Matrix effects in laser ablation molecular isotopic spectrometry

Spectrochimica Acta Part B 101 (2014) 204–212

Contents lists available at ScienceDirect

Spectrochimica Acta Part B journal homepage: www.elsevier.com/locate/sab

Matrix effects in laser ablation molecular isotopic spectrometry Staci Brown a,1, Alan Ford b,⁎, Charlemagne C. Akpovo a, Jorge Martinez a, Lewis Johnson a,2 a b

Department of Physics, Florida A&M University, 2077 E. Paul Dirac Drive, Tallahassee, FL 32310, United States Alakai Defense Systems, 197 Replacement Ave, Suite 102, Fort Leonard Wood, MO 65473, United States

a r t i c l e

i n f o

Article history: Received 24 March 2014 Accepted 30 August 2014 Available online 10 September 2014 Keywords: Laser-ablation molecular isotopic spectrometry LAMIS Matrix effect Boron isotope Isotopic analysis

a b s t r a c t Recently, it has been shown that laser-induced breakdown spectroscopy (LIBS) can be used for the detection of isotopes of elements via isotopic shifts in diatomic species in a technique known as laser ablation molecular isotopic spectrometry (LAMIS). While LAMIS works quite well for isotopic analysis of pure compounds under optimal conditions, it is desirable for it to be applicable for a variety of compounds and matrices. However, the LIBS plasma emission associated with LAMIS depends on several parameters, including the applied electric field of the laser pulse, the physical properties of the material being investigated, and the presence of additional elements other than the element of interest. In this paper, we address some of the pitfalls arising from these dependencies when using LAMIS for the determination of the relative isotopic abundance of boron-containing materials with varying chemical matrices. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Investigation of the isotopic composition of simple molecules dates back to the earliest days of molecular spectroscopy. Mulliken performed isotopic analyses of boron oxide (BO) using analysis of vibronic emission spectra in 1925 [1], while Jenkins et al. performed one of the first relative abundance analyses on the same molecule using both rotational and vibrational features [2]. Other studies of optical spectra of molecular isotopologues such as metal hydrides [3], silicon hydride [4], copper halides [5], iodine monochloride [6], carbon monoxide [7], and many other molecules containing isotopes of lower natural relative abundances were later demonstrated. These later analyses utilized microwave and infrared spectroscopy to obtain rotational and vibrational emission spectra produced from chemical reactions, thermal desorption, electrical breakdown, and a variety of other vaporization/ionization methods of vapors or gases [8]. Although spectroscopists were aware of the molecular isotopic composition influence on the rotational, vibrational, and electronic spectra in these studies, most experiments were not designed to precisely measure the relative abundance of isotopes. Instead, only molecules made of the isotopes of the highest abundances for a given element were studied

⁎ Corresponding author. Tel.: +1 573 329 1919. E-mail addresses: [email protected] (S. Brown), [email protected] (A. Ford), [email protected] (C.C. Akpovo), [email protected] (J. Martinez), [email protected] (L. Johnson). 1 Tel.: +1 773 802 0793. 2 Tel.: +1 850 599 8456.

http://dx.doi.org/10.1016/j.sab.2014.09.003 0584-8547/© 2014 Elsevier B.V. All rights reserved.

without enrichment because the major isotope makes up more than 92% of all atoms for most lighter elements [9]. For the inquiry of lower abundance isotopologue spectra, chemicals were enriched in the lower abundance isotopes to help elucidate spectroscopic parameters such as vibrational force constants or rotational constants [10–12]. However, the 19.9% and 80.1% natural abundances of 10B and 11B and the 76.8% and 24.2% natural abundances of 35Cl and 37Cl occurred in proportions such that multiple isotopes could be studied simultaneously without enrichment or purification. As a result, the ratio of intensities of features in optical spectra of isotopologues of boron and chlorine compounds was sometimes used to justify assignments of transitions, thereby not only showing an early awareness of relating optical spectra back to relative abundances of isotopes but also demonstrating the limitations of utilizing the information from isotopologues below a certain abundance threshold [13,14]. More recently, a method called laser ablation molecular isotopic spectroscopy (LAMIS) seeks to expand this use of optical emission for the determination of the relative abundance of isotopes in a sample via laser-induced breakdown spectroscopy (LIBS) [15–20]. Like LIBS, LAMIS begins with a high-fluence, high-power laser pulse that induces ablation, decomposition, and ionization of a (usually solid) sample. This process initially creates atoms and elemental ions from a target [21], but LAMIS avoids looking at these species because the energy separation between isotopes from atomic emission is often smaller than 10 GHz, which is narrower than the broadening caused from the strong electric fields of the LIBS plasma [22,23]. Instead, the technique focuses on molecules that eventually materialize through recombination processes under favorable plasma conditions [24]. Once molecules form, the analysis of rovibronic bands of isotopes of diatomic molecules in

