Spectrochimica Acta Part B 65 (2010) 1–14

Contents lists available at ScienceDirect

Spectrochimica Acta Part B

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s a b

Review

Calibration-Free Laser-Induced Breakdown Spectroscopy: State of the art

E. Tognoni ⁎, G. Cristoforetti, S. Legnaioli, V. Palleschi
Istituto per i Processi Chimico-Fisici del CNR, Area della Ricerca di Pisa, Via G. Moruzzi 1, 56124 Pisa-Italy

a r t i c l e i n f o a b s t r a c t

Article history: The aim of this paper is offering a critical review of Calibration-Free Laser-Induced Breakdown Spectroscopy
Received 14 August 2009 (CF-LIBS), the approach of multi-elemental quantitative analysis of LIBS spectra, based on the measurement
Accepted 10 November 2009 of line intensities and plasma properties (plasma electron density and temperature) and on the assumption
Available online 17 November 2009
of a Boltzmann population of excited levels, which does not require the use of calibration curves or matrix-
matched standards. The first part of this review focuses on the applications of the CF-LIBS method.
Keywords:
Quantitative results reported in the literature, obtained in the analysis of various materials and in a wide
CF-LIBS
Quantitative analysis range of experimental conditions, are summarized, with a special emphasis on the departure from nominal
composition values. The second part is a discussion of the simplifying assumptions which lie at the basis of
the CF-LIBS algorithm (stoichiometric ablation and complete atomization, thermal equilibrium, homo-
geneous plasma, thin radiation, detection of all elements). The inspection of the literature suggests that the
CF-LIBS method is more accurate in analyzing metallic alloys rather than dielectrics. However, the full
exploitation of the method seems to be still far to come, especially for the lack of a complete characterization
of the effects of experimental constraints. However, some general directions can be suggested to help the
analyst in designing LIBS measurements in a way which is more suited for CF-LIBS analysis.
© 2009 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Overview of CF-LIBS applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1. Definition of a metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2. Accuracy of CF-LIBS results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. The CF-LIBS approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1. Brief recall of the basic idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.2. Experimental requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3. Variants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4. Plasma non-ideality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.1. Failure of stoichiometric composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2. Failure of local thermal equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.3. Failure of spatial homogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.4. Failure of optical thinness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5. Completeness and other experimental issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.1. Reduced composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.2. Exploitation of Saha equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.3. Recognition of elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.4. Choice of analytical lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.5. Photon statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

⁎ Corresponding author. Tel.: +39 050 315 2223.

E-mail address: tognoni@[Link] (E. Tognoni).

0584-8547/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2009.11.006
2 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

1. Introduction survey of earlier applications of the CF-LIBS method is provided in

Table 1. Most of the CF-LIBS analyses were carried out on spectra
The Laser-Induced Breakdown Spectroscopy (LIBS) technique is, in acquired in atmospheric air. Only in a few cases different buffer gases
brief, a mix of great potential and severe limitations. It offers a strong and pressure values were adopted, mainly with the purpose of
potential for analysis in situ and in real time, not requiring a separate simulating real conditions expected for on site measurements. Thus,
sampling step. Since it does not need for sample preparation, it avoids Colao et al. [32] and Sallé et al. [33] simulated the Martian atmosphere
wet chemistry. This means quickness and versatility — for qualitative by using 9 mbar CO2. By the way, Sallé et al. [33] tested CF-LIBS
analysis [1–3]. The situation is different for quantitative analysis. quantitative analysis on spectra that were acquired in a stand-off
Versatility means very different measurement environments and very configuration at intermediate distance (3 m). In Herrera et al. [19],
different measurement outcomes. Even with a given experimental different pressure conditions (1000 mbar, 100 mbar and 0.1 mbar) of
configuration, the signal does not depend only on the concentration of an air environment were tested to identify the most favorable
the analyte, but also on the composition and aggregation state of the experimental conditions (see Section 2.1).
matrix. For a fine quantitative analysis of minor or trace elements, LIBS As for the excitation source, the majority of the reported CF-LIBS
requires standards, calibration curves, etc; in one word: time. And the analyses relied on nanosecond pulse lasers, while De Giacomo et al.
quickness is lost. Moreover, matrix-matched standards are simply [27] compared the results of standard-less analysis carried out on the
unavailable in many practical situations. spectra of copper alloy standard samples irradiated with nanosecond
Let's look into the matrix effect in more detail. It involves the and femtosecond pulses (see Section 2.1). The fundamental Nd:YAG
ablation mechanism, the production of free electrons, the absorption wavelength at 1064 nm was generally preferred, but several groups
of the laser pulse by the plasma and the resulting plasma temperature. used visible or ultraviolet radiation. The laser pulse energy varied
Several attempts to correct the calibration curves for matrix effects within two orders of magnitude, from about 1 mJ to more than
have been reported [4–6]. The Calibration-Free Laser-Induced 100 mJ.
Breakdown Spectroscopy (CF-LIBS) approach [7] was developed CF-LIBS has been applied so far to the analysis of spectra covering
following an alternative view; instead of looking at the matrix as an the near UV–visible–near IR spectral range, though in principle lines
external disturbing interference, it takes the matrix into account as a in the vacuum UV range are also eligible for the measurement. It is
part of the analytical problem: it analyzes the matrix together with clear that using an echelle-like instrument as the spectral resolving
the analyte. Ten years after its first presentation, CF-LIBS applications system allows saving measurement time, because a broadband
have been reported in the literature by several research groups. spectrum is acquired in a single run. However, CF-LIBS analysis of
Satisfactory results have been obtained; in other cases the accuracy wide range spectra collected by multiple acquisitions of successive
was poor. The reported CF-LIBS results have been validated, case by spectral windows with a grating monochromator has been also
case, using reference samples or by comparison with the results of documented with very good results [19]. The crucial points are: the
more established techniques. From a general perspective, the effort accurate characterization of the instrument spectral efficiency over
devoted to characterize the method is scattered in time and in geo- the whole wavelength range of interest and the stability of plasma
graphical distribution. A much more systematic approach would be (and instrumental) conditions during the measurement.
needed to promote the CF-LIBS maturation. In general, time resolved acquisitions have been performed to
Since we are convinced that the CF-LIBS may be a valuable tool for select plasma conditions where local thermal equilibrium (LTE) is
many analytical problems, especially in the rapid screening of fulfilled to a good approximation (see Section 4.2) and high signal-to-
unknown samples, we compiled this review according to a critical noise ratios can be achieved. The use of a compact detector was also
criterion. In fact, we believe that guiding the reader across difficulties reported, without details about the temporal settings of acquisition
and possible solutions is more helpful than just taking care of [36]. In Sun and Yu [18], the limitation due to the fixed integration
completeness of the literature overview. This paper is focused on the gate width (1 ms) of an Ocean Optics spectrometer was overcome by
discussion of the CF-LIBS method and does not take into account calculating the difference of two spectra collected at different delay
in detail other standard-less approaches for quantitative spectral values.
analysis which can be found in the literature regarding laser-induced
plasmas [8–11] or meteor spectra [12–14]. A summary of the results 2.1. Definition of a metric
achieved so far by the research groups who tested the CF-LIBS ap-
proach is provided in Section 2. Then, the theoretical basis of the CF- Most of the efforts devoted to the application of the CF-LIBS
LIBS approach in its most ideal formulation is briefly recalled in method were motivated by the difficulty of using the conventional
Section 3. Departures of the typical LIBS plasmas from the model of an calibration curve approach, for example for the lack of matrix-
ideal source, and their implications for CF-LIBS quantitative analysis matched standards (like in the analysis of archeological artifacts) or
are discussed in Section 4. Other specific issues, related to measure- for the need of executing measurement on site (like in planetary
ment and data processing, are described in Section 5. exploration). However, the accuracy of the CF-LIBS results regarding
unknown samples is not easily evaluated. In dealing with a single
2. Overview of CF-LIBS applications element in a sample, analysts evaluate the reliability of the mea-
surement by means of consolidated techniques of statistical analysis
Most frequently, the CF-LIBS analysis has been applied to solid [40,41]. Evaluating the overall performance of a method of multi-
materials: aluminum alloys [7,15–19], steel and iron alloys [15,18– element quantitative analysis is less common in the literature, though
20], precious alloys for jewelry [21,22], copper alloys [19,23–28], techniques of multivariate analysis are nowadays diffusing in
archaeological copper artifacts [24,29], glasses [23], pigments on spectroscopic laboratories [42]. The availability of a simple metric to
roman frescoes [30] and on parchments [31], soils and rocks [32–34], assess the global accuracy of CF-LIBS results would be therefore ex-
meteorites [35], coral skeletons [36], human hair [37]. tremely useful.
The composition of the ambient air was also estimated by CF-LIBS In a previous work [19] a classification of the CF-LIBS approach as a
[7], as well as the plasma plume composition during the irradiation of quantitative, semi-quantitative or qualitative analytical method was
an aluminum target in ambient air [38]. To our knowledge, the first attempted. The measurements reported in the literature were
attempt of CF analysis of an aqueous solution was reported by Tognoni reviewed and the elements contained in a sample were grouped in
et al. [39], in a work where spectral analysis based on CF-LIBS algo- several classes (matrix element, other major elements, minor ele-
rithms was applied to the experimental data obtained by ICP-AES. A ments, trace elements). It was noted that the best accuracy is usually
Table 1
Application of the CF-LIBS method to the quantitative analysis of different kinds of solid samples. The following settings are reported when available: excitation source (wavelength, pulse energy, irradiance, pulse width); spectrometer and
detector (spectral range, resolving power); timing (delay of acquisition after laser pulse, width of acquisition window); measurement statistics (number of cleaning shots, number of measurement shots, repetitions); source of the trusted
compositional values of the sample under investigation; reference.

Target Buffer gas Excitation source Spectrometer and detector Timing Measurement statistics Nominal values References

Aluminum alloys Air Nd:YAG 1064 nm Grating spectrometer and OMA, Delay 2000 ns Gate 1000 ns Reference sample [7]
spectral range 200–800 nm
Air Nd:YAG 1064 nm (10 GWcm− 2) Echelle spectrograph and ICCD, Delay 2000 ns Gate 1000 ns 40 shots Reference sample [15]
spectral range 230–900 nm, λ/Δλ ∼ 5000
Air Nd:YAG 1064 nm, 1.5 mJ Grating spectrometer and ICCD, Delay 500 ns Gate 300 ns 200 cleaning shots + 50 Reference sample [16]
spectral range 200–800 nm measurement shots
0.1 mbar air Nd:YAG 1064 nm, 90 mJ (4.4 GW cm− 2) Grating spectrometer and ICCD, Delay 50 ns Gate 100 ns Reference samples [17]
spectral range 220–700 nm
Air Nd:YAG 1064 nm, 200 mJ Ocean Optics LIBS2500-7, spectral Delay 2500 ns Gate 2500 ns 50 cleaning shots + 100 Reference samples [18]
range 200–980 nm measurement shots in 5 positions
1000, 100 and Nd:YAG 1064 nm, 90 mJ (4.4 GW cm− 2) Grating spectrometer and ICCD, Depending on pressure 10 shots × 3 repetitions Reference samples [19]
0.1 mbar air spectral range 220–700 nm
Steel, iron-based alloys Air Nd:YAG 1064 nm, (10 GW cm− 2) Echelle spectrograph and ICCD, spectral Delay 5000 ns Gate 1000 ns 40 shots Reference sample [15]

