, this is the first publication that correlates results from LIBS and μXRF for forensic banknote examinations.

μXRF
The micro-XRF analyses of the banknotes were carried out using a Bruker (Karlsruhe, Germany) Artax 800 spectrometer. This instrument is equipped with an Rh target polycapillary lens X-ray tube. The instrument has a Si drift X-ray detector, which has an active area of 10 mm 2 . All spectra were collected at 50 kV and 600 mA, while the live time count used was 45 s. The measured gross count rate was around 13,000 cps and the average dead time was 3%. The analytical spot size of the instrument is approximately 0.60 mm in diameter. The beam was focused on the analysis spot with the help of a laser and a camera, which are attached to the spectrometer. X rays with high energy can completely or partially penetrate through the sample; the penetration depth is dependent on the applied energy. XRF analysis of the paper is difficult due to its minimal thickness, low density, and the low concentration of the elements of interest. The mass attenuation coefficient of paper can be approximated by assuming that the paper is 99% cellulose by mass, and that the inorganic components of the paper and inks account for about 1% of the banknote's composition. Cellulose is an organic polymer with a density of 1.5 g•cm −3 [54] and a mass attenuation coefficient of 0.211 cm −3 •g at about 50 kV [55]. The depth of penetration can be estimated using the Beer-Lambert law (Eq. (1)).
where I and I 0 are the transmitted and incident photon flux, respectively; μ = ρ × σ is the absorption coefficient in cm −1 (ρ and σ are the density and the mass absorption coefficient, respectively); and t is the thickness of the sample. On the basis of this approach, the penetration depth was estimated at 14 cm. Drake [56] and Lera et al. [57] have reported X-ray penetrations of 12.5 cm (cellulose) and 4.6 cm (paper stamps), respectively, when a tube voltage of 40 kV is employed. Penetration depth, however, can be smaller in the actual samples as not only cellulose is present on the banknotes; the chemical and physical composition of the ink can also contribute. It was observed that the radiation is capable of penetrating through the entire banknote since its thickness remains below 0.15 mm. To avoid a situation where elements in the surface of the plate below the sample are fluoresced and detected, a special sample plate was used. An empty tube of about 7 cm in height was applied so even if the radiation passed through the sample, it only encountered air space. A more accurate estimation of the penetration depth of X-rays can be carried out by preparing a series of standards containing inks with a similar composition to those found in the banknotes and using an equivalent degree of penetration of the ink into paper substrates of various thicknesses. In this way, individual lines for each element present in the inks applied on paper can be described in terms of the analytical penetration depth of the technique. However, this methodology is beyond the scope of this work and is being considered as a potential topic for future research. Acquisition and evaluation of XRF spectra were carried out using Spectra 5.3 (Bruker AXS Microanalysis, Berlin, Germany). Using this software, the net peak areas were obtained M. Król, et al. Spectrochimica Acta Part B 169 (2020) 105898 after integral intensity calculation of the different spectral lines by Bayes deconvolution -the process of removing some signal distortions using the theoretically calculated profile as a reference.

LIBS
The analysis of Polish banknotes was performed using laser-induced breakdown spectroscopy system LIBS-6 (Applied Photonics, United Kingdom) with an integrated Q-switched Quantel Ultra Nd-YAG laser operating at λ = 1064 nm (Quantel, France) with a maximum energy of 150 mJ per laser pulse in 6 ns (max. Irradiance 25 GW•cm −2 ). The diameter of the ablation spot caused by the laser was dependent on the analysed material and ranged from 0.6-1 mm. All LIBS analyses were performed under atmospheric pressure in the air. The emitted light from the LIBS plasma was collected by an Avaspec-2048-2-USB2 fibre optic Czerny-Turner spectrometer (6-channel) with a CCD detector (Avantes, The Netherlands). The emission spectrum was collected in the UV-Vis range (185-904 nm) with a spectral resolution of about 0.1 nm. The analysed object was placed on a movable table at the focal point of the focusing lens at a distance determined by the nozzle about 70 mm from the optical head. The system has a camera, giving the possibility to observe the analysed object and to aim the laser beam at the appropriate point. The LIBS-6 system was controlled by LIBSoft V6.0.1 software (Applied Photonics, United Kingdom). LIBS spectra were evaluated, and their component spectral lines were assigned using Plasus SpecLine 2.13 software (Plasus, Germany) and an NIST database (in wavelength ranges of ± 0.1 nm). During the identification process, only emission lines with intensity ≥ threefold standard deviation (SD) of signals at the noise level were taken into account.