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212

the plasma that can have vibrational isotopic shifts of 100 GHz or more resolves the relative abundance of isotopes in a solid sample [25]. Due to technological improvements to LIBS-associated equipment over the past decade, the LAMIS technique has the potential to supplant commonly utilized technologies for isotopic analysis [16,20]. Current methods for abundance measurements include but are not limited to accelerator mass spectrometry (AMS), multiple collector inductively coupled plasma mass spectrometry (MC-ICP-MS), gas source mass spectrometry and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). However, these systems tend to be prodigious, complicated, and time-consuming. The availability of compact lasers and spectrometers now means that a LIBS instrument, including the power supply, can be small enough to be field-portable for in-place, non-contact sampling [26,27]. LIBS also has short analysis times, taking less than a second, as well as having simple experimental setups [28,29]. It requires no expendables and could function several tens of meters from a target in almost any phase [30,31]. Very few techniques can compete on this last point, which may be very important if potentially radioactive samples are to be probed for isotopic information with the LAMIS technique from a safe distance. Because of this ability for non-contact, standoff sampling, the advantages of LAMIS are most capably demonstrated for the analysis of nonvolatile, pure solids, but careful consideration of both the material being analyzed for a given instrument setup is still paramount to unlocking the full potential of this technique.. Laser-induced breakdown can produce many diatomic species suitable for isotopic analyses [32], though the amount of material that is ablated or entrained into a plasma depends not only on the applied electric field from the laser but also on the physical properties of the material under study [33–36]. Any additional elements that are present in a plasma other than the element of interest – including additional elements in pure compounds – can affect the extent of ablation and quality of spectra [37,38]. These additional elements potentially obscure or aid in the relative abundance assessment for a given element [39,40]. Collectively, the elements not being analyzed form a sort of matrix that can include more than just the elements in the starting compound. For example, the surrounding atmosphere can also contribute atoms to the plasma, and these atoms are necessary for LAMIS to be applicable in some cases [15,41]. The total amount of ablated sample, chemical composition, density, thermal conductivity, and other chemical and physical properties of a material results in similar issues for isotopic analysis as well. Changes to the signal-to-noise and signal-to-background ratio lead to inevitable adjustments to instrument settings for a given sample to ensure that molecular spectra are observed [42,43]. Addressing these potential problems is a necessary step in helping LAMIS become a more robust technique for the analysis of the relative abundance of isotopes in a sample. It should be noted that some effort has been devoted to make LAMIS applicable to multiple materials for isotopic analysis of the same element. The original LAMIS work demonstrated that emission from 13 C-enriched urea and natural-abundance graphite showed good separation in the C2 and CN LIBS spectra, but relative abundance analyses were not published on these materials [15]. In addition, naturalabundance boron nitride and isotopically-enriched boric acid samples were used in a follow-up publication where the isotopic composition of boron in a natural abundance sample was determined with an error of about 3.7% for 11B using a partial-least squares PLS model [16]. Even lower limits of detection of 9‰ (2σ) for 10B were achieved using improved pretreatment algorithms applied to LAMIS of BN and B2O3 [20]. Similar analyses have been carried out on boron carbide (B4C) samples that improved upon the first LAMIS analyses by demonstrating slightly lower limits of detection of 10B around 1% with femtosecond lasers to reduce continuum [18,19]. Although these works represent some extension of the method to different matrices, training a PLS model on the same or extremely similar matrices that are to be analyzed does not necessarily demonstrate that the method is capable of dealing with matrix effects if the matrix is not known. Furthermore, these analyses were

