Analysis of Multi-Elemental Thin Films Via Calibration-Free Laser-Induced Breakdown Spectroscopy Jörg Hermann, Emanuel Axente, Frédéric Pelascini, Valentin Craciun

Home , Pulsed laser deposition

Analysis of Multi-elemental Thin Films via Calibration-Free Laser-Induced Breakdown Spectroscopy Jörg Hermann, Emanuel Axente, Frédéric Pelascini, Valentin Craciun

To cite this version:

Jörg Hermann, Emanuel Axente, Frédéric Pelascini, Valentin Craciun. Analysis of Multi-elemental Thin Films via Calibration-Free Laser-Induced Breakdown Spectroscopy. Analytical Chemistry, American Chemical Society, 2019, 91 (3), pp.2544-2550. 10.1021/acs.analchem.8b05780. hal- 02114118

HAL Id: hal-02114118 https://hal.archives-ouvertes.fr/hal-02114118 Submitted on 23 May 2019

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Analysis of multielemental thin ﬁlms via calibration-free laser-induced breakdown spectroscopy

J¨orgHermann,∗,† Emanuel Axente,‡ Fre´de´ric Pelascini,¶ and Valentin Craciun‡

Aix-Marseille University, CNRS, LP3, 13009 Marseille, France, National Institute for Lasers, Plasma and Radiation Physics, 77125 Mˇagurele, Romania, and Cetim Grand Est, 67305 Schiltigheim, France

E-mail: [email protected]

Abstract This technique could be used even for thinner films provided that the film-composing ele- Elemental analyses of thin films with com- ments are not present in the substrate. plex composition are challenging as the standard analytical techniques based on measurement calibration are difficult to apply. We show Introduction that calibration-free laser-induced breakdown spectroscopy (LIBS) presents a powerful solu- Thin films attracted strongly growing inter- tion, enabling quantitative analyses of multiele- est during the last decades in many fields mental thin films with analytical performances of applications such as optics, optoelectronics, better than those obtained with other tech- biomedicine, and aeronautics. Consequently, niques. The demonstration is given for a nickel- great efforts have been dedicated to the devel- chromium-molybdenum alloy film of 150 nm opment of techniques for thin film production thickness that was produced by pulsed laser de- and pulsed laser deposition (PLD) emerged as position. The LIBS spectra were recorded in a versatile laboratory-based method after the experimental conditions that enable simple and successful synthesis of high-temperature supra- accurate modeling of plasma emission. Thus, conducting yttrium barium copper oxide layers 1 a calibration-free approach based on the cal- in the late 80’s. culation of the spectral radiance of a uniform The increasing demand of high-quality coat- plasma in local thermodynamic equilibrium was ings with well-defined optical, electronic or me- applied to deduce the elemental composition. chanical properties stimulated the development Supported by analyses via Rutherford backscat- of techniques for thin film characterization. tering spectrometry and energy-dispersive X- Beside the determination of cristallographical ray spectroscopy, the LIBS measurements ev- structure and optoelectronical properties, the idence nonstoichiometric mass transfer of the measurement of the elemental composition is of alloy during the thin film deposition process. great interest. Rutherford backscattering spectrometry (RBS) and X-ray photoelectron spec- ∗To whom correspondence should be addressed † trometry (XPS) are currently applied to mea- Aix-Marseille University, CNRS, LP3, 13009 Mar- sure elemental fractions within the deposited seille, France 2–4 ‡National Institute for Lasers, Plasma and Radiation layer. According to the lower depth res- Physics, 77125 Mˇagurele,Romania olution, energy-dispersive X-ray spectrometry ¶Cetim Grand Est, 67305 Schiltigheim, France (EDS) is suitable for analysis of layers with

