Professional Documents
Culture Documents
(2012) (Hahn DW, Omenetto N) (Part II) (Focal Point)
(2012) (Hahn DW, Omenetto N) (Part II) (Focal Point)
(2012) (Hahn DW, Omenetto N) (Part II) (Focal Point)
Laser-Induced Breakdown
Spectroscopy (LIBS), Part
II: Review of Instrumental
and Methodological
Approaches to Material
Analysis and Applications
to Different Fields
The first part of this two-part review focused on summarizing the current state-of-the-art of plications. An attempt is made to provide an
the fundamental and diagnostics aspects of analytical LIBS, providing a contemporary updated view of the role played by LIBS in the
laser-induced plasmas, only touching briefly snapshot of LIBS applications, and highlighting various fields, with emphasis on applications
upon concepts such as sensitivity and detection new directions in laser-induced breakdown considered to be unique. We finally try to assess
limits and largely omitting any discussion of the spectroscopy, such as novel approaches, instru- where LIBS is going as an analytical field,
vast panorama of the practical applications of mental developments, and advanced use of where in our opinion it should go, and what
the technique. Clearly a true LIBS community chemometric tools. More specifically, we dis- should still be done for consolidating the
has emerged, which promises to quicken the cuss instrumental and analytical approaches technique as a mature method of chemical
pace of LIBS developments, applications, and analysis.
(e.g., double- and multi-pulse LIBS to improve
implementations. With this second part, a more
the sensitivity), calibration-free approaches, Index Headings: Laser-induced breakdown
applied flavor is taken, and its intended goal is
hyphenated approaches in which techniques spectroscopy; LIBS; Instrumentation; Analyti-
such as Raman and fluorescence are coupled cal sensitivity; Limits of detection; Signal-to-
Received 22 December 2011; accepted 10 with LIBS to increase sensitivity and informa- noise ratio; Calibration; Matrix effects; Dou-
January 2012. tion power, resonantly enhanced LIBS ap-
* Author to whom correspondence should be ble-pulse LIBS; Multi-pulse LIBS; Microanal-
sent. E-mail: omenetto@chem.ufl.edu. proaches, signal processing and optimization ysis; Hyphenated methods; Chemometrics;
DOI: 10.1366/11-06574 (e.g., signal-to-noise analysis), and finally ap- Remote detection; Applications.
No sample preparation
Is the feature
a LIBS
characteristic?
YES
Comments
Eq.
Expression Description number
L ¼ AX Definition of optical conductance (luminosity, throughput, etendue, light gathering power). T3.1
k0
Rs ¼ Resolving power defined at the central peak of the line (k0). T3.2
Dks
k0
Ps [ Ls Rs ¼ As Xs Definition of luminosity-resolving power product. This term can be interpreted as the ‘‘spectral T3.3
Dks efficiency’’ of the optical collection apparatus.
Z
S } Ls Is ðkÞdk } Ls Is0 The expression for the line emission signal, spectrally integrated over the line profile, is T3.4
Dkline
proportional to the optical conductance of the system.
Z
k0
B } Ls IB ðkÞdk}Ls IB ðk0 Þ Dks }Ls IB ðk0 Þ The expression for the continuum background emission signal is proportional to the optical T3.5
Rs conductance of the system and the spectral resolution of the spectrometer.
Dks
S Ls Is0 pffiffiffiffiffiffiffiffiffi pffiffiffiffiffi
} rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi}K Ls Rs }K Ps This expression shows that in the shot-noise limit, the signal-to-noise due to plasma T3.6
NB Ls IB ðk0 Þ k0 background, the signal-to-noise ratio increases proportionally to the luminosity-resolving
Rs power product.
Ae Xe ¼ As Xs The optical conductance is a constant throughout the optical system. This is a fundamental T3.7
relation valid for any optical system.
ðDks hs Þ p
ðde Le Þ Xe ¼ Eq. T3.7 rewritten in terms of plasma and spectrometer parameters, i.e., reciprocal linear T3.8
Rd 4F2 dispersion, spectral bandwidth, and F-number.
Dks ¼ Rd ws Definition of the spectral bandwidth of the spectrometer. In most cases, this quantity defines T3.9
the spectral resolution, since it is much larger than the diffraction-limited spectral width.