E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

range 230–900 nm, λ/Δλ ∼ 5000
Air Nd:YAG 1064 nm, 130 mJ Ocean Optics LIBS2500-7, spectral Delay 830 ns Gate 3340 ns 50 cleaning shots + 100 Reference samples [18]
range 200–980 nm measurement shots in 5 positions
Air Nd:YAG 1064 nm Echelle spectrograph and ICCD, spectral Delay 6000 ns Gate 2000 ns 50 shots Certified samples [20]
range 200–1100 nm, λ/Δλ ∼7500
Precious alloys Air Nd:YAG 355 nm, 150 mJ Echelle spectrograph and ICCD, spectral Delay 10000 ns gate 60 shots Reference samples [21]
range 200–900 nm, λ/Δλ ∼ 7500 5000 ns
Air Ruby laser 694 nm, pulse width 25 ns spectrometer and CCD Delay about 2000–3000 ns Reference samples [22]
Bronzes and co pper Air Nd:YAG 1064 nm, 90 mJ (4.4 GW cm− 2) Grating spectrometer and ICCD, Delay 500 ns Gate 2000 ns 10 shots × 3 repetitions Reference samples [19]
alloys spectral range 220–700 nm
Air Ruby laser 694 nm, (100 GW cm− 2), pulse Spectrometer and CCD, spectral Delay ≥ 2000 ns 5 repetitions Certified samples [23]
width 25 ns range 280–770 nm
Air Nd:YAG 355 nm, 20 mJ Grating spectrometer and ICCD, Delay 1500 ns gate 1000 ns One 266 nm shot for cleaning Certified samples [24]
spectral range 180–850 nm
Air Nd:YAG 1064 nm Echelle spectrograph and ICCD, Delay 1500 ns Gate 2000 ns 20 shots Reference samples [25]
spectral range 200–900 nm, λ/Δλ ∼ 5000
Air Nd:YAG 1064 nm, 200 mJ, single and double Imaging spectrometer and ICCD Delay 1500 ns gate 2000 ns 30 shots with rotating sample Certified samples [26]
pulse with 45 µs delay
Air Nd:YAG 532 nm, 280 mJ, pulse width 7 ns Grating spectrometer and ICCD Delay 800 ns gate 5000 ns 30 shots Standard samples [27]
Air Nd:glass 527 nm, 2.5 mJ, pulse width 250 fs Grating spectrometer and ICCD Delay 50 ns gate 500 ns 50 shots with rotating sample Standard samples [27]
Air Nd:YAG 1064 nm, 100 mJ (5 GW cm− 2) Grating spectrometer and ICCD Delay 5000 ns Gate 1000 ns 100 shots with rotating sample Reference sample [28]
Air Nd:YAG 1064 nm, 20 mJ Echelle spectrograph and ICCD, [29]
spectral range 200–900 nm, λ/Δλ ∼ 5000
Glasses Air Ruby laser 694 nm, pulse width 25 ns Spectrometer and CCD Certified samples [22]
Pigments Air Nd:YAG 1064 nm Echelle spectrograph and ICCD, Delay 2000 ns gate 1000 ns 1 shot per spectrum [30]
spectral range 200–900 nm, λ/Δλ ∼ 5000
Air Nd:YAG 1064 nm Echelle spectrograph and ICCD, Delay 2000 ns gate 1000 ns 1 shot per spectrum [31]
spectral range 200–900 nm, λ/Δλ ∼ 5000
Soils and rocks 6–9 Torr of 95% Nd:YAG 355 nm Grating spectrometer and ICCD, Delay 500 ns gate 1000 ns 20 shots with rotating sample Certified samples [32]
CO2 spectral range 230–700 nm
−2
9 mbar CO2 Nd:YAG 1064 nm, 25 mJ (2 GWcm ) Echelle spectrograph and ICCD, Delay 200 ns gate 500 ns 400 shots × 50 repetitions ICP-AES, ICP-MS [33]
spectral range 280–1000 nm, λ/Δλ = 2600 measurements
−2
Air Nd:YAG 532 nm, 22 mJ (2.2 GW cm ) Grating spectrometer and PMT Delay 600 ns rotating sample ICP measurements [34]
Air Nd:YAG 1064 nm, 90 mJ (4.4 GW cm− 2) Grating spectrometer and ICCD, Delay 500 ns gate 2000 ns 10 shots × 3 repetitions Reference samples [19]
spectral range 220–700 nm
Meteorites Air Nd:YAG 532 nm (5 J cm− 2) Grating spectrometer and ICCD Delay 830 ns gate 5000 ns Reference sample [35]
Coral skeletons Air Nd:YAG 532 nm, 45 mJ (3x1012 Wcm− 2) Ocean Optics LIBS2000-7, spectral Delay 1500 ns 10 shots × 5 repetitions ICP-MS measurements [36]
range 200–980 nm
Human hair Air Nd:YAG 1064 nm, 150 mJ (100 GWcm− 2) Echelle spectrograph and ICCD, spectral Delay 1000 ns gate 2000 ns 20 shots scanning hair length ICP-MS measurements [37]
range 200–900 nm, λ/Δλ ∼ 5000

3
4 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

obtained for the matrix element, while for the remainder of the Different metrics could also be chosen (for example Euclidean
components in the sample the analysis can only be considered semi- distance), of course with correspondingly different features. The
quantitative. Observations of this trend were also reported by other widespread usage of the Manhattan metric is due to its simplicity.
authors [15,23,32]. In particular Burakov et al. [23] explained that a Even though this choice may (and we hope it will) be subjected to
homogeneous distribution of the minor and trace elements in the discussion, we deemed it useful to give a start to an overall evaluation
sample is unlikely. Moreover, it can be expected that noise, spectral of the CF-LIBS performance.
interferences and time settings are especially critical for the In the following we will consider a low distance value as an index
quantification of minor elements [23]. The simulations shown in ref. of good accuracy.
[15] revealed that an acceptable uncertainty of the order of 1% af-
fecting the major component corresponds to very large uncertainty 2.2. Accuracy of CF-LIBS results
values on the minor components. This is a consequence of the closure
equation (see below Section 3.1). The case of a Fe–Mn binary alloy, We display in Table 2 the distance values calculated for those CF-
shown in Fig. 5 in ref. [15], provides a clear example; the data in the LIBS results which have been reported in the literature together with
figure are the statistical representation of the effect of random trusted data. In fact, some groups tested the CF-LIBS performance by
intensity fluctuations and therefore relate only to the measurement analyzing certified samples, while in other cases the CF-LIBS results
precision, however they correctly illustrate the point. The estimated were compared with the results of more established analytical
concentration for Fe is 76.3 ± 0.7, i.e. the relative uncertainty on the Fe techniques such as ICP or EDX (see Table 1). The reader is directed
concentration is 1%. Correspondingly, the estimated concentration of to the research papers cited to compare reference and measured
Mn is 23.5 ± 0.7, i.e. the relative uncertainty on the Mn concentration concentrations of the individual elements. We also refer the reader to
is 3%. The conversion of a flat absolute error to large relative errors is the original papers for the list of the spectral lines used in the
especially severe for the determination of components with lower calculation. In Table 2 the global performance of the analytical method
concentrations (see the example given in ref.[15] for an aluminum is sketched. Results are grouped by classes of materials. The second
alloy); the problem applies to materials of any degree of complexity. column indicates the number of samples (with similar matrix)
Here, we want to propose a measure of the overall accuracy of the investigated in the same paper; the third column lists the number
CF-LIBS results. Given the list of elements in the sample, the nominal of elements included in the quantitative analysis (N in Eq. (1)). Ranges
composition and the CF-LIBS results can be both represented as are reported when samples with different numbers of components
vectors of percent concentration values. For example, for a copper were investigated. Several particular cases are outlined in the table.
alloy composed of Cu, Pb and Sn, the nominal composition may be The analysis is considered incomplete (inc) when only part of the
given as the vector C = (79, 11, 10), which means that the con- sample composition is reported in the paper; in these cases the
centration of Cu is 79 %, the concentration of Pb is 11% and that of Sn is distance is calculated only on the available list of elements. The
10%. Let us assume that the results of the CF-LIBS measurement are analysis is defined as reduced (r) when some of the components are
M = (75, 14, 11). The simplest algorithm for the calculation of a purposely excluded from the calculation and the closure condition is
distance between vectors, which is commonly used in computational applied to a reduced set of elements (see Section 3.1 for a recall of the
mathematics, is the Manhattan or city block distance: algorithm). Also in this case the distance is of course calculated on the
reduced set of elements. The distance values are reported in the fourth
N column (ranges are provided for more than one sample investigated
dist = ∑ abs jMi −Ci j ð1Þ in the same work). Finally, the use of variants (v) of the original CF-
i=1
LIBS formulation is indicated. Note that the distance values calculated
for refs. [20] and [18] correspond to the proposed SAC (self-absorption
where the index i runs over all the N elements of the vectors. Vectors corrected) and IRSAC (internal reference self-absorption corrected)
C and M must have the same size N, and element concentrations must methods, respectively. Results listed in the same papers as obtained
be sorted in the same order. We also assume that elements are sorted by “classical” CF-LIBS were very far from nominal values because
in a decreasing concentration order. With the example data listed strongly self-absorbed lines were included in the calculation instead
above, the distance of CF-LIBS results from nominal values is dist = 8.
As the concentration values are given in percent, the calculated
distance is a measure of the global departure from nominal com-
Table 2
position in percent units. Even though the above definition is quite
Departure of CF-LIBS results from the nominal concentration values (distance)
empirical, it is a monitor of the global quality of the CF-LIBS results. A calculated according to Eq. (1). Ranges refer to samples with similar matrix investigated
distance of a few units is a reasonably good result, while a distance of in the same work.
the order of tens can be considered as a poor result. The distance
Target n. samples n. elements Distance References
defined in Eq. (1) tends to depend strongly on the accuracy of mea-
Aluminum alloys 1 7 0.9 [7]
surement of the most abundant elements in the sample. For example,
1 5 1.8 [15]
an error of 5% on the matrix element (constituting let's say the 95% of 1 3 5.4 (2.1 v = CCF) [16]
the sample) gives a larger contribution to the distance than over- 1 3 (r) 1.6 (v = IRSAC) [18]
estimating by 100% a minor element (with concentration of about 1% Steel, iron-based 1 2 2.8 [15]
in the sample). A different alloys 2 2–3 (r) 3.1–5.4 (v = IRSAC) [18]
 choice  of definition would be possible by
3 4–5 2.5–5.3 (v = SAC) [20]
using the relative error MiC−C i
as the addendum in the sum in Eq. (1).
i Precious alloys 4 3–4 0.9–2.0 [21]
This alternative definition would give higher weight to the minor Bronzes and 10 2 0.6–1.8 [19]
elements in the sample. An example of the use of this second distance copper alloys 3 6–7 10.4–14.2 [23]
is provided in ref. [28]. In the current systematic evaluation of the CF- 1 4 5.1 [25]
Glasses 1 5 13.2 [23]
LIBS performance we preferred the definition stated in Eq. (1) for two
Soils and rocks 3 11–13 (inc) 5.1–6.1 [32]
main reasons: most of the works reviewed here measured sample 6 10–11 15–73 [33]
composition including elements with concentration above about 1 7 (inc) 3.6 (1.5 v = CC) [34]
1%; moreover, as illustrated in ref. [15], the accuracy on the most Meteorites 3 3–9 1–11 [35]
abundant element set a limit on the accuracy achievable for the minor Coral skeletons 1 6 (inc) 12.6 [36]
Aqueous solutions 3 8 (r) 3.8–10.0 [38]
elements.
 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14 5