1 thickness ≥ 1 µm. The analytical techniques over, these analyses were mostly carried out based on the calibration with standard sam- with very simple approaches that ignore self- ples such as inductively coupled plasma mass or absorption and thus suffer from low accuracy.15 atomic emission spectrometry (ICP-MS or ICP- In the present paper, we demonstrate that AES) are inapplicable.5 Indeed, ICP analyses quantitative analysis of thin films with com- are typically used for liquid sampling, while ex- plex composition are possible via calibration- tensive sample preparation (dissolution) is re- free LIBS measurements based on accurate and quired for solids.6 Although direct ICP analysis robust modeling. The method takes benefit of solids is possible via laser ablation (LA) sam- from experimental conditions that were recently pling, LA-ICP measurements are affected by found to generate a plume well described by the difficulties related to sample size, shape, hetero- simple model of a uniform plasma in local ther- geneity, and the lack of matrix-matched stan- modynamic equilibrium.16 dards.7 Beside the cost- and time-expensive use of RBS and XPS analyses, these techniques are limited Experimental section to materials having a relatively simple elemental composition.8 Moreover, RBS and XPS suf- LIBS setup fer low sensitivity and elements cannot be ac- The experiments were carried out with Nd:YAG curately measured in many cases if their mass laser pulses of 6 mJ energy and 4 ns dura- fraction within the thin film is below 1%. The tion. The laser was operated in the ultravio- poor limit of detection is usually not a handi- let spectral range (266 nm) to obtain efficient cap in thin film analysis as the deposited lay- energy coupling towards the sample material. ers mostly have a simple elemental composition: The beam was focused to a spot of 100 µm di- they include a few elements only with relative ameter on the sample surface, leading to a laser large abundances. fluence of about 100 J cm−2 that is high enough However, some materials such as wear- or to ensure stoichiometic ablation.17 The samples corrosion-resistant coatings have more complex were placed on a motorized sample holder in a elemental composition. Among them, superal- vacuum chamber that was filled with argon at loys such as nickel-chromium-molybdenum al- 5 × 104 Pa pressure. loys are of great interest. Their tribological Optical emission spectroscopic measurements properties critically depend on the elemental were performed by imaging the plasma with composition and accurate compositional anal- two lenses of 150 and 35 mm focal lengths onto yses including major and minor elements are the entrance of an optical fiber of 600 µm core required for quality control. diameter. The fiber was coupled to the en- As an emerging technique for fast multiele- trance of an echelle spectrometer with a resolv- mental analysis, laser-induced breakdown spec- ing power of 9 × 103. Photon detection was en- 9 troscopy was proposed for thin film analysis. sured using an intensified charge-coupled device However, due to the ablation depth being typ- matrix detector. The apparatus spectral width ically larger than several hundreds of nanome- and the apparatus response were measured as ters, accurate LIBS analyses based on calibra- functions of wavelength with appropriate cali- tion with standard samples were limited to films brated lamps. 10 of thickness ≥ 1 µm. To overcome the diffi- The spectra were recorded with a gate width culties due to calibration, thin films were an- ∆t short enough to make sure that the varia- 11–13 g alyzed via calibration-free LIBS methods. tions of electron density and plasma tempera- These techniques are based on modeling of ture during the recording were small compared the emission spectrum of the laser-produced to their absolute values. Typically, ∆tg ≤ td/2 plasma.12,14 The reported calibration-free anal- was applied, where td is the delay between the yses only concern thin films of simple composi- laser pulse and the detector gate. We denote tion, containing 3 elements at maximum. More-