205

not performed in the same spectroscopic region as previous LAMIS work; instead, a different electronic transition of BO in the visible portion of the spectrum was used. Other LAMIS work extracted strontium isotopic information from different halide salts of that element, but the same diatomic was not analyzed throughout [17]. Rather, the analyses of the emission from different strontium halide diatomics rather than strontium monoxide, which was produced through chemical reactions for each halide material, were demonstrated. This analysis illustrates the ability to extract isotopic information from different diatomics and shows the production of the same diatomic from different matrices, but it does not provide understanding of how the matrix can affect the analysis of a given diatomic. Though using different regions of a molecular spectrum, deriving alternative modeling techniques, or extracting information from different molecules for similar substrates does not invalidate the application of LAMIS to different matrices, it does preclude comparison of the previous calibration-free methods to more empirical ones for the analysis of a given molecule as the matrix is changed. Moreover, calibration-free models can be used to deduce the influence of the matrix on plasma properties, such as the local thermodynamic equilibrium (LTE) temperature, which are integral to understanding plasma behavior under a given set of experimental conditions. Because no LAMIS work has yet been published on matrices that do not contain the element of interest (e.g., a boron LAMIS analysis on a mixture, of which some compounds contain no boron), it is reasonable to assume that these calibration free models may become more important as LAMIS is used on samples that are not chemically pure. To extend the previous LAMIS studies with regard to matrix effects, the determination of the relative abundance of boron isotopes obtained from a study of BO emission spectra of multiple compounds and mixtures was examined in this work. LIBS spectra of boron oxide originating from different compounds, such as boron nitride (BN), boric acid (H3BO3), borax (Na2B4O7·10H2O), anhydrous borax (Na2B4O7), lanthanum borate (LaBO3), and elemental boron (B) were acquired. In addition, LIBS was performed on impure B-containing samples where all ablated materials did not contain boron. Each of these materials had different physical properties and elemental composition in order to test possible matrix effects for the LAMIS method. A calibration-free model similar to that of the original LAMIS work was used to determine properties of the plasma as the matrix was changed. For each compound or matrix, the effects of elements other than boron and the physical properties of the sample on the efficacy of LAMIS to determine the relative abundance of boron under the same experimental sampling conditions were examined. 2. Methods and materials An overview of the experimental setup for this work using LIBS is shown in Fig. 1. A Q-switched Nd3 +:YAG laser (Quantel Brilliant B) operating at 532 nm at either 20 mJ or 100 mJ of output energy with ~5 ns pulse duration was used. The beam was directed, expanded, and collimated using a set of mirrors and lenses, and before being focused using an achromatic lens (150 mm focal length) to a spot size of 1 mm at the sample surface. A photodiode connected to an oscilloscope was used to synchronize and monitor plasma onset and confirm detector settings (e.g., gate width and gate delay) after the onset of the nanosecond laser induced plasma. One mid-size fiber optic collimator with a focal length of 18 mm (Fiberguide Industries) was positioned orthogonal to the plasma to collect and direct light into an Acton Spectropro 750i (3600 gr/mm, 240 nm blazed grating) spectrometer with an Andor iStar 740 intensified charge-coupled device (ICCD) camera that was used for recording spectra. Before collecting spectra, the wavelengths of this spectrometer were offset corrected with an HG-1 (Ocean Optics) line source using a transition near 253.7 nm. To ensure that all samples maintained the same lens to sample distance, a helium–neon laser line level was used.

206

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212

Fig. 1. LIBS experimental setup for the analysis of boron oxide spectra. In the figure, the symbols are as follows: A = laser, C = computer, D = photodiode, E = beam expander, F = fiber optic collimating lens, I = iCCD camera, L = lens, M = mirror, O = oscilloscope, S = spectrometer, and T = digital-delay generator. The digital-delay generator also times the laser through a laser-control box connected to the laser by an umbilical, neither of which is shown in the figure.