of being discarded from dataset. More details on the variant methods maximum at about 2 µs delay and then decrease again. The relative
are given below (see Section 3.3). proportion (in number and intensity) of ionic and neutral lines is
It can be easily recognized that metallic alloys tend to give more rapidly changing during plasma evolution [19]: at the beginning, ionic
accurate results than organic samples and rocks. Among soils and lines dominate the emission spectrum, while later neutral lines
rocks, it is also evident that CF-LIBS results on natural samples [33] are become predominant. If the trend of distance with evolution time is to
more distant from nominal values than the results obtained for be considered significant, it would suggest that the analysis is most
pressed powders of certified soil samples [32]. A few cases do not accurate when one species (either ionic or neutral) is dominating the
represent a reliable statistical sample; however this is a hint that spectrum and not when their abundance is more balanced. It is
sample inhomogeneities may set a limit to the accuracy achievable. possible that this observation is specific to the experimental
We discuss other possible interpretations of this finding in Section 4.2, configuration and not general; however an explanation in terms of
regarding LTE. The general trend of distance values is contradicted by plasma inhomogeneity is possible (see discussion in Section 5.2). In
the results obtained by Wang et al. [34], which are notably good for a general, both early and late times are expected to be critical for the
soil sample. However, in this case, the list of the elements included in validity of LTE: at early times plasma evolution is more rapid, at late
the calculation is incomplete. times self-absorption tends to increase and the electron density is too
In a few cases systematic investigations have been carried out to low to satisfy the McWhirter criterion. It is possible that in the specific
explore the dependence of CF-LIBS results on several aspects of the case plasma inhomogeneity is the dominating effect.
experimental configuration. Herrera et al. [19] explored three dif- De Giacomo et al. [27] applied the CF-LIBS approach (with
ferent values of ambient gas pressure (1000, 100 and 0.1 mbar) for the variations, see below Section 3.3) to the analysis of spectra obtained
analysis of aluminum alloys. The test involved six aluminum alloy by irradiating three copper alloys with ns and fs laser pulses. In that
standards. The distance values calculated from the results reported in work, the authors limited the scope of the analysis to the minor alloy
the paper are displayed in Table 3. The elements included in the components: Ni, Sn, Pb. The distances calculated from the results
calculation were 10 for alloys S4 and SM10, and 9 for the other alloys. obtained with ns and fs irradiation are compared in Table 5 and
The acquisition time settings varied depending on the pressure (delay appear to be very close to each other. This is a first indication of the
and gate were 2 µs and 0.5 µs for 1000 and 100 mbar and 50 ns and suitability of femtosecond plasmas for standard-less quantitative
100 ns for 0.1 mbar, respectively). The temperature of the plasma was analysis.
quite similar in the three pressure cases, but the electron density at The above examples of systematic studies are very useful toward a
0.1 mbar was an order of magnitude lower than at atmospheric thorough characterization of the CF-LIBS approach. It is desirable that
pressure. Notwithstanding, the range of distances was similar in the further efforts will be addressed in this direction, extending the
three cases, though problems in the fulfillment of LTE would be investigation to as many variables of the measurement as possible.
expected at low pressure, due to the low electron number density (see
below, Section 4.2). 3. The CF-LIBS approach
In the same paper [19] a systematic study of the CF-LIBS
performance with varying acquisition delays after the laser pulse The algorithm used in the CF-LIBS procedure has been described in
and fixed 0.5 µs gate width has been reported. Two aluminum alloys detail elsewhere [7,15]. Here we review the main idea of this ap-
(B8 and SM10) were tested in this case, in a reduced air pressure of proach and the underlying hypotheses which are necessary to design
100 mbar. The distance values calculated from the results reported in a self-consistent algorithm.
the paper are listed in Table 4. For both samples, the distance was low The philosophy behind the approach rose from the observation
at a very short delay of 0.5 µs and then it tended to increase, to reach a that plasma emission in LIBS is very sensitive to changes in the matrix
of the sample. Turning the perspective up side down, changes in the
matrix can be measured through the analysis of a LIBS spectrum. In
Table 3 other words, Ciucci et al. [7] hypothesized that the spectrum contains
Departure of CF-LIBS results from the nominal concentration values (distance) of six all what is needed to derive the composition of the matrix. However,
aluminum alloy standards, depending on the ambient pressure during the measure- the laser-induced plasma is a very complex object. Its' realistic
ment. Distance values calculated from data in ref. [19].
description is not attainable with simple mathematical models. To
Target Distance manage the description of plasma macroscopic properties, several
1000 mbar 100 mbar 0.1 mbar simplifying assumptions were adopted. Let's consider them as 0th
order approximations:
B8 4.4 11 3.2
SM10 5.0 6.6 8.4
D33 5.4 4.6 i) the plasma composition is representative of the unperturbed
S4 4.2 2.1 target composition;
V14 6.1 3.2 ii) the plasma is in local thermodynamic equilibrium (LTE) in the
Z8 4.6 5.3
temporal and spatial observation window;
iii) the plasma can be modeled as a spatially homogeneous source;
iv) the spectral lines included in the calculation are optically thin.

Table 4
Evolution of distance of CF-LIBS results with acquisition delay, at a fixed 0.5 µs gate
width, for two aluminum alloys. Distance values calculated from data in ref. [19]. Table 5
Departure of CF-LIBS results from the nominal concentration values (distance) of three
Acquisition delay (µs) Distance copper alloy standards, depending on laser pulse width. Distance values calculated from
Alloy B8 Alloy SM10 data in ref. [27].

0.5 3.5 3.4 Target Distance

1 5.8 7.2
7 ns pulses 250 fs pulses
1.5 11.3 7.3
2 11.0 6.6 B4 3.2 3.5
2.5 7.9 B21 2.2 1.2
3 4.5 4.0 B22 1.2 1.9
6 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

The first assumption implies that laser ablation is stoichiometric and 3.2. Experimental requirements
that the material ablated from the target is also stoichiometrically
atomized (no element segregation occurs in the plasma). There is also a The correct application of the procedure requires the knowledge of
fifth assumption: the spectral range of measurement includes mea- the intensity of spectral lines emitted by all the elements of interest
surable lines from all the elements present in the sample, which is and of the plasma temperature. We recommend measuring the in-
actually an experimental requirement and not a characteristic of the tegral of the line over its wavelength profile, because it is a more
plasma description. The above hypotheses are quite rough approxima- robust quantity than the peak intensity, both if we consider Poisson
tions and in many cases one or more among them are not appropriate to statistical fluctuations and systematic effects, like the line broadening
describe the real plasma conditions. Evidences of the failure of the above dependence on plasma electron density. By inspecting the literature
hypotheses are discussed in Section 4 together with the consequences of we have found different expressions for the spectral line intensity,
releasing the single assumptions. The lack of completeness of the depending on the choice of the measurement units (counts or
spectrum and other experimental issues are discussed in Section 5. energy). We recommend using counts because this is the observable
measured by the detectors typically used in LIBS. In several cases,
3.1. Brief recall of the basic idea energy units are preferred in combination with a spectral efficiency
curve which is calibrated in energy [28].
In LTE, excited levels are populated according to the Boltzmann Plasma temperature can be measured by linear regression of
distribution and ionization states are populated according to the Saha– suitable spectral lines in the Boltzmann plot or in the Saha–Boltzmann
Boltzmann equilibrium equation. The graphical representation of a plot [15]. In the latter case the electron number density must be
group of spectral lines in a Boltzmann plane is commonly used to measured as well. In Saha–Boltzmann plot the uncertainty on the
determine the plasma temperature. Each spectral line is represented as a slope is apparently reduced, thanks to the wider abscissa range, but
point in the plane where the x coordinate corresponds to the energy of concerns have been raised on the reliability of the temperature
the upper level of the transition (x=Ei), and the y coordinate to the obtained by mixing emission from both the core and the periphery of
logarithm of the line intensity Iij divided by the transition probability Aij the plasma [43] (see discussion below, Sections 4.3 and 5.2).
and degeneracy gi (y = log(Iij/Aijgi)). It is intended that the observed line Measurement of the plasma electron density is also necessary for
intensity is corrected for the relative spectral efficiency of the detection CF-LIBS calculations if lines of an individual species are not detected
system. The plasma temperature is obtained by linear regression of the (see Section 5.2).
points representing lines of the same species, namely by taking the The spectral efficiency of the specific equipment used to acquire
reciprocal of the slope. The novelty of CF-LIBS [7] is in the use of the the spectrum must be measured in the spectral range of interest. In
second parameter optimized during the linear regression of the points in fact, differently from the analysis based on calibration curves, in CF-
the Boltzmann plane, i.e. the intercept value, qs, which is a function of LIBS the intensity values of different lines must be compared, even if
the number density (ns) of the individual species (s) in the plasma: originating in spectral regions very far from each other. Since the
  spectral efficiency of the grating, the sensor and other optical
Fns components is not flat in wavelength, the intensity of lines emitted
qs = log ð2Þ
Us ðTÞ at different wavelengths must be corrected before being used in the
CF-LIBS algorithm.
where the Us(T) is the value of the partition function of the species s at The CF-LIBS analysis can be applied in principle to any element in
the temperature T and F is an unknown parameter which accounts for the periodic table. However, relevant spectral information must be
the absolute efficiency of the detection system (the instrumental available regarding the elements and the emission lines to be
response is generally known in relative and not in absolute terms). measured. This includes first ionization energy and atomic mass for
The number density of each species can be thus derived by inversion each element, upper and lower level energy and statistical weight,
of Eq. (2). The total number density of the element of interest is and transition probability for each spectral line to be measured,
obtained by summing over the neutral and the ionization states with partition function value in the range of temperature of interest for
significant abundance in the plasma. Typically, this includes neutral each neutral and firstly ionized species. This kind of information can
(I) and singly ionized (II) species (and therefore: n = nI+nII). Once be collected from on line atomic spectroscopy databases [44,45].
the total number densities have been calculated for all the elements,
the relative abundance ca of element a is obtained as: 3.3. Variants
I II
χ na ðn + na Þ In their measurements on bronze samples, Colao et al. [24] found
ca = = a ð3aÞ
∑ na0 ∑ na0 significantly different values of temperature among the main con-
a0 a0
stituent Cu and the minor elements (Zn, Sn and Pb). They developed a
in terms of molar fractions or as: model where the plasma is represented as the superposition of two
non-interacting plumes described by different temperature values.
m na Ma ðn + na ÞMa
I II
However, the final results were deemed unsatisfactory. The distance
ca = = a ð3bÞ of their results from nominal concentrations has not been calculated
∑ na0 Ma0 ∑ na0 Ma0
a0 a0 in this review because confidence ranges were published in place of
mean values [24].
in mass abundance. In Eqs. (3a) and (3b) the sum is extended over all
The same group also developed a theoretical model of thermal
the detected elements a′, Ma is the atomic mass of element a and the
evaporation [25,46] to simulate the composition of the plume after
unknown scale factor F is canceled in the ratio. Eqs. (3a) and (3b) rest
the irradiation of bronze alloys, where elements with very different
on the hypothesis that the closure condition applies:
thermal properties are present and stoichiometric ablation is ques-
tioned. The authors reported the theoretical composition of the plume
∑ na0 = 100% ð4Þ
a0 which is expected according to their evaporation model. In Table 2,
the distance of the CF-LIBS results is calculated with respect to the
When a reduced set of elements is considered (like in the cases nominal sample concentration (dist = 5.1); the distance of the same
outlined in Table 2 with the “r”), the closure sum is conveniently results from the theoretical plume composition calculated through
adapted. the non-stoichiometric evaporative model is 4.3. However, the
 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14 7