2 Table 1: Mass fraction deduced from the LIBS Computational details analysis CLIBS, atomic mass mA, and first ion- Material ablation with ultraviolet laser pulses ization potential ∆ion of elements composing the nickel-chromium-molybdenum alloy (Hart of nanosecond duration in a near-atmospheric b.v., Inconel 625) used for thin film deposition. argon atmosphere generates a plasma with ideal properties for spectroscopic measurements, The CLIBS-values correspond to the recording delay of highest measurement accuracy (see combining two attributes that are usually not Fig. 7). found together: the plasma is (i) spatially uniform and (ii) well described by the model of local thermodynamic equilibrium (LTE).16 Con- Element CLIBS (%) mA (amu) ∆ion (eV) sequently, the spectral radiance of the plasma Ni 63.1 ± 3.6 58.70 7.64 can be calculated using a simple analytical so- Cr 21.2 ± 3.1 52.00 6.77 19 Mo 9.4 ± 1.4 95.94 7.09 lution of the radiation transfer equation. Nb 3.7 ± 1.2 92.91 6.76 We have implemented the calculation of the Fe 1.7 ± 0.3 55.85 7.90 spectral radiance in an iterative measurement Mn 0.07 ± 0.03 54.94 7.43 procedure that allows us to deduce the ele- Si 0.35 ± 0.10 28.09 8.15 mental composition from the best agreement Al 0.30 ± 0.07 26.98 5.99 between measured and computed spectra.20 Ti 0.24 ± 0.05 47.90 6.83 Co 0.10 ± 0.03 58.93 7.86 The procedure consists of two main measure- Ca 0.018 ± 0.005 40.08 6.11 ment loops. The principal loop includes the successive measurements of electron density ne, plasma temperature T and the fractions Ci of the time of observation t = td + ∆tg/2. To en- the N elements. For each measurement, the hance the signal-to-noise ratio, data acquisition corresponding parameter is varied in order to was performed by averaging over 100 ablation find the best agreement between the computed events, applying a single pulse on each irradia- and measured spectra in the wavelength ranges tion site. We note that the present experiments of the relevant atomic or ionic transitions. The were carried out with a spectroscopic appara- calculations include the precise description of tus of low efficiency (echelle spectrometer) and the spectral line profile, considering the domi- that the acquisition number can be reduced to nating line broadening effects, namely Doppler a few repetitions if a more efficient system is and Stark broadening. For spectral lines of un- used. The ablation depths measured via optical known Stark broadening parameters, the Stark microscopy were 0.55 and 1.0 µm for the bulk width and shift are measured in a separate ex- sample and the deposited thin film, respectively periment.21,22 The procedure also includes an (see SI 1). algorithm for the selection of the most appro- The thin film was produced via pulsed laser de- priate lines according to their optical thickness, position onto an electronic grade low resistiv- signal-to-noise ratio, interference with other ity Si (100) substrate of 99.999% purity. The transitions, and the accuracy of available spec- film thickness of 150 nm was deduced from RBS troscopic data. The data were taken in priority analysis (see Fig. S1). A detailed description of from the NIST database23 and completed by the PLD arrangement can be found in a pre- data from the Kurucz database.24 18 vious paper. The alloy target is irradiated in The calibration-free measurement algorithm −5 a vacuum chamber at 10 Pa pressure. Here, was successfully validated for glasses and al- −2 a moderate laser fluence of 2 J cm is applied loys.20,25 It was shown that the model validity to obtain a smooth film. The nickel-chromium- and thus the composition measurement accu- molybdenum alloy (see Table 1) was chosen as racy critically depend on the time of observa- sample material for its complex multielemental tion. For elements such as oxygen, character- composition. ized by an atomic structure with large energy gaps between the electronic excitation levels,

3 19 6 1×10 1×10 in LIBS plasmas. The pressure of the plasma

) 18 is shown to reach a minimum for T ' 7000 K.

-3 1×10 + ne Ni + 5 The minimum below the ambient gas pressure Ni Cr 1×10 17 + 1×10 Mo is expected after the moment when the plasma Cr 2+ reaches the maximum volume. The recompres- 2+ × 16 Mo + Mo 1 10 Nb + Ni Fe 4 sion afterwards is illustrated by the increase 1×10 Pressure (Pa) 2+ 2+ Nb × 15 Cr of pressure in the low-temperature range, sim-