Samples
The currency of the Polish Republic -Polish zloty (zl) -with denominations of 10, 20, 50 and 100 zl was examined. Banknotes were obtained from the National Bank of Poland (NBP) from two issues: before (old issue -OI) and after (new issue -NI) April 4th, 2014 (both brand new and used banknotes were obtained). In the analysis with μXRF, 24 samples of only brand-new polish banknotes: 12 OI (3 items per denomination), and 12 NI banknotes (3 items per denomination) were examined. In the case of LIBS examination, 52 samples were analysed: 20 brand new OI banknotes (5 items per denomination) and 20 used OI banknotes (5 items per denomination) as well as 12 brand new NI banknotes (3 items per denomination). Unfortunately, for security reasons, there is no available information on the elemental composition of printing inks used in banknotes. Prior to analysis, the banknotes were photographed, appropriately segregated and sealed in plastic bags. All banknotes were directly analysed using μXRF spectrometry and LIBS without sample pretreatment. Several areas of the banknotes on the obverse: A -F (based on denomination) and P (paper area) were analysed (see Table 1 and Fig. 1). After a careful inspection of area P on an OI banknote under the microscope (Motic, Polska), it was observed that this area had been printed on using a striped pattern ( Fig. S1 (Appendix)) Thus, unfortunately, there was no place on OI banknotes where a spectrum of pure paper could be registered. Four 'real-world' counterfeit banknotes (10, 20, 50 and 100 zl from the OI) provided by the Cash and Issue Department of the National Bank of Poland were also investigated (using μXRF and LIBS) in order to compare the results obtained with those corresponding to the genuine notes.

μXRF analysis
In order to verify the repeatability, the following examinations were performed: 1) area P of a 20 zl banknote was examined 5 times by single spot analysis in one place to estimate the influence of instrumental uncertainties (e.g. the laser shot-to-shot fluctuation); and 2) three areas (P, C, and D) of a 10 zl banknote were analysed 5 times at closely placed points (within approx. 0.25 cm 2 ). These three areas were selected since they have different physical features (see Table 1) and, due to that, a different chemical composition. The evaluation included a relative comparison of the raw spectra and obtained net peak areas. The results of these experiments are presented in Table 2.
In general, both matrix (e.g. absorption, secondary fluorescence) and instrumental parameter effects (e.g. background radiation, the inhomogeneous intensity distribution of the beam, sensitivity) can influence the precision of the XRF method. In our study, the precision was relatively good, the RSD values of net peak areas were lower than 10%. The lowest value was obtained for the main element that shows welldefined peaks (i.e., titanium). On the other hand, considering the repeatability of the method with 5 measurements of the same area at closely placed points, the RSD of the net peak areas oscillated around 20%. Such results showed that the peak areas strongly varied from one measurement point to another. The heterogeneity of a banknote's surface causes dissimilarities in elemental composition even for nearby measuring points. A solution to the problem of inhomogeneity could be to perform XRF elemental mapping of the entire object. X rays can go through the sample. The depth of penetration of X-rays is influenced by several factors including the density of the analysed material, the fluorescence energy of the chemical elements in the object, and the energy of the X-ray beam. In the case of banknotes -where samples are thin enough for the radiation to entirely penetrate through -elements from the other side of the banknote may be visible in the spectrum. This could be a potential source of error. Thus, it is recommended to take special precautions during the analysis of such thin objects.
Visual comparison of the 24 XRF spectra provided evidence that the elemental profiles of the examined areas were composed of: Ca, Ti, Fe, Mn, Co, Cu, Cr, and Zr in the case of OIs, whereas in current banknotes (NI) two more elements: Ni and Ba (in a very characteristic configuration) were found (see Fig. S2 and S3 (Appendix)). This confirms that the NI of banknotes is produced in a more modern way using ink with a different elemental composition. Evidently, the combination of elements detected on the various measuring areas was different (Table 3).
Furthermore, in lower denomination samples, fewer elements were detected than in 50 and 100 zl banknotes. According to these results, it could be ascertained that OI 10 and 20 zl Polish banknotes have the least varied elemental content. Probably due to their relatively lower value, they seem to be less protected in terms of security features [58]. Independently of the banknote denomination, the ratios of the main elements (net areas) showed interesting relations. In the case of OIs, three examined areas showed a unique elemental composition: 1) area A with Ti and Ca contents on a high level: After evaluating the spectra of NIs, two measuring areas attracted attention: 1) area A with high levels of Ti and Ca: [Ti] ≅ [Ca], and an additional element namely, Ba, in 50 and 100 zl banknotes, 2) area E (only on 50 and 100 zl notes) with content of Ti, Cr and Ni: Thus, these results along with a visual comparison indicated that these areas were the best options for verifying the authenticity of banknotes.