improvement is not as substantial as the one achieved by including the elements in the sample, and second it reduces the length of
ions in the calculation (see Section 5.2). calculation steps thus limiting the corresponding propagation of un-
Pershin and coworkers [16] also developed a method for taking certainty. However, this method often requires that the sample is
into account the selective evaporation of sample elements, occurring manipulated to add the internal standard, thus cancelling the ad-
during the irradiation and precisely in the melting-evaporation stage, vantage of inherent versatility which is offered by the CF-LIBS. There-
before the start of optical breakdown. A work function was defined, on fore, the internal standard normalization can be proposed in cases like
the basis of thermal parameters, for each target component and then the analysis of liquid solutions or powders. It must be also noted that
introduced as a correcting parameter in the expression of line in- the measurement uncertainty affecting the signal of the element
tensity used in CF-LIBS calculations. In consequence of the correction, chosen as the internal standard is directly propagated to the cal-
the modified method is proposed as CCF-LIBS. The effectiveness of the culated concentrations of the unknown elements. It is therefore
correction was demonstrated for an aluminum sample [16] (see also advisable to set the conditions to obtain a measure of the internal
Table 2). standard with high signal-to-noise ratio and high repeatability.
De Giacomo et al. [27] proposed a variant of the CF-LIBS method, Other variants specifically developed to take account of self-
called self-calibrated LIBS (SC-LIBS) with the aim of extending its absorbed lines are described in Section 4.4.
suitability to a broader range of experimental conditions. The
assumption of LTE was relaxed to a less demanding requirement of 4. Plasma non-ideality
partial LTE (see discussion in Section 4.2); neutral species were
assumed to be the only species with significant abundance and a In a previous paper we investigated the effect of some experi-
black-body model of the plasma continuum spectrum was used to mental and theoretical parameters on the precision and accuracy of
achieve an internal normalization of the calculated species densities. CF-LIBS results [15]. The study was carried out by means of a
The use of the background intensity at a given photon energy value numerical simulation, where parameters were varied one at a time.
allows the elimination of the experimental scale factor (F in Eq. (2)) The plasma was represented as a thin, homogeneous radiation source
and in practice substitutes for the closure equation. However this in thermodynamic equilibrium. In particular, we considered the effect
choice introduces a further strong assumption on the parameteriza- of line intensity fluctuations on precision, and the effects of
tion of the background intensity. uncertainty of the electron density, of the partition functions and of
In ref. [28] Aguilera et al. described a variant of the original CF-LIBS the spectral efficiency curve on the accuracy of the resulting con-
quantitative analysis where the Saha–Boltzmann plot was used not centration values. It was noted that the choice of partition function
only to estimate the plasma temperature, but also to derive the values directly affects the resulting concentrations, the electron
relative number densities of elements from the corresponding density only has an indirect effect through the evaluation of
intercept values. Technically, the intercept in the Saha–Boltzmann temperature (when the Saha–Boltzmann plot is used) and the line
plot is a function of the neutral species density and the corresponding intensities and spectral efficiency values have double effect on both
ionic species density is calculated via Saha equilibrium equation. temperature and concentrations [15]. Differently from the above
However, since the intercept is obtained by regression of a larger mentioned paper, where the plasma was considered as an ideal
number of spectral data, accuracy of the result is expected to improve. emitting source, in the following sections we discuss the implications
This method is recommended in the analysis of space resolved spectra of the departure from the assumptions listed in Section 3 and review
or when the plasma inhomogeneity can be neglected. In the same relevant papers available in the literature.
paper the plasma temperature was obtained by multiple-element
data regression in the Saha–Boltzmann plot, as earlier proposed in 4.1. Failure of stoichiometric composition
[47].
In case an internal standard of known concentration is present in Lack of sampling stoichiometry is a severe issue for analytic
the sample, this kind of information can be exploited to convert the techniques based on laser ablation in general (LA-ICP-AES, LA-ICP-MS,
relative proportions of a reduced set of elements in concentration LIBS, etc.) and not only for the CF-LIBS method.
values, even though the global composition of the sample is not In the other techniques laser ablated particles are transported
estimated. This approach was exploited in [39]. In that work the CF- toward the analytical chamber where they are excited and ionized by
LIBS approach was applied to ICP-AES spectral data obtained by an independent power supply. Departure of the analytical sample
measuring spiked solutions with a known concentration of Be. Even from stoichiometric composition of the original material (also called
though the matrix characteristic elements were not measured, the use fractionation) is possible in the stages of ablation, transport and
of an internal standard allowed the determination of the concentra- particles atomization. Studies of fractionation can be found especially
tion of the analytes. in the ICP literature: a complete review on this topic is out of the scope
Burakov et al. [23] mentioned the correction of the CF-LIBS results of this work, notwithstanding some references are provided as a
with respect to the known concentration of one of the components, starting point for the interested reader. In fact, even though transport
although they didn't exploit this possibility in the analysis reported in and independent excitation are not exploited in LIBS, laser ablation
their paper. techniques share many common problems.
Wang et al. [34] also proposed a modified CF-LIBS approach, based It should be noted that, in the approaches based on calibration
on the knowledge of the concentration of an internal standard. curves, the stronger requirement of matrix-matched standards makes
However, in the presentation of their method, the authors were not the fractionation issue less important, because it is assumed that
sufficiently clear in differentiating among element and species in the standard and unknown sample will behave in a similar way. In CF-
plasma. Apparently, the information on the internal standard element LIBS, instead, the results are based only on the analysis of the emission
concentration is used in the calculation of the number density of originated in the ablated plume, so that non-stoichiometric ablation
individual ionization species of the unknown elements. For the sake of or atomization will degrade the final accuracy.
correctness, the distribution of internal standard atoms among the Preferential ablation depends on the laser pulse characteristics
neutral and the ionic states should be first evaluated, because the such as wavelength, width, and irradiance value. Depending on these
normalization must be applied to homogeneous quantities: element parameters different ablation regimes dominate the process [38].
with element (like in [39]) and species with species. In their investigations, Mao et al. [49] found irradiance ranges that
The possibility of using an internal standard offers two main produced stoichiometric Zn/Cu molar ratio in the plasma; how-
advantages: first, it eliminates the need for measuring lines from all ever the stoichiometric regimes depended on the laser pulse width.
8 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

Fractionation is also a possible consequence of crater development plasma evolution, while numerical examples of the amount of ablated
during laser sampling on static targets [49]. Laser-produced aerosol elements entrapped in molecular form in the plasma are, to our
composition was found to depend on particle size, though its average knowledge, not available.
over the particle size distribution reflects the bulk composition [50]. Experimental studies of molecular emission in LIBS plasmas are
The irradiated material and the ambient gas also play a role in non- now appearing in the literature with increasing frequency, due to the
stoichiometric ablation [51–53]. Experimental and theoretical studies rising interest for the analysis of organic or biological samples. In
pointed their attention mainly on pairs of elements with very Baudelet et al. [62] it is shown that molecular emission in fem-
different melting points and vapor pressure values, such as Cu/Zn tosecond plasmas is greater than in plasma obtained by nanosecond
[48,54], Pb/U [49], etc.. Garcia et al. [53] suggested that the elements irradiation. Doucet et al. [63] combined chemometrics with intensity
with lower first ionization energy are preferentially ablated. Frac- of molecular emission in LIBS to measure the composition of phar-
tionation was not observed when elements with nearly the same maceutical formulations. The authors used standards to calibrate their
ionization energies were sampled. The sampling strategy is also method and no information was provided on the relative proportion
relevant: a comparison of surface scanning ablation and fixed- of atoms and molecules in the plasma.
position in-depth ablation was carried out by Gonzalez et al. [55], In the earlier investigations by the Niemax group, the concern for
who evidenced differences in the time-dependent fractionation incomplete atomization in the plasma was raised [54], especially
behavior. when elements with very different vapor pressure values are present.
Several papers reporting CF-LIBS analysis also discussed non- The suggestion given by the authors is to set the measurement time to
stoichiometric ablation and proposed models to correct the results for several microseconds after the laser pulse. As discussed below
selective evaporation. Fornarini et al. [25,46] developed a theoretical (Section 4.2), this is in conflict with the requirement of a sufficiently
model to describe laser ablation in quaternary bronze alloys high electron density, for the sake of fulfillment of thermal
(composed of Cu, Sn, Pb and Zn). The simulations showed that the equilibrium. The better guarantee of complete atomization in the
plasma plume can be substantially enriched in Cu and Zn compared to plasma seems to be the use of a reheating pulse. Surprisingly, at this
the bulk alloy, using either the 1064 nm or the 355 nm laser wave- time no systematic CF-LIBS analysis of double pulse spectra can be
length for ablation. In the model, the authors assumed a pure thermal found in the literature.
vaporization process as the mechanism responsible for ablation; even
though the results of the simulation reproduced quite closely their 4.2. Failure of local thermal equilibrium
experimental data, the initial assumption is not supported by
experimental evidence found by other groups [38]. Selective evapo- The LIBS analysis deals with the radiation emitted by the plasma.
ration of alloy components during the laser ablation was also modeled Therefore, a full equilibrium among the components of the plasma
by Pershin and coworkers for Al alloy [16] and bronze targets [26]. (electrons, atoms, ions and radiation), where all the distributions are
Using the calculated enrichment factors to correct the CF-LIBS results characterized by a unique value of the temperature, is not possible
they obtained more accurate estimations of the alloy composition. It [64]. The full equilibrium in fact would require that the plasma is
should be noted that other authors, in different experimental con- optically thick at all frequencies. The escape of photons from the
ditions, did not observe the effect of selective evaporation in brass plasma automatically implies spatial gradients of thermodynamic
samples [23]. Herrera et al. [19] reported that in 9 brass samples out of parameters. However, if the loss of energy due to escaping photons is
10 investigated, the CF-LIBS concentration of Zn was slightly over- negligible in respect of the total plasma energy balance, a local
estimated (by about 10%). thermodynamic equilibrium is still possible, where the distribution of
All the researchers that reported CF-LIBS results so far included in electrons, atoms and ions is governed by Maxwell and Saha–
the calculation of total heavy particles concentration only neutral Boltzmann equations [64,65]. Several investigations of the existence
atoms and singly ionized ions. The possibility to include also double of LTE in laser-induced plasmas can be found in the literature [57,66].
charged ions if needed is sometimes mentioned, but their abundance The assumption of LTE is particularly important in CF-LIBS because
is generally considered negligible and not estimated explicitly. In it allows deriving, from the population of the upper level of the
Bleiner et al. [51] the number density of singly and doubly charged Zn observed transitions, the global distribution of population in all the
ions is calculated at early times after the laser irradiation and after electronic levels of the species of the element considered. If LTE is not
100 ns the abundance of Zn III across the plume is lower than that of verified in the plasma, the CF-LIBS calculation will produce unreliable
Zn II. results. The development of a direct, quantitative check to verify the
It could probably be important to estimate the abundance of validity of LTE in the actual plasma would be extremely useful to the
molecules formed in the plasma, especially when measurements are analyst. The McWhirter criterion is widely used in LIBS analysis. It is
carried out in the late evolution times [56]. Plenty of studies of the based on the requirement that in LTE the rate of collision processes
chemical reactions between the material ablated by the laser and the must be dominant over the radiative processes. This criterion is more
surrounding gas can be found in the field of pulsed-laser deposition easily matched in the early stages of plasma evolution, when the
(PLD) [57], while in the more specific LIBS field they are less frequent electron number density is of the order of 1017–1018 cm− 3. Often,
[58,59]. Typically, LIBS measurements are carried out at atmospheric during late stages of plasma evolution the electron number density
pressure, which hinders the simple transfer of knowledge gathered drops to values lower than the McWhirter threshold [19]. However, it
for PLD. must be recalled that this criterion was derived for homogeneous and
The simulation of dynamic and kinetic evolution of a double pulse stationary plasmas. The condition required by the McWhirter
plasma under water was described in [60], where the spatial profile of criterion for the validity of LTE is therefore necessary but not
the molar concentration of the diatomic molecule TiO was calculated sufficient in laser-induced plasmas, which evolve in time and are
with a result at least three orders of magnitude lower than the atomic inhomogeneous. A second condition requires that the rate of variation
species concentration. In a subsequent paper De Giacomo et al. [61] of thermodynamic parameters (plasma temperature and electron
compared the molar concentration of the CuO molecule in the plasma density) is small compared to the time scale of the establishment of
obtained by irradiating a bronze sample submerged under water or in excitation and ionization equilibrium. This condition is more easily
air, with the result of higher abundance of the molecule for irradiation met in a late stage of plasma evolution, after the expansion is almost
under water. This kind of study is not extensive enough to cover a stopped, while it is unlikely fulfilled at the beginning of the expansion.
wide range of experimental cases. Moreover, the chemical kinetics For the same reasons, LTE is more difficult at a low ambient pressure
studies are typically presented as a tool to better characterize the rather than at atmospheric pressure, because in the former case the
 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14 9