Number density (cm 1 10 Fe 2+ ilar to the implosion that succeeds a classical Nb Fe 14 27 1×10 5000 10000 15000 20000 explosion. Temperature (K) Figure 1: Number densities of species vs tem- Results and discussion perature computed for a plasma in LTE with the elemental composition of the Inconel 625 The calibration-free LIBS measurement proce- alloy (see Table 1). The calculations were per- dure was applied to spectra recorded with dif- formed for the ne = f(T )-dependence deduced ferent gate delays in order to find the obser- from the measurements (see Fig. 4), assum- vation time for highest analytical performance. ing a negligible fraction of argon in the vapor The electron density was deduced from Stark plasma. The calculated kinetic pressure is pre- broadening of spectral lines having known Stark sented with respect to the secondary y-axis. broadening parameters, assuming a linear dependence between Stark width and electron density.28 As an example, the spectral profiles the equilibrium condition is only fulfilled for of the Ca II 393.36 nm transition (see Table S1) t < 1 µs, when n > 1×1017cm−3.20,25 Contrar- e measured for different delays are displayed in ily, metal atoms typically have many excitation Fig. 2 together with the computed line profiles. levels separated by small energy gaps. The According to the low fraction of calcium (see lifetime of local thermodynamic equilibrium is Table 1) the self-absorption of the resonance therefore significantly longer in case of metallic line is small and n can be measured with fair plasmas.25,26 e accuracy for t ≤ 1 µs, when the Stark width is The number densities of plasma species com- large enough. puted for LTE are displayed in Fig. 1 as func- Taking into account all sources of measure- tions of temperature for the elemental com- ment errors, the accuracy of n -measurements position of the considered nickel-chromium- e molybdenum alloy. For a clear presentation, only the most abundant elements are displayed. 0.50 +/- 0.10 µs measured According to the moderate differences in ion- computed ization energy (see Table 1), the T -dependence 0.85 +/- 0.15 µs of atomic and ionic number densities is sim- 1.25 +/- 0.25 µs ilar for all elements. For T < 7000 K, neu- Ca II 393.36 nm tral atoms dominate the plasma composition. 2.50 +/- 0.5 µs Their number densities are shown to reach a minimum for T ' 10000 K. The increase of the atomic number densities with temperature for Spectral radiance (arb. units) T > 10000 K is attributed to the high density 393.0 393.2 393.4 393.6 of the laser-produced plasma during the early Wavelength (nm) expansion stage.27 The high density favors the Figure 2: Measured and computed profiles of recombination processes, and the temperature Ca II 393.36 nm. The recordings were per- required to obtain a given degree of ionization formed during ablation of the bulk alloy sample increases with density. For the same reason, for different times of observation. doubly charged ions are generally not observed

4 19 1×10 5 (a) 15000

nickel ) -3

18 0 chromium 1×10 10000

-5 7000 17 molybdenum 1×10

niobium ) 5000 Temperature (K)

u -10 g Electron density (cm

ul 16 1×10 A -155 / 0.2 0.5 1 2 5 10 20 ε λ (b) Time (µs) 0 15000 +/- 200 K ln ( 13800 +/- 200 K Figure 4: Electron density and temperature vs -5 time measured for the bulk sample (filled sym- 11100 +/- 100 K bols) and the thin film (empty symbols). The -10 8800 +/- 100 K vertical error bars represent the measurement 6900 +/- 100 K uncertainties whereas the horizontal error bars -15 indicate the detector gate width. 5 10 15 20 Upper level energy (eV) measurements were achieved over the entire Figure 3: Saha-Boltzmann plots (a) for most considered time interval. abundant elements at t = 0.85 µs, and (b) for The electron density and plasma temperature chromium at 0.35 µs (squares), 0.5 µs (circles), measured for both the bulk sample and the de- 0.85 µs (triangles up), 2.5 µs (triangles down), posited thin film are displayed in Fig. 4 as func- and 6.0 µs (diamonds). In the logarithmic functions of time. It is shown that the properties of tion, ε, λ, Aul and gu are the emission coeffi- the plasma produced in both cases are equal cient, the wavelength, the transition probabil- within the measurement uncertainties. This ity and the upper level statistical weight of the behaviour is expected for laser ablation with lines, respectively. fluences strongly above the ablation threshold where the properties of the solid material play a minor role and the breakdown process is mainly is estimated to about 20% in the time interval governed by the attributes of the ablated va- from 0.5 to 2 µs. At later times, the measure- por atoms. The time evolution of n and T is ment uncertainty increases as the Stark widths e approximated by functions f(t) = AtB (contin- of most lines become small compared to the ap- uous lines) with the parameters A and B de- paratus spectral width while an decrease of ac- duced from the best fit. We obtain n (t) = curacy due to the simplified theoretical descrip- e 2.19 × 1020 t−1.07 and T (t) = 4.93 × 104 t−0.22, tion (temperature dependence is ignored) is ex- where t, n and T are in units of ns, cm−3 and pected for shorter times. e K, respectively. The combination of both ex- The alloy-forming most abundant elements are pressions gives the temperature dependence of characterized by rich emission spectra, but low electron density n (T ) = 0.0038 T 4.85 that was accuracy of the available spectroscopic data. e used as boundary condition for the calculations Thus, to reduce measurement errors due to in- of the LTE plasma composition displayed in accurate data, a large number of lines was in- Fig. 1. cluded in the analysis. This is illustrated by The capability of compositional measure- the Saha-Boltzmann plot displayed in Fig. 3 (a) ments is illustrated by the spectra displayed where the parallel slopes indicate the high ac- in Fig. 5. The two selected spectral ranges curacy of the deduced temperature values. The exhibit transitions of the most abundant ele- Saha-Boltzmann plots obtained for chromium ments of the nickel-chromium-molybdenum al- at different times (b) show that accurate T - loy. Compared to the line intensity distribution