Statistical analysis of XRF results
The analysis of μXRF data was carried out on three datasets, each described by three independent variables. Banknotes of all analysed denominations all together; 10 and 20 zl; and 50 and 100 zl were treated as separate datasets (γ, α, and β, respectively). Dependent variables, namely net areas of Ca, Ti and Fe normalized to Rh net area, were chosen because of their presence in all analysed samples. Three independent variables were considered: measurement area, issue, and denomination. Since the assumptions for the ANOVA were not met, non-parametric Kruskal-Wallis one-way analysis of variance by ranks was employed. This test is aimed at comparing samples from k groups of independent observations (k > 2) with the null hypothesis stating that the groups are sampled from populations with identical distributions with the alternative that at least two are sampled from different populations. It is based on the H statistic, calculated using sums for ranked data, and approximated by a chi-square distribution with k-1 degrees of freedom (χ 2 (k-1)) at α = 0.05. The statistical decision is to reject the null hypothesis when H > χ 2 (k-1), suggesting that the data provide sufficient evidence to conclude that there are significant differences among the groups. It was established, that statistically significant (p < .01) differences in all datasets were recognized when analysing Ca and Fe composition and different measurement areas: χ 2 (4) = 36.4, χ 2 (4) = 27.2, χ 2 (3) = 57.3 for Ca in β, α and γ dataset, respectively, and for Fe: χ 2 (4) = 15.4, χ 2 (4) = 22.1, χ 2 (3) = 24.0. This provides evidence to suggest that there is a difference between at least one pair of measurement areas. The Ti variable recognizes the difference among groups related to the issue (χ 2 (1) = 25.0, χ 2 (1) = 10.4, χ 2 (1) = 28.7 in β, α and γ datasets, respectively, (p < .01). However, this information is of limited use since visual inspection of banknotes provides an immediate answer.
In order to indicate a specific measurement area that differs significantly, a Nemenyi post-hoc test for pairwise multiple comparisons of the ranked data was applied to each of the datasets based on Ca and Fe content separately. The results are presented in Fig. 2. Areas A and C might be considered potential discriminatory points for banknotes regardless of their denomination or issue designation. The distinction between different datasets comprised of 10 and 20 zl banknotes as well as 50 and 100 zl banknotes seems to indicate that area D might be disregarded from the former dataset (10 and 20 zl) and the inclusion of measurement area E in the latter dataset is worth considering. Nevertheless, the dataset based on all banknotes confirms the results from visual interpretation (see section 3.1.).