Dks hs
M¼ ¼ The linear magnification of the collection optics, rewritten in terms of plasma and spectrometer T3.10
Rd de Le parameters, allows linking the diameter and length of the emission slab in the plasma with
the spectral and geometrical parameters of the spectrometer.
hs Rd de
Le ¼ T3.11
Dks
p de2 Le p de3 hs Rd
Vseen ¼ ¼ Expression defining the volume of the emitting plasma seen by the detector. T3.12
4 4 Dks
p de3 hs Rd
Nseen ¼ nT Vseen ¼ nT Expression defining the number of emitting atoms seen by the detector, given a total atom T3.13
4 Dks number density in the plasma.
a
Definition of parameters in the equations of the table.
Equations T.3.1–T.3.6: L = Luminosity of the monochromator (cm2 sr); Xm = Solid angle of the monochromator (sr); Rm = Resolving power of the
monochromator; k0 = Central wavelength of the spectral line (cm); Dkm = Spectral bandwidth of the monochromator (cm); Pm = Luminosity-resolving power
product (cm2 sr); S = Measured emission signal (counts); Dk‘ = Spectral line width of the transition (cm); Is(k) = Spectral line intensity (W cm-2 nm-1); Is0 =
Spectrally integrated line intensity (W cm-2); B = Background signal (counts); IB(k) = Spectral background intensity (W cm-2 nm-1); NB = Background noise
(counts). Note the proportionality sign in the equations, implying the presence of a factor which makes the units on either side of the equations consistent.
Equations T.3.7–T.3.13: Ae,s = Areas relative to the emission source and spectrometer (cm2); Xe,s = Solid angles relative to the emission and spectrometer (sr); ws
= Spectrometer slit width (cm); hs = Spectrometer slit height (cm); M = Linear magnification of the collection optics; F = F-number of the spectrometer; de =
Diameter of the emission slab (cm); Le = Length of the emission slab (cm); Rd = Reciprocal linear dispersion of the spectrometer (nm/mm); Dks = Spectrometer
bandwidth (nm).
ity, thus invalidating the comparison of ICCD detectors, making this combina- Chen et al.96 have described a spec-
the detectors. tion useful in LIBS work (see, e.g., early trometer coupling a two-dimensional
The advantage of covering a large instrumental developments and spec- CCD detector with an integrating grat-
spectral interval with a single laser shot trometer characterization88–90). As a ing consisting of 10 sub-gratings. The
is a welcome feature of many LIBS curiosity, it is interesting to note that instrument allows full spectral coverage
applications, in particular stand-off anal- the World Space Observatory Ultravio- from 200 nm to 1000 nm without any
ysis and characterization of samples let (WSO/UV) will comprise a high mechanical moving elements.96
producing complicated spectra where resolution double echelle spectrograph In concluding this section, we draw
spectral interferences are likely to occur. (HIRDES) covering the near vacuum the readers’ attention to a paper by
Echelle spectrometers are known to UV range (174–310 nm) and the Taschuk et al.100 comparing four com-
cover a large wavelength interval and vacuum UV range (102–176 nm) with mon LIBS detection systems. The au-
can be combined with both CCD and a resolving power greater than 55 000.93 thors considered two Czerney–Turner
FIG. 3. Flowchart of the essential steps involved in the implementation of a model in the theoretical calculation of the concentration of the
various species and a comparison with the experimentally obtained values (see text for discussion).
have weaknesses (e.g., fractionation LA- level considered adequate for the spe- render the calibration issue more diffi-
ICP-MS) when applied to trace analysis. cific analytical task investigated, and we cult to tackle and solve. However, this
Moreover, the fact that either internal refer the reader to the Applications issue is of essential relevance for the
standard calibration or external calibra- sections. In addition, as pointed out by further development and acceptance of
tion methods are needed is not a Russo et al.16 in the discussion of laser LIBS as a quantitative method, and it
drawback per se. All quantitative ana- ablation, the occurrence of fractionation will only be through understanding and
lytical techniques rely on calibration and does not preclude the use of laser appreciation of these processes that
standards. Ironically, the versatility, ablation sampling for accurate chemical LIBS analytical schemes will be opti-
which is the best advantage of LIBS analysis.16 Therefore, the problem is not mized to realize their full quantitative
due to its capability of directly address- that LIBS cannot perform quantitative analytical potential.
ing any sample, is inherently also the analysis, but rather how to make LIBS In conclusion, if quantitation is indeed
very reason that matrix-matched stan- become as accurate and reliable as the the only vulnerable aspect of LIBS (as
dards are not available in many applica- other spectroscopic methods. In this Achilles had his heel as the only
tions. Moreover, the lack of sample respect, it is also worth cautioning that vulnerable part of his body), this is
preparation, which is another unique comparing LIBS with bulk analysis where research should then be focused
and attractive feature of LIBS, will be methods, e.g., ICP-AES, may sometime and vigorously continued.