Spectra of the BO B2Σ+ → Χ2Σ+ transition were collected from 255 to 265 nm with this instrument. In addition to this spectral region being the first suggested for BO analysis by LAMIS, this region was chosen because it is much less congested than the A → X band at longer wavelengths, and it occurs in the deep UV where little sunlight is present. Avoiding interference from solar or external sources is essential for LAMIS to be utilized outdoors with consideration to the collection time of BO spectra, since the time for collection of the BO spectra could be sufficiently long to involve solar interference. This region synchronizes well with the envisioned application of LAMIS as a rapid, noncontact or standoff diagnostic for nuclear forensics and national security applications. Furthermore, this region represented a tradeoff between the capabilities of the available equipment and the relative maximum signal strength of BO emission in the deep-UV with current optics; BO emission at shorter wavelengths was stronger, but also contained strong rotational transitions of high J from the Δv = 0 band or emission from B(II) at short gate delays, which could have complicated analyses with the algorithms mentioned later.

Collection of time-resolved LIBS emission of BO from a BN disk partitioned from a BN rod (McMaster-Carr, 25 mm diameter, 150 mm length) showed consistent and near optimal intensities of BO in the chosen wavelength region using a 35 μs gate width after a gate delay of 25 μs, as seen in Fig. 2. Three sets of LIBS spectra in the region of interest for BO produced from BN were collected at different gate delays for each energy setting with a gate width of 35 μs, which was chosen to provide an adequate temporal window over which to capture much of the BO emission for different materials. These spectra had their minima subtracted and were then normalized to remove effects from overall emission, which involved continuum at gate delays shorter than a few microseconds. Next, a portion of the 11BO B2Σ+ → Χ2Σ+ (Δv = − 2) band head was summed together for each spectrum and these sums were then averaged to produce a mean intensity at a given gate delay, which were fit with a spline to produce a smooth curve. As seen in Fig. 2, the smoothed curves plateaued around 20 μs for the 20 mJ setting and began a slow descent at the same time for the 100 mJ setting. Moreover, gate delays shorter than about 20 μs were believed to produce

Fig. 2. Plots of the integrated area for a region containing the 11BO band head obtained from baselined and area-normalized LIBS spectra of BN as a function of gate delay of the iCCD for both laser energies used. Each point is the mean area of three locations with 30 shots at a given location for a given gate delay. Smooth curves represent a 5-knot spline fit between the points.

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212

contaminated spectra from elemental emission [16]. Therefore, a slightly longer gate delay of 25 μs was used for both laser energies to keep the experimental setup simple and to avoid as much contamination as possible from species other than BO, though BO spectra generated from some of the materials discussed later did show elemental emission at this gate delay. Data was collected on different natural-abundance, boroncontaining materials following the initial optimization of experimental settings. Commercial grade boric acid (99% H3BO3), common borax laundry detergent additive (99 + % Na2B4O7·10H2O), anhydrous borax (≥ 99% Na2B4O7, Sigma-Aldrich #71997), lanthanum borate (99% LaBO3, Sigma-Aldrich, #772887), and boron metal (99.7% B, Sigma-Aldrich #266620) were each analyzed under the same conditions at the two different laser energy settings. Except for the elemental boron, which were small ~ 0.25 cm3 irregularly-shaped chunks rather than a powder, samples were made by pelletizing 0.5 to 1.0 g of material without a binding agent under approximately 160 MPa for 30 min in a hydraulic pellet compressor (Carver, Inc., USA). The resulting pellets had an average diameter of 13 mm and an average thickness of 2.41 to 4.71 mm. Spectra were taken at 50 different locations on each pellet, with each location being analyzed 30 times under identical instrument settings. Collection took approximately 1 h for all shots, minimizing instrument drift. The isotopic composition analysis of BO in a sample was performed by a calibration-free model where experimental spectra were fit via least-squares to theoretical rovibronic emission spectra. This type of model was documented in several past works, including the original LAMIS paper, and consisted of a number of steps that mostly involved formation of the theoretical energies and relative intensities of transitions from the predominant BO isotopologues (i.e., 10B16O or 11B16O) [15,24,44,45]. Once the transition energies and relative intensities were known, spectra of each isotopologue were convoluted with an instrument response function. This function broadened the transitions such that when they were vertically scaled and added together, they produced a result resembling the experimental rovibronic BO spectra. Calculating the theoretical energies of the BO molecular transitions required predetermined constants taken from the literature [46–50]. These values were derived from an approximated molecular Hamiltonian that estimated the electronic, vibrational, and rotational energies of the BO molecules as a function of user-supplied quantum numbers. The constants used for this work came from Mélen et al. [51], who did an extensive analysis of the A → X and B → X transitions of BO. Once these constants were known, energy levels arising from different electronic states (labeled as A, B, or X in this case) were calculated. Transition energies between different electronic states were then determined based on selection rules corresponding to the total spin (S) and