LIBS optimization
In order to obtain reliable and most informative emission spectra of banknotes, an optimization procedure of the LIBS method was conducted. The main criterion for the selection of optimal analysis parameters was to obtain the largest possible number of signals and relatively high signal intensity (the discriminatory elements mainly occurred in trace amounts) without saturation of the spectrum. The amount of sample consumed and the size of resultant craters were also taken into consideration, in order to reduce the amount of ablated paper.
In the first step, the Q-switch delay [μs] was optimized. It is inversely proportional to the laser power. Maximum power = 25 MW is equal to Q-switch delay = 135 μs. For this purpose, a new 10 zl banknote was used and a series of measurements were carried out in the range of Q-switch delay from 135 to 255 μs. The optimal value was found to be 135 μs requiring the maximum power of the laser.
The next optimization step focused on the detector parameters, including gate delay (1.27-1.35 μs) and integration time (1.1-1.4 ms). In the case of gate delay, increasing the time resulted in decreasing signal intensities. Therefore, all measurements were performed with a gate delay of 1.27 μs. Regarding the integration time, the best spectra were obtained when their recording lasted 1.4 ms.
It was found that the conditioning shot, used for cleaning contamination from the surface before analysis, not only did not improve the spectrum but also caused spectral distortion. In fact, it was observed that the layer of colouring matter on the surface of the paper was so thin that a single laser shot was sufficient to penetrate through it and also cause ablation of the paper. Thus, a single accumulation was found optimal both for reducing sample destruction (Fig. S4 (Appendix)) and for obtaining an adequate and reliable spectrum. However, it was observed that a spectrum based on five shots yielded a significant enhancement of the S/N ratio. Therefore, data were collected from 5 different points of the same measuring area.    In order to evaluate the repeatability of the results, analyses of different measuring areas (5 repetitions) were performed on new 10 zl banknotes, both from OI and NI. The positions (wavelengths) of the spectral lines detected for the analysed samples were repeatable. Successive spectra only differed in the intensities of the signals (see Table 4), which are mainly related to the laser shot-to-shot fluctuation, the unrepeatable phenomenon of plasma creation and cooling, and the chosen place of analysis (the heterogeneity of the sample composition). The intensities of the most intense persistent lines for several elements varied from one measuring area to another, with an RSD value that ranged from 11.1 to 41.6. The lowest RSD was obtained for pure paper (10 zl NI), while the highest corresponded to metallic printing ink. Thus, to improve repeatability, the average spectrum obtained from five measurements in one measuring area was recorded after gently moving (within an area of 0.25 cm 2 ) the banknote, which was placed on a sample plate.

LIBS analysis
LIBS spectra from measuring area P for: 1) 5 brand new OI banknotes, 2) 5 used OI banknotes, and 3) 3 brand new NI banknotes were recorded under the optimized experimental conditions. Spectra registered for banknote papers were quite different from the spectrum of standard office paper [59]. The banknote paper gave a richer spectrum with a higher number of lines originating from neutral species (I), singly charged ions (II), and molecules (see Fig. 3). The presence of several elements such as Ti, Ca, Na, K, Al, and Cr was revealed. The presence of Ti was most likely related to the use of titanium dioxide during the beating process of cotton paper pulps in order to improve the fibrillation of cotton fibres and the distribution of fillers. The elemental characteristics of banknote paper might be used as a first step in the identification of fake banknotes.
Only quantitative differences in the elemental composition of paper were observed after comparing brand new banknotes from the same issue. On the other hand, there were dissimilarities between banknote papers from NI and OI. This was probably a consequence of the striped pattern of printing ink found in OI banknotes. Only the spectrum of NI banknotes (Fig. 4, red line) can be assumed to be derived solely from the emission spectrum of blank paper. In this study, the banknote was treated as an integrated system consisting of paper and colouring matter produced with high production standards, and accordingly, it was analysed as an inseparable sample.
To verify how impurities influence results, brand new banknotes were compared with used ones. For this purpose, averaged spectra corresponding to the same elemental measurement areas on new and used banknotes (of various denominations) were superimposed and studied signal by signal. In general, the results allowed the banknotes to be separated into two groups. The first group contained banknotes that do not differ from each other qualitatively, while in the second group the spectra of the used banknote contained additional spectral lines. For instance, after careful observation of the spectrum of a 10 zl banknote, several additional emission lines were visible (Fig. 4). These were persistent lines originating from Fe I: 371.965, 373.457, 373.670, 374.572, 374.942). The possible reason for the appearance of new spectral lines is the fact that these banknotes were used by people and show signs of contamination derived from many different sources, including hands, metal parts of wallets, purses, or pockets. The places showing higher exposure to contaminants are measuring areas B, C, and P, independently of the denomination, D on 10 and 20 zl bills, and F on 100 zl banknotes. On the other hand, the same emission spectra for the Table 4 The repeatability of peak intensities (the most intense ones) of elements obtained for the analysed banknotes using LIBS.