plasma evolution is more rapid. Finally, a third condition is set to the in 11 regions along the plasma propagation axis, z. The spectra
length scale of thermodynamic parameters, which must be larger than obtained were thus spatially resolved along the plasma axis but
the diffusion length of the atoms in the time scale of equilibrium. A integrated along the line of sight. We calculated the distance for the
detailed discussion of the fulfillment of these three criteria in laser- CF-LIBS results obtained from spectra at different values of z, as
induced plasma is given in Capitelli et al. [67]. reported in Table 6. The trend displayed by the distance values reveals
In a recent review of the topic, Cristoforetti et al. [68] offer several that the most accurate CF-LIBS results are obtained from the emission
practical examples extracted from real spectra. The dependence of the of the central region of the plume, closer to the propagation front. This
threshold value imposed by the McWhirter criterion on the elements trend was explained by the authors with the relatively higher
that compose the plasma, via the parameter ΔE, is discussed in the temperatures and electron number densities observed in these
paper. Plasmas composed of metallic elements satisfy the McWhirter regions of the plasma, which may lead to conditions closer to local
criterion at low to moderate electron density values, while non- thermodynamic equilibrium.
metallic elements require higher electron density values. This is Aguilera et al. [28] applied the CF-LIBS analysis to radially resolved
related to the electronic configuration and in particular to the higher spectra extracted from measurements carried out in ambient
density of excited levels in the low energy range in metals. A atmosphere on a copper based alloy. The radial profile of distance
difference is also found among metallic and non-metallic elements was evaluated in the paper, according to a Euclidean definition. Here
regarding the time scale for establishment of equilibrium, with the we calculated the distance using the definition of Eq. (1), for the sake
latter requiring longer times. These considerations offer an interesting of consistence to the other cases, as reported in Table 7. It is worth
explanation of the experimental observation that better quality CF- noting that the trend is qualitatively the same for the two definitions
LIBS results are obtained in metals than in non-metallic samples (see of distance: the most accurate results correspond to the spectra of the
Table 2). Moreover, in ref. [68] it is illustrated that minor and trace central regions of the plasma. Also in this case it was pointed out that
elements may be subjected to strong problems of spatial gradients of in the external regions the electron density declines to values unable
concentration in the plasma, making their quantification less reliable. to fulfill the McWhirter criterion thus contradicting the hypothesis of
In De Giacomo et al. [27], the self calibrated analysis, variant of CF- existence of local thermodynamic equilibrium. It is also important to
LIBS, was based on the assumption that partial LTE is present in the recall that in this case only a subset of elements (not including the
plasma. In that work plasmas were generated by using nanosecond major one) were included in the calculation.
and femtosecond laser sources. Considering the different ablation By inspection of Tables 6 and 7 the comparison of spatially re-
mechanisms and plasma evolution patterns at play in the two cases, solved and spatially integrated spectra provides contradictory results.
the authors deemed it useful to relax the more general equilibrium In the measurements described by Aguilera et al. [28], the CF-LIBS
hypothesis to the requirement of partial LTE (pLTE). They selected a analysis of space integrated spectrum gives a poor result, while in the
number of analytical lines with excitation energy in the range 30.000– measurement by Herrera et al. [17], apparently the errors compensate
50.000 cm− 1 and succeeded in obtaining quantitative results for the each other and the distance of the space integrated spectrum is in line
investigated elements. In our view, the use of spectral transitions from
the high excitation energy range may surely be helpful in avoiding
self-absorption bias effects, so to obtain more reliable estimations of
Table 6
the plasma temperature; however, this choice doesn't solve the Profile of distance of CF-LIBS results along the plasma propagation axis for an aluminum
problem of departures from LTE, if present. In fact, even thought an alloy. Slabs are numbered from target surface toward the plasma front. Distance values
accurate estimation of the plasma temperature is obtained, CF-LIBS calculated from data in ref. [17].
and related approaches resort to an extrapolation of the total particle Plasma slab Distance
density of a given species from the measured Boltzmann distribution
1
over the higher energy excited levels. If the population of the ground 2
state is not in equilibrium with the higher states, the result of our 3 13.2
extrapolation will not reflect the total species density. This problem is 4 7
intrinsic to the CF-LIBS approach and may appear insoluble. Since 5 6.8
6 7.4
quite accurate results have been achieved by many different groups, it
7 5.3
is possible that deviations compensate each other to some extent. In 8 3.4
fact, if the plasma is not in LTE, ground states will be all overpopulated 9 4.8
or, alternatively, underpopulated, so that the bias will affect all species 10 3
in the same direction and in the final closure equation it will be partly 11 6.8
Spatially integrated spectrum 4.5
canceled.

4.3. Failure of spatial homogeneity

Table 7
Plasmas generated by means of a laser pulse are transient and Radial profile of distance of CF-LIBS results for a copper based alloy. Distance values
inhomogeneous in nature. Characterization of plasma features have calculated from data in ref. [28], corresponding to a reduced composition (3 elements
been reported in the literature [69–79]. Several researchers proposed accounting for the 7% of the total mass composition- remainder Cu 93%).

models to take into account the plasma inhomogeneity and its Radial position (mm) Distance
consequences on the spectral line intensity [57,80,81]. A recent work 0.03 6.7
quantified the degree of inhomogeneity of typical LIBS plasmas and 0.2 6.9
outlined the difference between the local values of plasma temper- 0.43 2.2
ature and the apparent value obtained from the analysis of the space 0.66 8.2
0.88 11.8
integrated radiation [43].
1.11 16.2
The CF-LIBS approach was tested in the analysis of space resolved 1.34 18.6
spectra in the paper by Herrera et al. [17], which regarded an 1.57 19.2
aluminum alloy sample irradiated in a reduced atmospheric pressure 1.8 19.89
of 0.1 mbar. Plasma emission was recorded with an imaging 2.02 11.9
Spatially integrated spectrum 25.2
spectrometer coupled to a CCD and the resulting image was sectioned
10 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

with the resolved ones. It is worth pointing out that the measure- the theoretical expected intensity for each line λ according to the
ments were carried out in different pressure conditions and the relation:
profiles span different axis: the plasma propagation axis in ref. [17],
the plasma radius in ref. [28].
ij Iλmn Aij gi Em −Ei = kB T
Iλ = R
e ð7Þ
4.4. Failure of optical thinness Amn gm

In real plasmas self-absorption of spectral lines is very common. In

a quantitative procedure like CF-LIBS, the use of unrecognized self- The authors propose to iteratively evaluate the temperature from
absorbed lines is especially detrimental. Self-absorption affects most the spectral intensities corrected according to Eq. (7), up to
severely transitions with a low energy upper level, whose population convergence. They recognize that the method is useful when the
appears lower than real. Including self-absorbed lines in the lines of one species are all or for a large number substantially thin.
Boltzmann plot leads to overestimation of the plasma temperature When the lines of a species are all or almost all self-absorbed (which is
and underestimation of the intercept (and therefore of the concen- sometimes the case for the most abundant elements), the iterative
tration) of the most abundant species, with consequences over the process does not guarantee that the final temperature value is
evaluation of the global sample composition. It is particularly accurate. In this case, the authors suggest using in the iteration the
important, therefore, to recognize self-absorbed lines and then to average temperature value obtained from the other species. This
decide how to deal with them. The first step is not trivial: the simple simplified approach yields Boltzmann plots which look artificially
inspection of the line profile by naked-eye may be misleading. Several well aligned. However, their information content is limited to the
research groups have devoted their efforts to model self-absorption intensity of the reference line and to the lowest temperature value,
effects on spectral line intensity in LIBS [80–86]. calculated from a species not subjected to self-absorption. Moreover,
An automated algorithm (SAC, self-absorption correction) for the authors realized that the choice of the reference line is very
quantification of self-absorption of the lines used in the CF-LIBS critical, and they use, in place of its experimentally measured intensity
analysis and correction of their intensity was proposed in [20]. In (which is subjected to statistical fluctuations and, possibly, unrecog-
particular, the self-absorption parameter SA was defined as the nized systematic bias), the value obtained by the linear regression in
departure of the measured line peak intensity I(λ0) from the linear the Boltzmann plot.
extrapolation of the curve of growth in the optically thin regime to the Moon et al. [87] measured lines self-absorption by comparing
same emitters number density of the actual measurement, I0(λ0): spectra acquired with and without a duplicating spherical mirror
placed behind the plasma plume. They demonstrated that this method
−kðλ Þℓ
Iðλ0 Þ ð1−e 0
Þ is especially effective to identify outliers in the Saha–Boltzmann plot
SA≡ = ð5Þ
I0 ðλ0 Þ kðλ0 Þℓ for temperature evaluation. The authors also recalled that the data
obtained are the result of line-of-sight measurements, and therefore
with can only be interpreted in terms of an average over the plume profile
[87], a point which applies to all the above mentioned approaches.
 
λ40 n n g ℓ The effects of self-absorption were found out to grow with time
κðλ0 Þℓ = Aki gk i 1− k i ð6Þ
8πc gi gk ni Δλ0 during plasma evolution [86,87].
The current overview suggests that the different methods
where κ(λ0) is the absorption coefficient at the line central proposed should be exploited to discriminate optically thin from
wavelength, which takes into account both the absorption by lower optically thick lines, and only thin lines should then be used for CF-
level atoms and the stimulated emission of the upper level atoms LIBS calculation. In fact, the theoretical and empirical corrections are
under the effect of the radiation; ℓ is the optical path length inside the both based on a simplified assumption of homogeneous plasma,
plasma plume; c is the speed of light, λ0 is the central wavelength of which, as discussed above, is unrealistic. Moreover, the degree of self-
the transition, Δλ0 is the FWHM of the line and the other symbols absorption can be for some spectral lines too large for any correction
were already defined. All the parameters can be retrieved from to be applied. This is the case for resonance lines of elements of the
spectral database or from the measured spectrum, with the exception first and second group in the periodic table, even for concentration
of the number density of each species, which in CF-LIBS is obtained in values of several hundreds of ppm. In these cases the detection system
arbitrary units. The absolute number density values are therefore should be sensitive enough to detect not-resonant lines. Burakov et al.
derived by imposing the condition of plasma neutrality and equating [23] reported that the resonance lines of alkali elements were strongly
the total ionic density (scaled by the unknown factor F) to the known self-absorbed and the others were too weak to be measured, thus
electron density value. Finally, the extent of self-absorption affecting degrading the performance of CF-LIBS analysis of a glass sample.
each individual line is calculated, and the observed integral intensity
can be appropriately corrected in the Boltzmann plot (the effect on 5. Completeness and other experimental issues
the integral intensity scales as SA0.46 [86]). The correction to the
intensity of the self absorbed lines leads to a more accurate estimation In practical cases it is not always possible to include all the
of the temperature and of the intercepts, and the execution of the elements which are present in the sample into the CF-LIBS calculation.
algorithm may be iterated up to convergence. This is obvious for the trace elements whose concentration is lower
El Sherbini et al. [86] proposed to quantify the amount of self- than the instrumental detection limit. Sometimes one or more ele-
absorption of individual spectral lines by comparing the observed ments cannot be quantified because they don't have transitions in the
Lorentzian width with the value expected from Stark broadening. The available spectral range, or because all their lines are self-absorbed,
method can be applied to spectral lines with known Stark broadening etc. The closure condition, Eq. (4), is meaningful only if all the ele-
coefficient, and to measurements with sufficiently high resolution. ments are included in the sum. If the undetected element is a trace
Sun and Yu [18] proposed a simplified procedure for correcting one, the relative error affecting the concentration of the other
self-absorption effects in calibration-free analysis. They choose a elements is negligible. When one or more among the characteristic
reference line, λR, for every species in the plasma, with supposedly elements of the matrix cannot be measured, the direct application of
negligible absorption, then calculate the temperature from the Eq. (4) would give an overestimation of the concentration of the
Boltzmann plot using the uncorrected lines intensity and extrapolate detected elements.
 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14 11