207

molecular orbital angular momentum (Λ), which were also available in the literature. Intensities for the theoretical spectra were determined from a similar set of equations and selection rules. Both the B and X electronic states of BO were 2∑ states and followed the transition strengths for rovibronic spectra found in Ochkin [52]. Franck–Condon factors that determined vibrational transition probabilities between the two electronic states were taken from Liszt et al. [53] Other effects on the intensities, such as the r-centroid contribution altering the electronic–vibrational strengths, were neglected because they were deemed too insignificant to impact the results of this work [54]. Line shapes were assumed to be Voigt-shaped due to the presence of pressure, Stark, and Doppler broadening convolved with the spectrometer-broadening function [55–59]. For this work, the Voigt function was assumed to be approximated as a linear combination of Gaussian and Lorentzian line shapes. Each transition was assumed to have varying center wavelength and intensity parameters, but constant full-width at half maximum (FWHM), along with constant Gaussian and Lorentzian contributions. For molecular transitions under the experimental conditions in this work, the use of a constant FWHM was acceptable, as FWHM differences were unlikely to be resolvable with the equipment used. By simulating spectra with different peak widths through Monte Carlo comparisons to experimental spectra, a FWHM of 4.1 cm−1 for the line shape was determined. Once the relative transition intensities and line shapes were calculated, initial fitting parameters had to be set. For this work, the plasma temperature and relative intensity scale parameters for the spectra of the isotopologues, which encapsulated the number densities of the isotopologues, were used. The plasma temperature parameter was known to be between 2500 K and 10 000 K from plasma emission of molecules in previous studies, both published and internal [60,61]. It was not known if the plasmas were in LTE. Regardless, simulations using separate temperatures for electronic, rotational, and vibrational contributions would have likely caused over-fitting, and individual temperature values for these contributions are difficult to determine due to the speculated formation of BO through poorly understood chemical reactions. Thus, a single average parameter for the temperature was most appropriate [62]. The intensity of the offset-corrected experimental spectrum compared favorably to the simulated spectrum generated by setting concentrations to natural abundance BO (80.1% 11BO and 19.9% 10 BO) at 5000 K, which mirrored the temperature used in the original LAMIS work [15]. It should be noted that 16O was the only isotope of oxygen in the simulations in order to limit the number of parameters, as other isotopes make up less than 0.25% of naturally occurring oxygen. The least-squares fits were performed using custom software written in Matlab [63,64], which was accessible through the graphical

Fig. 3. The graphical interface of the least-squares fitting program used to fit the theoretical spectra to experimental ones.

208

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212 5

4

10

x 10

3

x 10

9 2.5 8 7

Intensity (a.u.)

Intensity (a.u.)

2 6 Na B O 2 4

7

5 4

LaBO

3

1.5

1 3

Na2B4O7•10H2O

BN

2 0.5 1 0

B

H BO 3

256

3

258

260

262

0

256

Wavelength (nm)

258

260

262

Wavelength (nm)

Fig. 4. Mean LIBS spectra for each of the pure compounds studied in this work at 20 mJ laser energy. Spectra have been offset for clarity. The edge pixels of these digitized spectra (λ b 255.1 and λ N 263.9 nm) represent unintensified light and dark current from the CCD rather than signal from LIB. The apparent atomic transitions in the LaBO3 spectrum have been clipped on the y-axis to focus on weaker BO emission that may have been present.