Sample
Description new and used banknotes were obtained in the case of area A on 10 and 20 zl notes, and areas A and E on 50 and 100 zl bills. A possible explanation for these relationships is based on their continuous use. For example, when taking a Polish banknote from a wallet, a person frequently grasps it in the middle (lengthwise), covering the portrait with a finger. However, when a Polish banknote is handed over to another person, usually the giver grasps it around the red serial number whilst the recipient takes the banknote from the opposite side. It is very important to take contamination into account when carrying out analysis of used banknotes. It must always be determined whether the differences are caused by counterfeiting or contamination. Analyses can be performed after attempts to gently clean the surface of the banknote. In doubtful situations, it is recommended to compare the elemental profile in other measuring areas with high discriminatory potential and with less likelihood of contamination. The elemental composition obtained from LIBS for selected measuring areas is listed in Table 5. The results differed from those originating from μXRF analysis but also showed technological dissimilarities in OI and NI banknotes manifesting in different elemental compositions of the same measuring areas (e.g. Ba in area A). On the other hand, the most common element found in almost every measuring area independently of issue is Mn. In some cases, XRF detected a larger number of elements in a particular measuring area, e.g. area A or B (compare Tables 3 and 5). However, there were also areas, for instance, C, D, and E, where a richer elemental composition was found by the LIBS technique. It is very difficult to find a clear correlation here. The different penetration depth and sensitivity of the two techniques were probably the reasons for these variations. It is worth mentioning that in the case of LIBS, the criteria applied during the interpretation of spectral lines were crucial [60,61]. For instance, Ni yielded a large number of low-intensity signals (almost at the noise level), which cannot be interpreted as significant spectral lines. What is more, there were cases in which many intense lines came from one element (e.g. Fe), but none of them was labelled as persistent in the NIST database. Under the established interpretation criteria, there were no grounds to claim that it was present. Furthermore, in contrast to XRF, LIBS elemental profiles from OI banknotes were relatively richer than those from their NI counterparts. A selection of the most discriminating areas was possible taking into account such qualitative information. In the case of OIs, the most unique elemental profile was in area C, whereas in the case of NIs there were three characteristic areas: 1) A for Ba, 2) C for Co, and 3) E for Mg.

Statistical analysis of LIBS results
Similarly to the data obtained from XRF, three datasets described by three independent variables were investigated: γ -banknotes of all denomination, α: 10 and 20 zl, and β: 50 and 100 zl banknotes. The intensities of spectral lines (persistent ones) of three elements (Ti, Ca and Zr) present in all analysed banknotes (only in areas A and C) were selected as dependent variables. Independent variables were the same as in the case of XRF data namely-measurement area, issue, and denomination.
A post-hoc test was not performed since only two areas (A and C) were analysed. These two areas can be considered potential discriminatory points for Polish banknotes, which is in accordance with the visual assessments.