challenged if extensive chemical or be meaningless, since bulk methods will Calibration and Curves of Growth.
physical (mechanical) sample manipula- determine the average concentration of The relation between the (spectrally
tion prior to laser ablation were the only the analyte in a sample mass that is integrated) thermal radiance, Bthermal, of
way to achieve good quantitative ana- much larger than that sampled by an emission line as a function of the
lytical data. LIBS.23 number density of the atomic (ionic)
LIBS has been shown to be capable of The complexity associated with laser– species in the plasma is exhaustively
performing quantitative analysis at a sample and laser–plasma interactions treated in all books related to spectro-
ground to be measured exactly at uation of the true background may directly applied to liquid samples,
the spectral position of the line. If not be easy, especially with low including, for example, liquid jets
such blank is not available, the spectral resolution spectrometers and droplets, water is not a common
background is taken with a sample and complex spectra, where spec- reference matrix for LIBS. As a
containing the analyte at very low tral interferences may easily occur. result, LOD charts, typical of ICP-
concentrations. Note that the eval- (ii) iDespite the fact that LIBS can be OES and ICP-MS, collecting sin-
Eq.
Expression Description number
n o1=2
rsum ¼ ½rðBÞp;koff-peak 2 þ ½rðSÞnet;k0 2 þ 2h ½rðBÞp;koff-peak ½rðSÞnet;k0 This expression refers to a single isolated spectral line (signal S) T4.1
superimposed on a flat background (signal B). The standard deviation
resulting from the sum of the two signals contains the correlation
coefficient h.
n o1=2
rdiff ¼ ½rðBÞp;koff-peak 2 þ ½rsum 2 -2h ½rðBÞp;koff-peak ½rsum This expression relates the standard deviation of the difference of the two T4.2
signals to the standard deviation of the sum and the correlation
coefficient.
ðrsum Þ2exp þ ðrB Þ2exp -ðrdiff Þ2exp
h¼ The correlation coefficient cannot be calculated directly from the standard T4.3
2 ðrsum Þexp ðrB Þexp deviation of the sum of the two signals, but can be evaluated from the
standard deviations of the sum of the difference and of the background.
rsum ðr2 þ r2S þ 2hrB rS Þ1=2
ðRSDÞsum ¼ ¼ B This expression shows how the relative standard deviation of the sum of T4.4
ðS þ BÞ ðS þ BÞ the analyte and background signals is related to the individual standard
deviations and the correlation coefficient.
2
1=2
rP r2 1
ðRSDÞratio ¼ 2
þ P2is -2h rP rPis This expression is relevant when an internal standard is used to normalize T4.5
P Pis P Pis the signal fluctuations. P and Pis refer to the signal peaks measured for
the analytical line and that of the internal standard chosen.
a
The equations reported in this table have been taken from H-Y. Moon, PhD Dissertation (University of Florida, 2009), and will be submitted for publication.
LODs scale with the inverse square sponding standard deviation and relative this, the approach has merits and it is
root of the number of shots used: standard deviation values (assuming a easy to implement in the practical LIBS
this, however, needs to be demon- Gaussian probability distribution func- routine. For example, it indicates how
strated experimentally. tion of the data). Note that the back- the measurement precision evolves dur-
ground and signal plus background are ing the plasma evolution while at the
The Standard Deviation–Relative acquired simultaneously and may there- same time evaluating the degree of
Standard Deviation Approach. We fore show some correlation. Moreover, correlation between the signal and the
have recently become interested in the only noises considered are back- background, thus helping to assess the
applying to LIBS spectra an approach ground noise and signal noise. Since the feasibility of single-shot normalization
that was first described in the case of a background cannot be measured under approaches based upon the spectral
highly heterogeneous system like atmo- the peak, it is assumed that the mea- continuum. An interesting outcome is
spheric aerosols.220 We call this ap- surement ‘‘off-peak’’ is representative of the possibility of directly spotting the
proach the standard deviation–relative the ‘‘true’’ background under the line. presence of self-absorption in some of
standard deviation (SD-RSD) approach The total noise expressions resulting the peaks observed (see the brass
(see Fig. 5). It is based upon single-shot from the application of error propagation spectrum in Fig. 5).