5.1. Reduced composition as an example. If we detect in the spectrum the C I line at 247.8 nm,
we can estimate the number density of neutral carbon atoms (via the
The closure equation can also be applied to an incomplete list of intercept in the Boltzmann plane); for the number density of carbon
components, in order to obtain the relative proportion of the elements ions, we resort to the Saha equation and find out the result. For typical
in a sort of reduced composition. Following this idea, De Giacomo et al. conditions in LIBS plasmas, the C II density is much lower than C I
measured the relative proportion of the minor components of copper- density. The abundance of C in the sample is therefore mostly based
based alloys [27]. An analogous approach was applied to spatially on the direct measurement and in a low percentage on the Saha
resolved spectra of a copper alloy by Aguilera et al., who measured the extrapolation. Now suppose we are measuring Cu I lines and not Cu II
proportion of the three minor elements, while the copper lines were lines, due to spectral range limitations in the experimental apparatus.
neglected, being self-absorbed [28]. In the last two cases, however, In typical LIBS conditions, Cu is about 90–95 % ionized: so we must
partially modified CF-LIBS algorithms were implemented, as de- estimate Cu II density from the Saha equilibrium, based on the
scribed in Section 3.3. experimental data of Cu I. Low abundance means low intensity and
The CF-LIBS approach can also be exploited to determine thus strong fluctuations: the measurement error may be strongly
concentration ratios. An example is given in ref. [37] with the CF- amplified in the calculation. Generally speaking, it is advisable to
LIBS analysis of hair. At the experiment design stage, the simplicity of design the measurement layout in such a way to detect the most
measurement was preferred to the completeness of information that abundant species in the plasma.
could be obtained. The hairs were irradiated in ambient air, and
elements common to the sample and buffer gas could not be quan- 5.3. Recognition of elements
tified. However, the concentration ratio between mineral elements
(Na/K, Na/Mg, etc.) was obtained. This approach was also adopted to In the analysis of unknown samples, spectral lines must be checked
measure Al/Cu and Na/Cu concentration ratios with the aim to and assigned to a probable emitting element. For CF-LIBS quantitative
discriminate between azurite and lazurite pigments on parchments analysis, all the components of the sample must be correctly re-
[31]. cognized. According to the literature, this step of the analysis is
commonly carried out by the experienced analyst. In ref. [17] Herrera
5.2. Exploitation of Saha equilibrium et al. outlined that this process involves a great deal of decision-
making and requires much knowledge and experience. Colao et al. [32]
It may also happen that only spectral lines from one of the species mentioned the need of a complementary measurement to identify the
of a given element are detected. In this case it is still possible to elements present in the unknown sample as a minor drawback of the
include the element into the CF-LIBS calculations, provided the num- CF-LIBS approach. An automatic routine able to identify the major
ber density of the missing species is estimated via the Saha equi- elemental components of a LIBS spectrum, to be joined to CF-LIBS
librium equation. This step requires a knowledge of the electron quantitative analysis, would be extremely useful. Semi-automatic
number density, which can be estimated from the spectrum, by analysis software for peaks identification was presented by Mateo et al.
measuring the Stark broadening of suitable spectral lines with [88]. A different approach, based on algorithms developed in the field
known broadening parameters. Fornarini et al. showed in ref. [25] of text retrieval, is proposed by Amato et al. [89]. The software is
the dependence of the CF-LIBS results on the hypotheses assumed by designed to automatically scroll a list of peaks, providing at the end a
the analyst. In fact, the authors report the analysis of a certified ranking list of probability for the emitting elements.
quaternary bronze, and in particular the results obtained including
only the neutral species and those obtained calculating also the 5.4. Choice of analytical lines
abundance of ionic species through the Saha equilibrium equation.
The distance in the first case (not reported in Table 2 for clarity) is Beside the matching of lines and emitting elements, also the choice
12.3, while it drops to 5.1 after taking account of ionic species. De of suitable lines for quantitative analysis is a crucial step. In De
Giacomo et al. [27], on the contrary, based their analysis only on the Giacomo et al. [35] four criteria were given for selecting appropriate
abundance of neutral species after evaluating that ions abundance spectral lines: 1) no resonance lines for any element and lower level
was negligible in their measurement conditions. higher than 6000 cm− 1 for high concentration elements, to minimize
When using the Saha equilibrium to derive pieces of information self-absorption effects; 2) no lines with spontaneous emission rate
not directly present in the spectrum two points must be borne in lower than 2 × 106 s− 1, because this is the same time scale of plasma
mind. First, the approximation of the plasma as a homogeneous variations; 3) no ionic lines to avoid bias due to departures from Saha
sphere neglects the existence of spatial structures inside the plume. equilibrium; 4) no spectral lines with high intensity, to avoid
Temperature, electron density and thus ions density may be rather overestimation due to integration along the optical path.
different in the center of the plume compared to the peripheral Aydin et al. [90] proposed an interesting method of selection of
regions. Earlier investigations did not reveal significant differences in spectral lines to be used in a Boltzmann plot. Their aim was to build a
the spatial distribution of elements inside the plume [28]. Thus, if the systematic and objective procedure to improve the precision of the
Saha equation is applied to spatially-integrated spectra to calculate temperature calculation. A pre-selection of spectral lines is carried out
the density of non-detected species which are all neutral or all ionic, based on the calculation of theoretical emission coefficients. The lines
the result is probably self-consistent. If the available data refer partly with coefficients higher than the fixed threshold are then checked for
to neutral and partly to ionic species, the use of the Saha equation spectral interference by other components in the sample. Afterwards,
leads to a confusing picture of the plasma, where information related the pre-selected lines are assessed using experimental data within an
to the central region is superimposed on information related to the iterative optimization step where those data points which appear as
borders. This applies, however, also to the case where all the species outliers in all time windows are discarded. Owing to its transferability,
are directly measured from the spectrum, especially if the plasma is this method seems to be appropriate to set a quantitative and objective
strongly inhomogeneous, so that the majority of the ionic emission is basis for the choice of spectral lines to be used in CF-LIBS analysis.
produced in conditions rather different from those where atomic
emission is generated. 5.5. Photon statistics
The second point is that the estimation of the number density of
the non-detected species from the Saha equilibrium may introduce The last criterion mentioned by De Giacomo et al. [35] and
significant errors in the calculation. Let's compare the case of C and Cu reported above in Section 5.4 is in conflict with the recommendations
12 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

given by Mauchien and coworkers [33,91], who studied the which knowledge is still to be collected or consolidated. Based on the
dependence of relative standard deviation (RSD) of LIBS measure- concept of distance of the results from the nominal concentration
ment on the spectral line intensity. It must be noticed however that values, here proposed, the analyses carried out on metallic alloys
different conclusions might be due to the investigation of different samples were quite successful, whilst the analyses of other kinds of
experimental conditions (such as experimental apparatus, buffer gas materials like rocks or dielectrics were shown to be more problematic.
and especially detection time). In ref. [33] Mauchien and coworkers In any case, CF-LIBS can be proposed as an analytical method
evaluated the RSD of a large number of spectral lines over 50 mea- suited for the global characterization of the matrix of an unknown
surements on several samples of natural rocks. Their study demon- sample, while it can only provide a qualitative estimation of the trace
strated that, beside sample heterogeneity (which of course depends components. This is an intrinsic limit of CF-LIBS, which stems from the
on the kind of material investigated), shot noise is a major source of closure equation, because even a small uncertainty affecting the result
uncertainty of measurement. Using detectors such as an ICCD, a var- for the major components is translated into a huge relative error
iable multiplication factor relates the observed intensity in arbitrary affecting the minor components [15]. However, this limit can be easily
units to the number of photons illuminating the photo-cathode. The bypassed when an internal standard is available.
observed intensity of 25,000 a.u. corresponds in the case under con- The literature reviewed here is generous of information and
sideration to 312 photoelectrons, with an incidence of shot noise of suggestions: many kinds of materials have been investigated, many
6%. The authors recommend therefore using in quantitative analysis variants of the original algorithm have been proposed. However, it is
only lines with a sufficiently high intensity, in order to limit the effect impossible to ignore the general lack of a systematic method. In many
of shot noise. A comparison of the incidence of shot noise in CCD and works only one or a few samples were analyzed, thus offering poor
Intensified CCD is proposed by Mueller et al. [92]. With steel samples, statistical information. It is desirable that systematic studies, involving
they obtained comparable signal-to-noise ratios. large numbers of samples and variable experimental settings will be
It is evident that statistical fluctuations have proportionally a undertaken, to confirm and enlarge the knowledge gathered for
stronger impact in those cases in which a low photon yield is ob- example through the works by Herrera et al.[19] and De Giacomo et al.
tained. Difficulties have been reported about the analysis of trans- [27].
parent crystal samples, because a small fraction of the laser pulse It would be extremely interesting to compare the CF-LIBS
radiation is absorbed, so that small changes in the state of the sample performance with single and double pulse spectra, in the hypothesis
surface causes large variation of the absorbed energy and conse- that the latter should be closer to LTE. It would also be worth
quently of the plasma emission intensity [23]. investigating in a more systematic way the use of spectral lines of only
Elements with low concentration in the sample are subjected to one ionization stage, to check its effectiveness to eliminate the pro-
strong signal fluctuations. However, the signal from major compo- blems related to plasma inhomogeneity. It is also expected that a
nents may also undergo similar problems. Often the LIBS measure- greater exchange of data and knowledge with plasma modelists
ments are carried out in ambient atmosphere. This is a good practice would be beneficial.
in view of the plume confinement which slows the dynamical evo- The above mentioned topics are only some suggestions for the
lution. The drawback is that the ablation of the target surface is much possible directions which would deserve to be investigated. The
reduced due to the shielding of the trailing part of laser pulse by the general feeling is that a larger involvement of the scientific com-
growing plasma. At the end of the irradiation, on average over the munity would be necessary in order to develop the knowledge needed
whole plume, atoms from air elements are much more abundant than to take the best out of the CF-LIBS approach.
atoms from the target. For some materials and experimental settings,
the numerical abundance of target atoms in the plasma may be of the Acknowledgements
order of 1% of the total [38]. This corresponds to number density
values of the order of 1015 cm− 3. It is easily derived that trace ele- The authors thank Dr. Giuseppe Amato of ISTI-CNR for helpful
ments which in the target have concentration of 10 ppm, in the discussion about metrics.
plasma have numerical density of 1010 cm− 3. Very strong fluctuations
can be expected for the signal emitted by these species. Double pulse References
is usually beneficial in this view, because it enhances the target
[1] A.W. Miziolek, V. Palleschi, I. Schechter (Eds.), Laser Induced Breakdown
ablation, reduces the number density of gas atoms in the plume and
Spectroscopy, Cambridge University Press, Cambridge, 2006.
leads to an overall improvement of the reproducibility of the mea- [2] D.A. Cremers, L.J. Radziemski, Handbook of Laser-Induced Breakdown Spectros-
surement [93]. copy, Wiley, New York, 2006.
The incidence of the different noise sources in shot-to-shot micro- [3] J.P. Singh, S.N. Thakur (Eds.), Laser Induced Breakdown Spectroscopy, Elsevier
Science, Amsterdam, 2006.
probe LIBS measurements was also investigated by Mermet et al. [91] [4] C. Chaléard, P. Mauchien, N. Andre, J. Uebbing, J.L. Lacour, C. Geertsen, Correction of
using a steel sample. That work outlined especially the role of matrix effects in quantitative elemental analysis with laser ablation optical
instrumental drift during the measurement. Even though a drift in emission spectrometry, J. Anal. At. Spectrom. 12 (1997) 183–188.
[5] U. Panne, C. Haisch, M. Clara, R. Niessner, Analysis of glass and glass melts during
intensity could appear insignificant for CF-LIBS analysis, care should the vitrification process of fly and bottom ashes by laser-induced plasma
be applied in checking the effect of thermal equilibration on the spectroscopy. Part I: Normalization and plasma diagnostics, Spectrochim. Acta
dispersive elements, because wavelength-dependent shifts may affect Part B 53 (1998) 1957–1968.
[6] S.I. Gornushkin, I.B. Gornushkin, J.M. Anzano, B.W. Smith, J.D. Winefordner,
the detection system response. Effective normalization technique for correction of matrix effects in laser-induced
In principle, the analysis of single shot spectra would be preferable, breakdown spectroscopy detection of magnesium in powdered samples, Appl.
for a better guarantee of self-consistence, in order to avoid the effect Spectrosc. 56 (2002) 433–436.
[7] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti, E. Tognoni, New procedure for
of flicker noise [91]. In fact, single shot events have been shown to be
quantitative elemental analysis by laser-induced plasma spectroscopy, Appl.
dominated by high irreproducibility, so that the central limit theorem Spectrosc. 53 (1999) 960–964.
may not apply to their distribution [94]. Therefore, the usual average [8] I.B. Gornushkin, A.Y. Kazakov, N. Omenetto, B.W. Smith, J.D. Winefordner,
Radiation dynamics of post-breakdown laser induced plasma, Spectrochim. Acta
of the signal over many shots may not be statistically representative.
Part B 59 (2004) 401–418.
[9] I.B. Gornushkin, A.Y. Kazakov, N. Omenetto, B.W. Smith, J.D. Winefordner,
6. Conclusions Experimental verification of a radiative model of laser-induced plasma expanding
into vacuum, Spectrochim. Acta Part B 60 (2005) 215–230.
[10] A.Y. Kazakov, I.B. Gornushkin, N. Omenetto, B.W. Smith, J.D. Winefordner,
The inspection of the earlier literature regarding the applications Radiative model of post-breakdown laser-induced plasma expanding into
of CF-LIBS revealed some interesting trends and many aspects on ambient gas, Appl. Opt. (2006) 2810–2820.
 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14 13