5 oun)wr esrdvaEDS. via (black measured values de- were reference columns) gate The S2). different Table sample (see with abundant lays 625 recorded most Inconel spectra of the from for fractions deduced Mass elements 6: Figure (a) sample alloy bulk the of for ablation (b) molybdenum, during film chromium, performed thin from were deposited transitions recordings the of The and radiance nickel. spectral and computed niobium and Measured 5: Figure Mass fraction (%) analysis. LIBS high calibration-free the of indicates accuracy spectra mea- computed between agreement and good sured The Tables S3). (see and delays S2 gate spectra of different fac- analysis with by a recorded confirmed by was four niobium defi- of and tor The molybdenum process. of deposition ciency laser pulsed during the transfer indi- mass composition, nonstoichiometric the a of cating in change de- We lower significant much (b). a film are duce thin lines the re- for Cr recorded with or spectrum transitions Ni Nb relative to and the spect Mo (a), sample of bulk intensities the for observed S1. Table in listed are measurements 100 10 1 Spectral radiance (arb. units) Ni Cr Cr 312.0 + + Cr Cr Mo II312.19 Mo II312.19 + + Cr Cr Cr + + 312.5 Mo Cr Cr Nb II312.75 Nb II312.75 + + Nb II313.08 Nb II313.08 Cr Cr 313.0 Nb + + t 1 = Cr + Mo II313.25 0.35 0.23 EDS 12.50 6.00 2.50 1.25 0.85 0.50 Fe . 25 313.5 Ni Ni Wavelength (nm) Cr µ µ µ µ µ µ µ ± Cr µ s s s s s s s + s + 0 Cr Cr . 25 + 6 + Mo II313.87 µ 1. 319.0 314.0 iu nlsso aluminum. pre- of with analysis accordance con- in vious entire interval, the over time model sidered LTE the the The of indicate validity fractions times. element measurement constant various almost for 6 presented ele- Fig. are abundant in alloy bulk most the for the measured of ments fractions mass The n b vlae rmtedffrnewt the EDS. with via difference measured the values algorithm from evaluated LIBS (b) pre- calibration-free and the (a) by time fraction dicted measurement Fe and vs Mo of measurements errors Relative 7: Figure measurement fraction the estimated we ysis, spectra from of recorded analysis LIBS observed calibration-free is the values for reference EDS the with ment ofidtegt ea o otacrt anal- accurate most for delay gate the find To Measured error (%) Computed error (%) composition the for used lines spectral The s. 20 40 60 20 40 60 0 0 0.2 (b) (a) computed measured Nb II318.93 0.5 Nb II319.11 t 0 = Nb II319.14 . o6 to 5 2 1 Mo I319.40 Time ( Mo II319.21 Fe bulk Mo bulk Fe film Mo film Fe bulk Mo bulk µ Nb II319.50 µ s. Mo I319.40 s) 319.5 25 Nb II319.50 odagree- good A 5 Cr (b) (a) Cr + + 320.0 10 errors for all measurement times. The char- 100 10 bulk target (a) acteristic time evolution is displayed in Fig. 7 thin film for molybdenum and iron. According to their 1 abundances, the measurement errors of Mo and 0.1 Fe are representative for elements with mass 0.01 fractions of the order of 10% and 1%, respec- Mass fraction (%) LIBS (b) tively. 1.0 RBS The errors deduced from the spectra simula- bulk / C 0.5 tion are presented for both the bulk sample and film the thin film (a). The error evaluation takes C into account all relevant error sources, including 0 Ni Cr Mo Nb Fe Al (Si) Ti Co Mn Ca (i) intensity measurement errors due to signal- to-noise ratio and uncertainty of the appara- Figure 8: (a) Elemental compositions of the tus response, (ii) accuracy of spectroscopic data bulk alloy target and the thin film deduced from and, (iii) self-absorption. For both elements, the spectra recorded for t = 1.25 and 2.50 µs. the most accurate measurements were achiev- (b) The deduced fraction ratio Cfilm/Cbulk is able in the time interval from 1 to 2 µs. For compared the to ratio measured via RBS for smaller recording times, the measurement error the most abundant elements (see Table S4). increases mostly due to enhanced Stark broadening and the increase of the continuum emis- The elemental compositions measured for both sion intensity. This leads to stronger line in- the bulk alloy sample and the thin film are dis- terferences and lower signal-to-noise ratio. For played in Fig. 8 (a). To illustrate the charac- t > 2 µs, the self-absorption of the most in- teristic changes, the fraction ratios C /C tense lines (having small upper level energies) film bulk are presented in (b). The amount of silicon (in increases. Thus, the measurements have to be parentheses) within the thin film could not be performed with lines of reduced intensities for measured due to the contribution of the Si sub- which the signal-to-noise ratio is low. In ad- strate to the LIBS signal. The Si-fraction was dition, the uncertainty of the electron density thus set equal to the value measured for the measurement increases as a consequence of line bulk sample. We emphasize that this adjust- narrowing. The minimum error of about 10 to ment is done for the presentation of the final 15% obtained for the intermediary time interval result only. The LIBS measurements procedure is mainly due to the uncertainty of the available was operated with the silicon amount from both spectroscopic data. thin film and substrate. The measurement errors evaluated from the dif- A good agreement with the fraction ratios ference between the mass fractions measured by measured via RBS is observed for the most LIBS and those obtained via EDS are displayed abundant elements for which RBS measure- in Fig. 7 (b) for the bulk sample. Similar to the ments were possible (see Fig. S1). Compared behavior of the computed errors (a), the time to their abundance in the bulk target, the frac- evolution of the measured errors (b) reach a tions of molybdenum and niobium in the film minimum located in the time interval from 1 to are strongly reduced. Accordingly, the amount 2 µs. However, the measured errors are system- of the major elements (nickel and chromium) is atically smaller, indicating that the accuracy of slightly increased. the used spectroscopic data is better than the The deficiency of Mo and Nb in the thin film ev- given uncertainty. The measured error is not idences a nonstoichiometric mass transfer dur- presented for the thin film (b) as both RBS and ing the pulsed laser deposition process. This EDS analysis had too low accuracy. A further phenomenon is often observed in PLD and ex- reduction of the measurement error down to 5% plained by different mechanisms. Among them, could be readily expected with the availability the mass-dependent (see Table 1) angular dis- of more accurate spectroscopic data.