acquisition and processing of individual statistics are collected in Table IV. A final merit of the approach resides
spectra. As pointed out by Michel,44 an The ‘‘spectra’’ obtained by plotting in its capability of evaluating the
investigation into current research draws the standard deviation and relative usefulness of using an internal standard
attention to a range of applications for standard deviation versus wavelength for normalization purposes. From Eq.
single-shot LIBS, yet exposes the infan- (pixels) can now be compared with the T4.5, it is easy to predict the effect
cy of quantitative single-shot LIBS. We conventional spectrum obtained by plot- resulting from ratioing the analyte signal
therefore hope that the following con- ting the average signal versus wave- by an internal standard (see later). If we
siderations may trigger some additional length (pixels). Based upon the ex- assume that the two signals have
interest in the application of this mea- pressions given in Table IV, one can comparable magnitudes and noises (a
surement approach. predict the shape of the spectrum that reasonable assumption in most practical
The principle is the following. With a would be obtained for different limiting cases), the relative standard deviation of
given target and under given operating cases, in which the dominant noise may the ratio will improve only in the case
conditions, a certain number of single- be due to the background or to the where the noises in the signal and in the
pulse spectra are obtained (see Fig. 5). signal. As shown in Fig. 5, even under internal standard are correlated, i.e.,
Within the spectral region selected, for the simplifying assumptions made, mul- when h = 1. If the noises affecting the
each pixel (or equivalently for each tiple outcomes are possible, since the analyte signal and the internal standard
spectral resolution element) the average observed behavior will depend upon the are anti-correlated, i.e., if h = -1, the
signal levels for the background, signal source of limiting noise, the relative RSD will worsen by a factor of 2
plus background, and net signal are magnitudes of the signal and back- compared with the RSD of the analyte
computed, together with their corre- ground, and their correlation. Despite signal alone. Finally, if the two noises
FIG. 5. Illustration of the SD-RSD approach applied to single-shot measurements. The application of the statistical expressions given
in Table IV result in the simulated spectral plots of the RSD shown below the 55 single-shot spectra of Cu in a brass standard. The
simulations (a–d) show that, at the wavelengths corresponding to the two copper lines, the RSD can remain the same, increase, or decrease
depending upon the limiting noise of the measurement, the relative values of the signal and background and their degree of correlation. The
two spectra on the right, in panel (e), were obtained with Al-alloy standard R14 and show the SD (black trace) and the RSD (gray trace),
n the vicinity of the Mg 285.213 nm line. Note how the RSD can increase or decrease, depending upon the line chosen. Finally, the FWHM of
the signal at the Cu 324.754 nm spectrum (f) is smaller than that observed for the SD spectrum (g), indicating the presence of self-
absorption. Note here that the wavelengths listed correspond to the instrumental readings. [Reproduced with permission from Heh-Young
Moon, PhD Thesis, University of Florida, 2010.]
Eq.
Expression Description number
ed ¼ 1-expð-/s tobs Þ »/s tobs The detection efficiency is considered in the case of a repetitive transient behavior like that T5.1
ed ¼ 1-expð-/s tres Þ»/s tres encountered in LIBS. One single shot is considered. The two temporal parameters refer to T5.2
the residence time of the atoms in the observation volume and the actual observation time of
the emission. Their relative values define whether a measurement is stationary or non-
stationary.
gu -Eu ws hs Xe T0
/s ¼ ðAu‘ Þ e kT gD This expression relates the mean flux of signal counts per atom in the probe volume (s-1) to T5.3
ZðTÞ 4pAe the parameters of the transition and to the parameters of the optical detection system.
S ¼ ð/s tres Þ Nprobed ð/s tD Þ Nprobed The product /s tres becomes /s tD by taking into account the integration time used in the T5.4
single-pulse acquisition.
Nprobed ¼ Nsample eablation evaporization The total number of atoms probed will depend upon the efficiency of ablation and T5.5
vaporization. Note that the excitation efficiency is included in the Boltzmann factor. This
NA
Nsample ¼ ms wa number is related to the total number in the solid by the sample mass, the Avogadro number T5.6
Ma and the atomic mass and weight fraction of the analyte.
gu -Eu ws hs Xe T0
S ¼ ðAu‘ Þ e kT ðgD Þ ðtD Þ This final expression gives the number of counts per pulse predicted for a given transition, a T5.7
ZðTÞ 4pAe given mass ablated, a given ablation and vaporization efficiency, and a given electro-optical
collection/detection efficiency. By using the numerical values collected in the caption below,
NA
3 ms wa eablation evaporization we obtain ed = 3.1 3 10-7 counts atom-1 S = 2.91 3 105 wa.