[11] P. Yaroshchyk, D. Body, R J.S. Morrison, B.L. Chadwick, A semi-quantitative aqueous solutions: double ratio plasma diagnostics combined with an internal
standard-less analysis method for laser-induced breakdown spectroscopy, standard, J. Anal. At. Spectrom. 24 (2009) 655–662.
Spectrochim. Acta Part B 61 (2006) 200–209. [40] J.M. Mermet, Quality of calibration in inductively coupled plasma atomic emission
[12] J. Borovicka, A fireball spectrum analysis, Astron. Astrophys. 279 (1993) 627–645. spectrometry, Spectrochim. Acta Part B 49 (1994) 1313–1324.
[13] J.M. Trigo-Rodriguez, J. Llorca, J. Borovicka, J. Fabregat, Chemical abundances [41] J.M. Mermet, Limit of quantitation in atomic spectrometry: an unambiguous
determined from meteor spectra: I. Ratios of the main chemical elements, concept? Spectrochim. Acta Part B 63 (2008) 166–182.
Meteorit. Planet Sci. 38 (2003) 1283–1294. [42] N.M. Faber, B.R. Kowalski, Propagation of measurement errors for the validation of
[14] J.M. Trigo-Rodriguez, J. Llorca, J. Fabregat, Chemical abundances determined from predictions obtained by principal component regression and partial least squares,
meteor spectra — II. Evidence for enlarged sodium abundances in meteoroids, J. Chemom. 11 (1997) 181–238.
Mon. Not. R. Astron. Soc. 348 (2004) 802–810. [43] J.A. Aguilera, C. Aragón, Characterization of a laser-induced plasma by spatially
[15] E. Tognoni, G. Cristoforetti, S. Legnaioli, V. Palleschi, A. Salvetti, M. Mueller, U. resolved spectroscopy of neutral atom and ion emissions. Comparison of local
Panne, I. Gornushkin, A numerical study of expected accuracy and precision in and spatially integrated measurements, Spectrochim. Acta Part B 59 (2004)
calibration-free laser-induced breakdown spectroscopy in the assumption of ideal 1861–1876.
analytical plasma, Spectrochim. Acta Part B 62 (2007) 1287–1302. [44] NIST Atomic Spectra Database (version 3.1.5), [Online]. Available: [Link]
[16] V.L. Lednev, S.M. Pershin, Plasma stoichiometry correction method in laser- [Link]/asd3 [2008, November 3]. National Institute of Standards and Technology,
induced breakdown spectroscopy, Laser Phys. 18 (2008) 1–5. Gaithersburg, MD.
[17] K. Herrera, E. Tognoni, N. Omenetto, I.B. Gornushkin, B.W. Smith, J.D. Winefordner, [45] P.L. Smith, C. Heise, J.R. Esmond and R.L. Kurucz, Atomic spectral line database,
Comparative study of two standard-free approaches in laser-induced breakdown built from atomic data files from R.L. Kurucz' CD-ROM 23, [Link]
spectroscopy as applied to the quantitative analysis of aluminum alloy standards [Link]/cgi-bin/ssi/test/kurucz/[Link].
under vacuum conditions, J. Anal. At. Spectrom. 24 (2009) 426–438. [46] L. Fornarini, V. Spizzichino, F. Colao, R. Fantoni, V. Lazic, Influence of laser
[18] L. Sun, H. Yu, Correction of self-absorption effect in calibration-free laser-induced wavelength on LIBS diagnostics applied to the analysis of ancient bronzes, Anal.
breakdown spectroscopy by an internal reference method, Talanta 79 (2009) Bioanal. Chem. 385 (2006) 272–280.
388–395. [47] J.A. Aguilera, C. Aragón, Multi-element Saha-Boltzmann and Boltzmann plots in
[19] K. Herrera, E. Tognoni, B.W. Smith, N. Omenetto, J.D. Winefordner, Semi- laser-induced plasmas, Spectrochim. Acta Part B 62 (2007) 378–385.
quantitative analysis of metal alloys, brass and soil samples by calibration-free [48] X.L. Mao, A.C. Ciocan, O.V. Borisov, R.E. Russo, Laser ablation processes
laser-induced breakdown spectroscopy: recent results and considerations, J. Anal. investigated using inductively coupled plasma-atomic emission spectroscopy
At. Spectrom. 24 (2009) 413–425. (ICP-AES), Appl. Surf. Sci. 127–129 (1998) 262–268.
[20] D. Bulajic, M. Corsi, G. Cristoforetti, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni, [49] O.V. Borisov, X.L. Mao, R.E. Russo, Effects of crater development on fractionation
A procedure for correcting self-absorption in calibration free-laser induced and signal intensity during laser ablation inductively coupled plasma mass
breakdown spectroscopy, Spectrochim. Acta Part B 57 (2002) 339–353. spectrometry, Spectrochim. Acta Part B 55 (2000) 1693–1704.
[21] M. Corsi, G. Cristoforetti, V. Palleschi, A. Salvetti, E. Tognoni, A fast and accurate [50] J. Koch, A. von Bohlen, R. Hergenroder, K. Niemax, Particle size distributions and
method for the determination of precious alloys caratage by laser induced plasma compositions of aerosols produced by near-IR femto- and nanosecond laser
spectroscopy, Eur. Phys. J. D 13 (2001) 373–377. ablation of brass, J. Anal. At. Spectrom. 19 (2004) 267–272.
[22] M.V. Bel'kov, V.S. Burakov, V.V. Kiris, N.M. Kozhukh, S.N. Raikov, Spectral standard- [51] D. Bleiner, Z. Chen, D. Autrique, A. Bogaerts, Role of laser-induced melting and
free laser microananlysis of gold alloys, J. Appl. Spectrosc. 72 (2005) 376–381. vaporization of metals during ICP-MS and LIBS analysis, investigated with
[23] V.S. Burakov, V.V. Kiris, P.A. Naumenkov, S.N. Raikov, Calibration-free laser spectral computer simulations and experiments, J. Anal. At. Spectrom. 21 (2006) 910–921.
analysis of glasses and copper alloys, J. Appl. Spectrosc. 71 (2004) 740–746. [52] C. Nouvellon, C. Chaléard, J.L. Lacour, P. Mauchien, Stoichiometry study of laser
[24] F. Colao, R. Fantoni, V. Lazic, L. Caneve, A. Giardini, V. Spizzichino, LIBS as a produced plasma by optical emission spectroscopy, Appl. Surf. Sci. 138–139
diagnostic tool during the laser cleaning of copper based alloys: experimental (1999) 306–310.
results, J. Anal. At. Spectrom. 19 (2004) 502–504. [53] C.C. Garcia, H. Lindner, A. von Bohlen, C. Vadlab, K. Niemax, Elemental
[25] L. Fornarini, F. Colao, R. Fantoni, V. Lazic, V. Spizzicchino, Calibration analysis of fractionation and stoichiometric sampling in femtosecond laser ablation, J. Anal.
bronze samples by nanosecond laser induced breakdown spectroscopy: a theoretical At. Spectrom. 23 (2008) 470–478.
and experimental approach, Spectrochim. Acta Part B 60 (2005) 1186–1201. [54] J.B. Ko, W. Sdorra, K. Niemax, On the internal standardization in optical emission
[26] S.M. Pershin, F. Colao, V. Spizzichino, Quantitative analysis of bronze samples by spectrometry of microplasmas produced by laser ablation of solid samples,
laser-induced breakdown spectroscopy (LIBS): a new approach, model, and Fresenius J Analyt Chem 335 (1989) 648–651.
experiment, Laser Phys. 16 (2006) 455–467. [55] J.J. Gonzalez, A. Fernandez, X.L. Mao, R.E. Russo, Scanning vs. single spot laser
[27] A. De Giacomo, M. Dell'aglio, O. De Pascale, R. Gaudiuso, R. Teghil, A. Santagata, G.P. ablation (λ = 213 nm) inductively coupled plasma mass spectrometry, Spectro-
Parisi, Ns- and Fs-LIBS of copper-based-alloys: a different approach, Appl. Surf. Sci. chim. Acta Part B 59 (2004) 369–374.
253 (2007) 7677–7681. [56] M. Adamson, A. Padmanabhan, G.J. Godfrey, S.J. Rehse, Laser-induced breakdown
[28] J.A. Aguilera, C. Aragón, G. Cristoforetti, E. Tognoni, Application of calibration-free spectroscopy at a water/gas interface: a study of bath gas-dependent molecular
laser induced breakdown spectroscopy to radially resolved spectra from a copper- species, Spectrochim. Acta Part B 62 (2007) 1348–1360.
based alloy laser induced plasma, Spectrochim. Acta Part B 64 (2009) 685–689. [57] J. Hermann, C. Dutouquet, Local thermal equilibrium plasma modeling for
[29] M. Corsi, G. Cristoforetti, M. Giuffrida, M. Hidalgo, S. Legnaioli, L. Masotti, V. analyses of gas-phase reactions during reactive-laser ablation, J. Appl. Phys.
Palleschi, A. Salvetti, E. Tognoni, C. Vallebona, A. Zanini, Archaeometric analysis of 91 (2002) 10188–10193.
ancient copper artefacts by laser-induced breakdown spectroscopy technique, [58] G. Colonna, A. Casavola, M. Capitelli, Modelling of LIBS plasma expansion,
Microchim. Acta 152 (2005) 105–111. Spectrochim. Acta Part B 56 (2001) 567–586.
[30] I. Borgia, L. Burgio, M. Corsi, R. Fantoni, V. Palleschi, A. Salvetti, M.C. Squarcialupi, E. [59] V.S. Burakov, N.V. Tarasenko, N.A. Savastenko, Plasma chemistry in laser ablation
Tognoni, Self-calibrated quantitative elemental analysis by laser-induced plasma processes, Spectrochim. Acta Part B 56 (2001) 961–971.
spectroscopy: application to pigment analysis, J. Cult. Heritage 1 (2000) [60] A. Casavola, A. De Giacomo, M. Dell'aglio, F. Taccogna, G. Colonna, O. De Pascale, S.
S281–S286. Longo, Experimental investigation and modelling of double pulse laser induced
[31] M. Bicchieri, M. Nardone, P.A. Russo, A. Sodo, M. Corsi, G. Cristoforetti, V. Palleschi, plasma spectroscopy under water, Spectrochim. Acta Part B 60 (2005) 975–985.
A. Salvetti, E. Tognoni, Characterization of azurite and lazurite based pigments by [61] A. De Giacomo, M. Dell'aglio, A. Casavola, G. Colonna, O. De Pascale, M. Capitelli,
laser induced breakdown spectroscopy and micro-raman spectroscopy, Spectro- Elemental Chemical Analysis of Submerged Targets by Double-Pulse Laser-
chim. Acta Part B 56 (2001) 915–922. Induced Breakdown Spectroscopy, Anal. Bioanal. Chem. 385 (2006) 303–311.
[32] F. Colao, R. Fantoni, V. Lazic, A. Paolini, F. Fabbri, G.G. Ori, L. Marinangeli, A. Baliva, [62] M. Baudelet, L. Guyon, J. Yu, J.-P. Wolf, T. Amodeo, E. Fréjafon, P. Laloi,
Investigation of LIBS feasibility for in situ planetary exploration: an analysis on Femtosecond time-resolved laser-induced breakdown spectroscopy for detection
Martian rock analogues, Planet. Space Sci. 52 (2004) 117–123. and identification of bacteria: a comparison to the nanosecond regime, J. Appl.
[33] B. Sallé, J.L. Lacour, P. Mauchien, P. Fichet, S. Maurice, G. Manhes, Comparative Phys. 99 (2006) 084701.
study of different methodologies for quantitative rock analysis by laser-induced [63] F.R. Doucet, P.J. Faustino, M. Sabsabi, R.C. Lyon, Quantitative molecular analysis
breakdown spectroscopy in a simulated Martian atmosphere, Spectrochim. Acta with molecular bands emission using laser-induced breakdown spectroscopy and
Part B 61 (2006) 301–313. chemometrics, J. Anal. At. Spectrom. 23 (2008) 694–701.
[34] L. Wang, C. Zhang, Y. Feng, Controlled calibration method for laser induced [64] J.A.M. van der Mullen, Excitation equilbria in plasmas; a classification, Phys. Rep.
breakdown spectroscopy, Chinese Optics Letters 6 (2008) 5–8. 191 (1990) 109–220.
[35] A. De Giacomo, M. Dell'aglio, O. De Pascale, S. Longo, M. Capitelli, Laser induced [65] H.R. Griem, Validity of local thermal equilibrium in plasma spectroscopy, Phys.
breakdown spectroscopy on meteorites, Spectrochim. Acta Part B 62 (2007) Rev. 131 (1963) 1170–1176.
1606–1611. [66] O. Barthelemy, J. Margot, S. Laville, F. Vidal, M. Chaker, B. Le Drogoff, T.W. Johnston,
[36] S. Pandhija, A.K. Rai, In situ multielemental monitoring in coral skeleton by CF- M. Sabsabi, Investigation of the state of local thermodynamic equilibrium of a
LIBS, Appl. Phys. B 94 (2009) 545–552. laser-produced aluminum plasma, Appl. Spectrosc. 59 (2005) 529–536.
[37] M. Corsi, G. Cristoforetti, M. Hidalgo, S. Legnaioli, V. Palleschi, A. Salvetti, E. [67] M. Capitelli, F. Capitelli, A. Eletskii, Non-equilibrium and equilibrium problems in
Tognoni, C. Vallebona, Application of laser-induced breakdown spectroscopy laser-induced plasmas, Spectrochim. Acta Part B 55 (2000) 559–574.
technique to hair tissue mineral analysis, Appl. Opt. 42 (2003) 6133–6137. [68] G. Cristoforetti, A. De Giacomo, M. Dell’Aglio, S. Legnaioli, E. Tognoni, V. Palleschi, N.
[38] G. Cristoforetti, S. Legnaioli, V. Palleschi, E. Tognoni, P.A. Benedetti, Observation of Omenetto, Local thermodynamic equilibrium in laser-induced breakdown spectros-
different mass removal regimes during the laser ablation of an aluminium target copy: beyond the McWhirter Criterion Spectrochim, Acta Part B 65 (2010) 86–95.
in air, J. Anal. At. Spectrom. 23 (2008) 1518–1528. [69] J.A. Aguilera, C. Aragón, Temperature and electron density distributions of laser-
[39] E. Tognoni, M. Hidalgo, A. Canals, G. Cristoforetti, S. Legnaioli, V. Palleschi, Towards induced plasmas generated with an iron sample at different ambient gas
a calibration-less ICP-AES method for the determination of trace elements in pressures, Appl. Surf. Sci. 197 (2002) 273–280.
14 E. Tognoni et al. / Spectrochimica Acta Part B 65 (2010) 1–14