7 tributions of atoms and ions in the plume cal thermodynamic equilibrium. Analyses per- and non-stoichiometric ablation have been re- formed for different delays between the laser ported.29,30 Operated with low laser fluence pulse and the detector gate reveal that the most to minimize droplet generation, the ablation accurate measurements are obtained in the time mechanism in PLD is close to thermal evap- interval from 1 to 2 µs. In this condition, the oration for which the elemental compositions measurement accuracy is mostly limited by the in the liquid and in the gas phase differ as a uncertainty of the used spectroscopic data, and consequence of the element-dependent values of the reported errors of about 15% can be reduced boiling temperature and vaporization enthalpy down to 5% if accurate data are available. (in Table S6).17 It was recently shown that the Analyzing both the bulk alloy sample and the composition of a thin film deposited via PLD 150 nm thin film produced by pulsed laser de- from a copper-tungsten alloy target could be position from the sample, a nonstoichiometric tuned from pure copper to 97% tungsten by mass transfer during PLD is evidenced. The changing laser fluence and angular position of calibration-free LIBS measurement procedure is the substrate.31 The mass-dependent sticking expected to be applicable to thinner films if the probability of atoms and ions on the thin film film-composing elements are not present in the was also considered.32,33 In the present experi- substrate. ment, the contribution of nonstoichiometric vaporization is indicated by EDS analysis that show an enrichment of molybdenum and nio- Acknowledgments bium on the alloy target in the zone irradiated Jacques Perrière is acknowledged for fruitful by the laser during pulsed laser deposition (see discussions on RBS analyses. The research Table S5, Figs. S2 and S3). leading to these results has received funding The calcium fraction within the thin film is also from LASERLAB-EUROPE (Grant Agreement found to be significantly smaller than the value No. 654148, European Union’s Horizon 2020 measured for the bulk sample. This change is research and innovation programme). EA and attributed to the nonuniform distribution of Ca VC acknowledge the financial support from Ro- within the bulk sample, characterized by a max- manian national projects NUCLEU-LAPLAS imum value at the surface and a rapid decrease and STAR 161. as a function of depth (see Fig. S4).