Ma
a
The equations reported in this table have been derived from the analytical considerations given in Winefordner et al.17 The reader is referred to this paper for more
details as well as for the citation of the original literature from which the expressions were first derived [see, e.g., C.Th.J. Alkemade, Appl. Spectrosc., 35, 1 (1981)].
Definition of the parameters used in Eqs. T5.1–T5.7: ed = Detection efficiency; tobs = Observation time (s); tres = Residence time of the atoms in the observation
(probe) volume (s); /s = Mean flux of signal counts per atom in the probe volume (s-1); Au‘ = Einstein coefficient of spontaneous emission (s-1); gu = Statistical
weight of the upper level of the transition; Z(T) = Partition function at the temperature T of the plasma; Eu = Energy of the upper level of the transition (eV); k =
Boltzmann constant (eV/K); T = Plasma excitation temperature (K); ws = Monochromator entrance slit width (cm); hs = Monochromator entrance slit height (cm); Xe
= Emission solid angle (sr); T0 = Overall transmission of the optics and the environment; Ae = Plasma emission area (cm2); gD = Detector efficiency (counts/photon);
tD = Detector integration time (s); ms = Sample mass (g); wa = Weight fraction of analyte in the sample; Ma = Analyte atomic mass (g/mol); NA = Avogadro number
(atoms/mol); eablation = Ablation efficiency; evaporization = Vaporization efficiency.
Numerical values of the parameters used in Eq. T5.7: Au‘ = 108 s-1; gu/Z(T) = 1; exp(-Eu/kT) = 3.3 3 10-3, with Eu = 4 eV and T = 0.7 eV (8123 K); ws 3 hs
= 0.005 cm2; Xe = 0.047 sr (F/4.1); T0 = 0.1; Ae = 0.1 cm2; gD = 0.05; tD = 10-6 s; ms = 10 ng (10-8 g); wa = Weight fraction of analyte in the sample; Ma = (Cu)
= 64 (g/mol); NA = 6 3 1023 (atoms/mol); eablation = 0.1; evaporization = 0.1.
are independent, i.e., if h = 0, the RSD terms of counts per atom per pulse, of 10% (i.e., S = 100) is 0.034%, or 340
of
pffiffiffithe ratio will worsen by a factor of a typical LIBS experiment. In this case, ppm. Although it is tempting to play
2. This represent the classical case in a conventional laboratory setup is con- with the parameters to improve such
which the measurement is signal-noise sidered. The numerical values of the limit, one has to realize that the single-
limited, and both signals (analyte and optical and detector parameters are shot sensitivity as calculated is well
internal standard) are affected by shot typical of other calculations performed within the order of magnitude of the
noise. These results are of course well by Winefordner et al.17 for other spec- typical experimentally reported values in
known, and similar considerations in the troscopic methods. As explained in the literature.
case of ICP-AES measurements have Table V, the usefulness of the final
been pointed out recently.259 expression (Eq. T5.7) is that it can be MATRIX EFFECTS AND
We finally note that this approach numerically evaluated for any analyte NORMALIZATION
gives information about the limiting and other similar experimental setups, APPROACHES
noise of the measurement but does not provided that the spectroscopic parame-
identify the type of noise (e.g., shot, ters are known and the electro-optical Definitions and Methodology. Ma-
flicker) present. This can be identified by detection parameters can be evaluated. trix effects are of critical importance in
the slope of the plot of the signal-to- Assuming the values given in the several analytical spectroscopy tech-
noise ratio versus signal. table caption, and taking copper as a test niques, and laser-induced breakdown
Numerical Evaluation of the Abso- element, the signal counts are given by spectroscopy (LIBS) is not an exception.
lute Sensitivity of a LIBS Measure- Eq. T5.7. Assuming that the limiting In fact, it is not uncommon that a
ment. Table V contains a selection of noise of the measurement is signal shot significantly different signal response is
expressions that are needed in order to noise, the weight fraction of Cu calcu- observed at a given spectroscopic tran-
calculate the absolute sensitivity, in lated for a relative standard deviation of sition of an element present at the same
FIG. 6. Illustration of the different outcomes resulting from choosing the concentration units in weight fraction or mole fraction. The top
part of the figure shows the relation between the two units, while the inset highlights the different slopes obtained for one element in
two different matrices and the single slope resulting from the use of mole fractions in the x-axis. The two simulated plots in the lower part of
the figure are meant to highlight the different conclusions reached regarding the presence of matrix effects. See text for discussion.