[70] J.A. Aguilera, C. Aragón, J. Bengoechea, Spatial characterization of laser-induced [82] I.B. Gornushkin, J.M. Anzano, L.A. King, B.W. Smith, N. Omenetto, J.D. Winefordner,
plasmas by deconvolution of spatially resolved spectra, Appl. Opt. 42 (2003) Curve of growth methodology applied to laser-induced plasma emission
5938–5946. spectroscopy, Spectrochim. Acta Part B 54 (1999) 491–503.
[71] J.A. Aguilera, J. Bengoechea, C. Aragón, Spatial characterization of laser induced [83] I.B. Gornushkin, C.L. Stevenson, B.W. Smith, N. Omenetto, J.D. Winefordner,
plasmas obtained in air and argon with different laser focusing distances, Modeling an inhomogeneous optically thick laser induced plasma: a simplified
Spectrochim. Acta Part B 59 (2004) 461–469. theoretical approach, Spectrochim. Acta Part B 56 (2001) 1769–1785.
[72] C. Aragón, F. Penalba, J.A. Aguilera, Spatial characterization of laser-induced [84] C. Aragón, J. Bengoechea, J.A. Aguilera, Influence of the optical depth on spectral
plasmas: distributions of neutral atom and ion densities, Appl. Phys. A 79 (2004) line emission from laser-induced plasmas, Spectrochim. Acta Part B 56 (2001)
1145–1148. 619–628.
[73] O. Barthélemy, J. Margot, M. Chaker, M. Sabsabi, F. Vidal, T.W. Johnston, S. Laville, [85] J.A. Aguilera, J. Bengoechea, C. Aragón, Curves of growth of spectral lines emitted
B. Le Drogoff, Influence of the laser parameters on the space and time by a laser-induced plasma: influence of the temporal evolution and spatial
characteristics of an aluminum laser-induced plasma, Spectrochim. Acta Part B inhomogeneity of the plasma, Spectrochim. Acta Part B 58 (2003) 221–237.
60 (2005) 905–914. [86] A.M. El Sherbini, T.M. El Sherbini, H. Hegazy, G. Cristoforetti, S. Legnaioli, V.
[74] A. De Giacomo, Experimental characterization of metallic titanium-laser induced Palleschi, L. Pardini, A. Salvetti, E. Tognoni, Evaluation of self-absorption
plasma by time and space resolved optical emission spectroscopy, Spectrochim. coefficients of aluminum emission lines in laser-induced breakdown spectroscopy
Acta Part B 58 (2003) 71–83. measurements, Spectrochim. Acta Part B 60 (2005) 1573–1579.
[75] A. De Giacomo, M. Dell'aglio, D. Bruno, R. Gaudiuso, O. De Pascale, Experimental [87] H.-Y. Moon, K.K. Herrera, N. Omenetto, B.W. Smith, J.D. Winefordner, On the
and theoretical comparison of single-pulse and double-pulse laser induced usefulness of a duplicating mirror to evaluate self-absorption effects in laser
breakdown spectroscopy on metallic samples, Spectrochim. Acta Part B 63 induced breakdown spectroscopy, Spectrochim. Acta Part B 64 (2009) 702–713.
(2008) 805–816. [88] M.P. Mateo, G. Nicolàs, V. Pinon, J.C. Alvarez, A. Ramil, A. Yanez, Versatile software
[76] A. De Giacomo, M. Dell'aglio, R. Gaudiuso, G. Cristoforetti, S. Legnaioli, V. Palleschi, for semiautomatic analysis and processing of laser-induced plasma spectra,
E. Tognoni, Spatial distribution of hydrogen and other emitters in aluminum laser- Spectrochim. Acta Part B 60 (2005) 1202–1210.
induced plasma in air and consequences on spatially integrated laser-induced [89] G. Amato, F. Sorrentino, G. Cristoforetti, S. Legnaioli, V. Palleschi, E. Tognoni,
breakdown spectroscopy measurements, Spectrochim. Acta Part B 63 (2008) Unassisted element identification from Laser Induced Breakdown Spectra with
980–987. automatic ranking techniques inspired by text retrieval, Euro-Mediterranean
[77] M. Corsi, G. Cristoforetti, M. Hidalgo, D. Iriarte, S. Legnaioli, V. Palleschi, A. Salvetti, Symposium on Laser-Induced Breakdown Spectroscopy, Tivoli, 28th September-
E. Tognoni, Temporal and spatial evolution of a laser-induced plasma from a steel 1st October 2009, oral communication.
target, Appl. Spectrosc. 57 (2003) 715–721. [90] U. Aydin, P. Roth, C.D. Gehlen, R. Noll, Spectral line selection for time-resolved
[78] M. Corsi, G. Cristoforetti, M. Giuffrida, M. Hidalgo, S. Legnaioli, V. Palleschi, A. investigations of laser-induced plasmas by an iterative Boltzmann plot method,
Salvetti, E. Tognoni, C. Vallebona, Three-dimensional analysis of laser induced Spectrochim. Acta Part B 63 (2008) 1060–1065.
plasmas in single and double pulse configuration, Spectrochim. Acta Part B [91] J.M. Mermet, P. Mauchien, J.L. Lacour, Processing of shot-to-shot raw data to
59 (2004) 723–735. improve precision in laser-induced breakdown spectrometry microprobe,
[79] G. Cristoforetti, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni, Characterization of Spectrochim. Acta Part B 63 (2008) 999–1005.
a collinear double pulse laser-induced plasma at several ambient gas pressures by [92] M. Mueller, I.B. Gornushkin, S. Florek, D. Mory, U. Panne, Approach to detection in
spectrally- and time-resolved imaging, Appl. Phys. B 80 (2005) 559–568. laser-induced breakdown spectroscopy, Anal. Chem. 79 (2007) 4419–4426.
[80] V. Lazic, R. Barbini, F. Colao, R. Fantoni, A. Palucci, Self-absorption model in [93] G. Cristoforetti, Orthogonal double-pulse versus single-pulse laser ablation at
quantitative laser induced breakdown spectroscopy measurements on soils and different air pressures: a comparison of the mass removal mechanisms, Spectro-
sediments, Spectrochim. Acta Part B 56 (2001) 807–820. chim. Acta Part B 64 (2009) 26–34.
[81] C.A. D'angelo, D.M.D. Pace, G. Bertuccelli, D. Bertuccelli, Laser induced breakdown [94] A.P.M. Michel, A.D. Chave, Analysis of laser-induced breakdown spectroscopy
spectroscopy on metallic alloys: solving inhomogeneous optically thick plasmas, spectra: the case for extreme value statistics, Spectrochim. Acta Part B 62 (2007)
Spectrochim. Acta Part B 63 (2008) 367–374. 1370–1378.