Supporting Information Available: Ad- Conclusion ditional data and information on EDS and RBS analyses are given. This material is avail- The present results demonstrate that composi- able free of charge via the Internet at http: tional analyses of multielemental thin ﬁlms can //pubs.acs.org/. be performed via calibration-free laser-induced breakdown spectroscopy with analytical performances better than those obtained with other References techniques. The demonstration was given for a (1) Dijkkamp, D.; Venkatesan, T.; Wu, X. D.; nickel-chromium-molybdenum alloy composed Shaheen, S. A.; Jisrawi, N.; Minlee, Y. H.; of ﬁve elements with mass fractions > 1% and Mclean, W. L.; Croft, M. Appl. Phys. Lett. several elements of lower abundance. To ensure 1987, 51, 619–621. simple and accurate modeling of the plasma emission spectrum, the LIBS experiments were (2) Sie, S. H. In Surface analysis methods in carried out with ultraviolet nanosecond laser materials science; O’Connor, D. J., Sex- pulses in a near-atmospheric argon pressure. ton, B. A., Smart, R. S. C., Eds.; Springer, These conditions were previously shown to en- Berlin, 2003; Vol. 23; pp 229–246. able the formation of a uniform plasma in lo-

8 (3) Jeynes, C.; Barradas, N. P.; Szil´agyi, E. (16) Hermann, J.; Grojo, D.; Axente, E.; Ger- Anal. Chem. 2012, 84, 6061–6069. hard, C.; Burger, M.; Craciun, V. Phys. Rev. E 2017, 96, 053210 1–6. (4) Kibel, M. H. In Surface analysis methods in materials science; O’Connor, D. J., (17) Mao, X. L.; Ciocan, A. C.; Russo, R. E. Sexton, B. A., Smart, R. S. C., Eds.; Appl. Spectrosc. 1998, 52, 913–918. Springer, Berlin, 2003; Vol. 23; pp 175– 202. (18) Craciun, D.; Bourne, G.; Socol, G.; Ste- fan, N.; Dorcioman, G.; Lambers, E.; (5) Olesik, J. W. Anal. Chem. 1991, 63, A12– Craciun, V. Appl. Surf. Sci. 2011, 257, A21. 5332–5336.