FIG. 7. Signal normalization approaches in LIBS. Two normalization parameters have been chosen, i.e., the plasma temperature and
the ablated mass, which is proportional to the acoustic signal resulting from the ablation process. [The temperature plot has been taken
from our own data, while the two figures on the left were reproduced, with permission, from Ref. 229, and the two on the right from
Ref. 40, with permission.]
(TOFMS), where the results obtained Whatever the definition of absolute different laboratories and different in-
were barely affected by the sample versus standardless analysis may be, the strumentation, have made the analytical
matrix.272 Notably, chemometric ap- complexity of the problems associated spectroscopic community look at these
proaches were discussed as powerful with the functional dependence between approaches with skepticism. This is well
means of analyzing spectroscopic data signal and concentration, and difficulty summarized in the following (anony-
without using reference spectra.270 of reproducing the same results in mous) statement: ‘‘For many decades,
FIG. 8. The left part of the figure depicts a flowchart showing the essential steps of the calibration-free approach first proposed by Ciucci et
al.273 (see text for discussion). The right part of the figure shows a comparison between the results obtained for several standard Al alloy
samples with CF-LIBS and Monte Carlo-LIBS (measurements obtained at ~0.1 mbar). The results are shown for three spatially resolved
positions (A, D, and G) taken at different points in the axial direction of the plasma. [Reproduced from Ref. 288, with permission.]
spectral lines and accurate detector and sample compositions need still were observed at very early delay
spectral efficiency are not surpris- further scrutiny.281 times (150–200 ns) and should not
ing: more important is the finding (v) iiiOnly singly ionized species are affect the accuracy of CF-LIBS.
that poor knowledge of partition considered in the closure equation. (vi) iiThe technique has been applied to
functions and electron number Doubly ionized species have been double-pulse LIBS,280 although a
density seem to play less impor- predicted and indeed observed for systematic study is still lacking
tant roles. Moreover, the most boron (B) and Fe in LIBS work by here.
relevant aspects of plasma ideality Gaft et al.292 However, these ions (vii) iDespite its limitations, it is a fact
Eq.
Expression Description number
gu e-Eu =kB T gu e-Eu =kB T
I ¼ nA;n V plasma WAu‘ A;n
¼ nplasma V plasma WvA aA;n Au‘ A;n This expression is a simple modification of that presented in our first T6.1
U ðTÞ U ðTÞ review (see Table IV), with the introduction of some new symbols to
emphasize its application to neutrals and ions, and the mole fraction of
the analyte.
IDP nplasma VDP
plasma A A;n
vDP aDP UA;n ðTSP Þ -kEBu T 1 -T 1
Rn ¼ ¼ DP
plasma plasma A A;n U A;n ðT Þ
e DP SP The ratio of the intensities measured for neutral species in a single pulse T6.2
ISP nSP VSP vSP aSP DP and in a double pulse experiment depends upon the total mass ablated,
the excitation energy of the line chosen, and the plasma temperature
obtained in the two operational schemes.
E
IDP nplasma plasma A
VDP U ðTSP Þ -kBj T 1 -T 1
A;i A;i
vDP aDP
Ri ¼ ¼ DP e DP SP The corresponding expression obtained for ionic transitions is identical to T6.3
ISP nSP VSP vSP aSP UA;i ðTDP Þ
plasma plasma A A;i
that for the neutrals, except that the proper superscript is now used and
the excitation energy refers to the ion excited state.
! " #
nplasma
DP
plasma A
VDP vDP A;n A;n
aDP U ðTSP Þ Eu 1 1
lnðRn Þ ¼ ln þ ln A;n A;n - - The usual transformation in a logarithmic plot shows the dependence of T6.4
nplasma
SP
plasma A
VSP vsP aSP U ðTDP Þ kB TDP TSP the slope on the difference between SP- and DP-plasma temperatures,
while the intercept indicates the difference in the ablated mass.
! " #
nplasma
DP
plasma A
VDP vDP A;i A;i
aDP U ðTSP Þ Ej 1 1
lnðRi Þ ¼ ln þ ln A;i A;i - - A similar result is obtained when the logarithmic plot is made using T6.5
nplasma
SP
plasma A
VSP vsP aSP U ðTDP Þ kB TDP TSP ionic transitions.