Case studies results
After the analysis of the collected original banknotes, μXRF and LIBS analyses of suspected counterfeits were conducted using the same Fig. 4. Fragment of LIBS spectra of three 10 zl banknotes registered in area P: brand new OI banknote (green), used OI banknote (blue), brand new NI banknote (red). Peaks originating from impurities are indicated by arrows. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Table 5 Elemental profiles obtained by LIBS analysis for various areas of Polish banknotes of all denominations. methodology. The four different specimens of 10, 20, 50 and 100 Polish zloty banknotes from OI suspected to be counterfeited were investigated. All measurements were performed on area C -the black serial number -which was previously statistically selected as one of the most discriminatory parts of the banknote. In all four cases, it was possible to discriminate between counterfeit and authentic banknotes. The comparison criterion was based on the qualitative presence/absence of elements and semi-quantitative data. The variation of elemental profiles within area C of individual counterfeits and between areas C of the four counterfeits was smaller than the variation for genuine banknotes. A comparison of the obtained results revealed that every suspected banknote was made with the use of standard office paper, not banknote paper. Both the element profiles and the ratios of detected elements of the 4 examined samples were different from those of genuine banknotes.

Conclusions
The effectiveness of LIBS in the analysis of Polish banknotes was assessed and compared with that of μXRF. Although both techniques gave reproducible and comparable results, particular caution is advisable when evaluating the data. In LIBS, problematic spectral noise manifested by the presence or absence of different emission lines may be observed. The noise was dependent on the characteristic plasma evolution, which is a consequence of the laser fluctuation as well as the detector gate delay selected. This complicated the elemental interpretation. However, once this problem was solved by careful selection of valid spectral lines, reproducible spectra concerning qualitative differences between the same areas of banknotes were obtained. It was found that there was a consistent composition of the ink and the paper between different banknotes of the same denomination and issue, respectively.
This confirms the use of high-quality control standards in the production of banknotes starting from the paper and ending with a variety of security features. However, the radiation employed in μXRF has the ability to penetrate entirely through the sample resulting in misinformation, since elements from the second side of the banknote may be detected in the spectrum.
Considering the quantitative repeatability of the LIBS method, RSD values oscillated around 28%. After comparing it with μXRF, it was found that the RSD of the net peak areas was lower (about 20%). However, in both cases, the intensities/areas of the lines strongly varied from one measuring point to another. A possible way of overcoming the inhomogeneity problem is to work with both techniques in elemental imaging (mapping) mode. Scanning a sample beneath the X-ray microbeam or the laser beam allows detailed hyperspectral elemental images to be obtained. These contain both spatial and spectral information which can be used to study element distribution.
The results also indicate that LIBS allows us to distinguish individual features of a banknote. Identification of the obtained lines provided enough evidence about specific elemental profiles for each particular measurement area. However, LIBS elemental compositions were different from those detected using μXRF. This is a complex issue, presumably related to the physicochemical principles (e.g. penetration of radiation) and sensitivity of each technique. The protocols employed during the interpretation of LIBS spectral lines -at least two persistent lines of an element were required to give grounds to claim that it was present -were also very important. The sensitivity of LIBS is theoretically greater. Due to the lack of adequate standards, it was very difficult to estimate the actual difference. Regarding the interpretation criteria, there were cases in which a lot of intense lines coming from one particular element were observed in the LIBS spectrum, but none of them was persistent. Such signals should not be unequivocally interpreted as significant spectral lines.
Nevertheless, thanks to a very careful analysis of the LIBS emission spectra, many differences between the measuring areas were observed. The most characteristic spectrum was obtained for the points C, A, and E (on 50 and 100 zl banknotes, especially in the case of NI). This is consistent with μXRF results and was confirmed by statistical analysis. In conclusion, the developed LIBS method can be used for the analysis of banknotes since the results can be relevant for judicial purposes. However, the number of banknotes used in the study was relatively small and therefore larger datasets are needed to report the power of the test, to estimate error rates and uncertainty, and to be able to generalize the above-mentioned conclusions. Given its low cost, high sample throughput, and ease of use, the application of LIBS to banknote examination has been shown to be as effective as μXRF. In summary, LIBS is a viable alternative to μXRF, but a multi-method approach provides an adequate more effective and reliable way of comparing and validating the results.
Future research includes analysing a larger number of used banknotes in order to better assess the most frequently found impurities. This will allow us to evaluate the adequacy of the proposed methods for routine forensic analysis. It is worth emphasizing at this point that additional studies involving more 'real-world' counterfeit banknotes should also be performed.