(6) Groh, S.; Diwakar, P. K.; Garcia, C. C.; (19) Cooper, J. Rep. Prog. Phys. 1966, 29, 35– Murtazin, A.; Hahn, D. W.; Niemax, K. 130. Anal. Chem. 2010, 82, 2568–2573. (20) Gerhard, C.; Hermann, J.; Mer- (7) Cahoon, E. M.; Almirall, J. R. Anal. cadier, L.; Loewenthal, L.; Axente, E.; Chem. 2012, 84, 2239–2244. Luculescu, C. R.; Sarnet, T.; Sentis, M.; (8) Climent-Font, A.; Fernández- Viöl, W. Spectrochim. Acta Part B: Atom. Jiménez, M. T.; Wätjen, U.; Perrière,J. Spectrosc. 2014, 101, 32–45. Nucl. Instr. Meth. Phys. Res. A 1994, (21) Cirisan, M.; Cvejić, M.; Gavrilović, M. R.; 353, 575–578. Jovicević, S.; Konjević, N.; Hermann, J. J. (9) Cabalin, L. M.; Laserna, J. J. Anal. Chem. Quant. Spectrosc. Radiat. Transfer 2014, 2001, 73, 1120–1125. 133, 652–662. (10) In, J.-H.; Kim, C.-K.; Lee, S.-H.; Choi, J.- (22) Burger, M.; Hermann, J. Spectrochim. H.; Jeong, S. Thin Solid Films 2015, 579, Acta Part B: Atom. Spectrosc. 2016, 122, 89–94. 118–126. (11) Widjonarko, N. E.; Perkins, J. D.; (23) Kramida, A.; Ralchenko, Y.; Reader, J. Leisch, J. E.; Parilla, P. A.; Curtis, C. J.; NIST Atomic Spectra Database (version Ginley, D. S.; Berry, J. J. Rev. Sci. In- 5.5.6). 2018; http://physics.nist.gov/ strum. 2010, 81, 073103 1–8. asd, National Institute of Standards and Technology, Gaithersburg, MD. (12) Axente, E.; Hermann, J.; Socol, G.; Mer- cadier, L.; Beldjilali, S. A.; Cirisan, M.; (24) Smith, P. L.; Heise, C.; Esmond, J. R.; Luculescu, C. R.; Ristoscu, C.; Mi- Kurucz, R. L. Atomic spectral line hailescu, I. N.; Craciun, V. J. Anal. At. database built from atomic data files Spectrom. 2014, 29, 553–564. from R. L. Kurucz CD-ROM 23. 2011; http://www.pmp.uni-hannover.de/ (13) Davari, S. A.; Hu, S.; Pamu, R.; Mukher- cgi-bin/ssi/test/kurucz/sekur.html. jee, D. J. Anal. At. Spectrom. 2017, 32, 1378–1387. (25) Hermann, J.; Lorusso, A.; Perrone, A.; (14) Ciucci, A.; Corsi, M.; Palleschi, V.; Strafella, F.; Dutouquet, C.; Torralba, B. Rastelli, S.; Salvetti, A.; Tognoni, E. Appl. Phys. Rev. E 2015, 92, 053103 1–15. Spectrosc. 1999, 53, 960–964. (26) Cristoforetti, G.; Lorenzetti, G.; Leg- (15) Bulajic, D.; Corsi, M.; Cristoforetti, G.; naioli, S.; Palleschi, V. Spectrochim. Acta Legnaioli, S.; Palleschi, V.; Salvetti, A.; Part B: Atom. Spectrosc. 2010, 65, 787– Tognoni, E. Spectrochim. Acta Part B: 796. Atom. Spectrosc. 2002, 57, 339–353.

9 (27) Hermann, J.; Axente, E.; Craciun, V.; Taleb, A.; Pelascini, F. Spectrochim. Acta Part B: Atom. Spectrosc. 2018, 143, 63– 70.

(28) Tognoni, E.; Palleschi, V.; Corsi, M.; Cristoforetti, G.; Omenetto, N.; Gor- nushkin, I.; Smith, B. W.; Wine- fordner, J. D. In Laser-induced breakdown spectroscopy; Miziolek, A. W., Palleschi, V., Schechter, I., Eds.; Cam- bridge University, Berlin, 2006; pp 122– 194.

(29) Gonzalo, J.; Afonso, C. N.; Perri`ere,J. J. Appl. Phys. 1996, 79, 8042–8046.

(30) Ohnishi, T.; Lippmaa, M.; Yamamoto, T.; Meguro, S.; Koinuma, H. Appl. Phys. Lett. 2005, 87, 241919 1–3.

(31) Klamt, C.; Dittrich, A.; Jaquet, B.; Eberl, C.; Doering, F.; Krebs, H.-U. Appl. Phys. A: Mat. Sci. Proc. 2016, 122 .

(32) Itina, T. E. J. Appl. Phys. 2001, 89, 740– 746.

(33) Droubay, T. C.; Qiao, L.; Kaspar, T. C.; Engelhard, M. H.; Shutthanandan, V.; Chambers, S. A. Appl. Phys. Lett. 2010, 97 .