A
kSaha ðTÞ
aA;i ¼ aA;n These two relations show the definition of the ionization fraction as a T6.6
ne function of the electron number density and the Saha constant (see
3=2
A ðTÞ ¼ ð2pme kB TÞ 2UA;i ðTÞ -EA =k T
kSaha e ion B also Part 1, Table IV1).
h3 UA;n ðTÞ
A;n 1
aDP;SP ¼ A =ne
These two expressions represent the neutral fraction of the analyte atoms T6.7
1 þ kSaha in the plasma (top) and the singly ionized fraction of the analyte atoms
in the plasma.
A;i 1
aDP;SP ¼ T6.8
1 þ ne =kSaha
A
! " #
nplasma
DP
plasma A
VDP vDP A;n
aDP UA;n ðTSP Þ
lnðRi Þ ¼ ln þ ln
A;n U A;n ðT Þ
Using Eq. T6.6 results in this final expression of the log of the T6.9
nplasma
SP
plasma A
VSP vsP aSP DP enhancement for ionic transitions. The plot is made using in the x-axis
A
e
the sum of the ionization energy of the atom and excitation energy of
Ej þ Eion 1 1 3 TDP n
- - þ ln þ ln eSP the ion. The intercept in this case is seen to be a function of the
kB TDP TSP 2 TSP nDP temperature and electron number density, in addition to the ablated
mass.
a
The equations reported in this table are the result of a collaborative effort between the University of Florida (N. Omenetto) and the CNR Pisa, Italy (E. Tognoni, G.
Cristoforetti, and V. Palleschi). The experimental data obtained using the above equations have not yet been submitted for publication.
Definition of the parameters used in Eqs. T6.1–T6.7: I DP,SP = Spectrally integrated (over the line profile) net intensity (counts s-1); gu = Statistical weight of the
A;s
upper level; nA -3
DP;SP = Number density of atoms of the analyte in the plasma (cm ); nDP;SP = Number density of neutral (s = n), singly ionized atoms (s = i) of the
analyte in the plasma (cm-3); nplasma
DP;SP = Total number density of atoms and ions in the plasma (cm-3); = vA A plasma
DP;SP = (n /n )DP,SP = Molar fraction of the analyte in
the plasma; TDP,SP = Plasma excitation temperature (K); neDP;SP = Electron number density (cm-3); kB = Boltzmann constant (eV K-1); h = Planck constant (erg s);
plasma
me = Electron mass (g); EA ion = ionization energy of the analyte (eV); U
A,s
(T) = Partition function of the analyte in ionization stage s; VDP;SP = Plasma volume
(cm3); WDP,SP = Fraction of plasma volume seen by the detector; Eu (eV) = Excitation energy of upper level of the neutral line; Ej (eV) = Excitation energy of upper
level of the ionic line; Ei (eV) = Excitation energy of lower level of the ionic line; Rn = (IDP/ISP)neutral = Intensity ratio for neutral lines; Ri = (IDP/ISP)ionic = Intensity
ratio for ionic lines.
each burst consisting of six individual performance by taking advantage of rial ablation are again discussed as
pulses of 50–80 ns duration, separated multi-pulse operation, while still main- possible causes of signal enhancement
from each other by 15 ls.348 Galbács et taining the advantage of being a com- in multi-pulse LIBS.353 Applications
al.350 developed a micro-LIBS system pact and portable system.350 have been reported in the analysis of
based on a Nd-GGG (gadolinium galli- We note here that one could adopt the steel samples using a vacuum UV
um garnet) laser and produced from 1 to nomenclature used to describe the char- Paschen–Runge spectrometer, with de-
11 laser pulses of 8 to 10 ns duration in a acteristic output of a free-electron laser, tection limits below 10 ppm for the
burst (called shot by the authors) of which consists of macropulses of 5–10 elements C, P, S, Al, Cr, Cu, manganese
100–200 ls: the inter-pulse separation ls duration, formed by a train of micro- (Mn), and molybdenum (Mo).347,349
was 25–50 ls. As the authors point out, pulses of a few picoseconds.346 Other applications include the quantita-
their system improves the analytical Plume re-heating and increased mate- tive analysis of gold alloys,352,354 me-
FIG. 10. Complementary use of LIBS and Raman in the identification of pigments in art works. [Reproduced courtesy of Anastasia
Giakoumaki and Demetrios Anglos (IESL-FORTH and University of Crete, Heraklion, Crete, Greece).]
instruments were used, then hybrid units LIBS was much smaller than that for ficial for both techniques. Raman can
were developed consisting of single Raman.398 Giakoumaki et al.400 assem- identify the presence of quartz in a
laser and a single detector, and more bled a hybrid unit designed to investi- mineral sample, but this is not sufficient
recently new laser types and spectrom- gate samples and objects of cultural to differentiate between different miner-
eters have been implemented in the heritage. Only one laser at 532 nm was als. Using LIBS on the same sample
combination. used and the spectra were taken with a spot as the Raman spectra reveals
Marquardt et al.395 described a fiber- nonintensified CCD. Pigment identifica- significant differences in the elemental
optic probe system for LIBS, Raman tion in paintings represents a typical composition of the minerals, making
point analysis, and Raman imaging application of the combined techniques their differentiation easy. On the other
analysis of particles on soil substrate: (see also Refs. 396 and 397 and Fig. 10). hand, the LIBS spectrum of a copper ore
different lasers were used to generate the More recently, Hoehse et al.401 have reveals the presence of TiO2 from the Ti
plasma and the Raman signal. Sharma et described a two-laser LIBS-Raman mi- spectral lines in the spectrum, but does
al.398 and Wiens et al.399 described the cro-analysis system with a dual-arm not give information about its polymor-
analysis of biogenic and inorganic high-resolution Echelle spectrometer phic forms (anatase, rutile, brookite),
minerals in a standoff configura- and a single nonintensified CCD. The whose differentiation is provided by the
tion.398,399 One laser system was used, system compares well with state-of-the- Raman spectra.401
the Raman excitation being provided by art Raman scanning microscopes, while Enhance Analytical Sensitivity and
a frequency-doubling crystal in front of LIBS provides parts per million (ppm) Detection Power: LIBS-LIF and LA-
the LIBS laser and a super-notch filter in detection limits and large spectral cov- LIF. The diagnostic value of the fluo-
the path of the detector.399 Both spectra erage (290–945 nm).401 The authors rescence technique has been discussed
were taken at the same location (at 10 m illustrate clearly how the combination in detail in our first review and will
distance): the sampling diameter for of LIBS and Raman is mutually bene- therefore not be repeated here. The
FIG. 13. Left: (a) Microscope image of a hexagonal packed micro-lens array with circular micro-lenses of diameter ~0.9 mm. (b) Side-view
and (c) 458 view of a laser-induced breakdown in ambient air generated by a 200 mJ/pulse laser beam focused with a micro-lens array. The
laser beam direction is from the right to the left side of the photographs. Right: Overlay of three consecutive emission spectra of the laser
plasma produced with focusing by (d) a single lens and (e) a micro-lens array in air at a rotating slag sample. The improvement in
reproducibility obtained with the micro-lens is evident. [Reproduced from Ref. 427, with permission.]
and a multichannel spectrometer cou- Several applications using microchip Zn at 213.9 nm).436 Other applications
pled with an ICCD. The laser was and low-energy high-repetition-rate la- include 2D images of latent fingerprints
characterized by an energy per pulse of sers were reported on low alloy steels using 400 nm pulses of 120 fs dura-
4 mJ, which could be decreased with a and aluminum targets.436–441,444 It is tion,443 elemental distribution of paper
diaphragm to a few lJ, giving a spatial interesting to note that self-reversal was coatings,430 topographical and chemical
resolution adjustable from 5 to 15 lm. observed in strong resonance lines (e.g., analysis of a meteorite, a basaltic rock,
FIG. 14. Selected examples of applications of chemometrics to LIBS experimental data. Top right: PCA scores of chemical agent simulants
constructed using spectral regions from (A) single-shot and (B) average spectra.552 [Reproduced with permission from Ref. 552.] Bottom
right: Three-dimensional scatter plot of a biological data set in the principal component coordinates. Each data point represents a single
spectrum acquired with a single laser shot.662 [Reproduced with permission from Ref. 662.] Top left: Principal component analysis (PCA)
plot of PC1 (47%) versus PC2 (26%) scores for carbonate samples (closed symbols) and geological reference standard materials (open
circle).597 [Mineral images: Courtesy of Geology.com; Bacteria image by Bruce Arey/2010 EMSL. Plot reproduced with permission from Ref.
597.] Bottom left: X-loadings for PC1 (47%) and PC2 (26%) by wavelength. The most influential peaks in PC1 belong to Ca(þ) and Mg(-),
while the most influential peaks in PC2 belong to Mg (þ) and Fe (-).597 [Reproduced with permission from Ref. 597.]