interface seen in Fig. 3. As in the original LAMIS work [15], only the portion of the spectra between 255.1 nm and 256.8 nm representing the Δv = −2 transitions were used for these fits in order to avoid possible errors arising from uncertainty in the rotational distortion constants. Data over this region was checked for peak alignment to the theoretical energies for fitting and was shifted and scaled along the x-axis until a reasonable match to theoretical peak positions was made. The intensity scale parameters of the two isotopologues of BO, the LTE temperature, and the offset of the spectra were then adjusted until a good match was obtained to the experimental spectra over this reduced spectral region. At the end of the fits, a ratio of the intensity scale parameters to their sums for 11BO and 10BO determined the relative abundance of boron in the starting material. The software for fitting the spectra and example models for repeating the least-squares fits here is planned for release in the near future so that others may benefit from this tool for LAMIS analysis. 3. Results and discussion To maximize the likelihood of getting ample BO emission, the lower laser energy of 20 mJ was used first. Optimal settings for the instrument were unknown, but the 20 mJ of laser energy was sufficient for BO formation judging from the preliminary studies on the BN disk from McMaster-Carr. Mean LIBS spectra of BO emission from this effort were plotted in Fig. 4, which revealed the typical BO emission of the 1500 total transients collected from each material. The spectra in Fig. 4 showed that each of the tested materials gave rise to different signal strengths for the same number of transients. The B, BN, H3BO3, Na2B4O7·10H2O, and Na2B4O7 all had similar signals, which indicates that the amount of oxygen and boron in the sample relative to other elements does not drastically affect production of BO at the gate delays used in this work. This result was significant, since the relative percent of boron by mass and by number density, ΧB, in each material varied, as shown in Table 1. For a given laser shot, elemental B was expected to generate the most elemental boron in the plasma and borax or LaBO3 the least. This assumption was based on the chemically pure solid matrix of B requiring less energy per atom/molecule to be consumed in the ablation process overall. Unfortunately, the B sample had irregular edges that caused the laser focus to change slightly as different spots were selected for study, meaning that the compound

with the strongest average signal using similar reasoning should be BN. BN was the densest compound with the fewest ancillary atoms attached to B. Also, because BO was generated by reactions in the plasma rather than from fragment production during the ablation, the oxygen contained in most of the compounds was of little consequence in the emission of BO. Indeed, BN was observed to have the strongest BO emission signals. A mysterious feature appeared near 259.3 nm for both the Na2B4O7 materials under study. This feature was not identified as an atomic transition due to its width at delay times that preclude Stark broadening. In all likelihood, it was a molecular emission from an undetermined species in the plasma, and its exact identification was not obvious from available data without performing a series of time-consuming theoretical simulations. However, its influence on the determination of the isotopologues was deemed minimal, since it was over 2 nm outside the spectral window used for the analyses here. Analyses from the least-squares fits on the LIBS spectra averaged from each location for each material yielded both relative abundance of 10B and plasma temperature information. The average and standard deviation of these parameters for a given spot on a material were assembled into Table 2. In addition, the area underneath the fitted experimental spectrum was determined to provide an idea of the intensity variation of the collected spectra. A typical example of the match between the fit output and the experiment of the LS-fitting procedure was similar to that in Fig. 5, which shows a typical BO spectrum produced from boric acid plotted against the fitted, theoretical BO emission. Matches between the experimental and predicted intensities generally agreed quite well, but there were some variations in between the two. These variations were believed to be due to rovibronic couplings that were not included in the Table 1 Relative amounts of boron in each compound studied in this work. Boron material

%B (w/w)

ΧB

B (elemental) BN H3BO3 LaBO3 Na2B4O7 Na2B4O7·10H2O

100 44 17 6 9 5

1.00 0.500 0.143 0.200 0.308 0.0930

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212 Table 2 Least-squares fitted parameters from libs using 20 mJ laser energy. Boron material

10

B (elemental) BN H3BO3 LaBO3 Na2B4O7 Na2B4O7·10H2O

20.4 (2.4) 19.4 (1.5) 20.0 (4.9) 76.3 (3.4) 22.1 (1.9) 19.5 (4.5)

a b

B abundance (%)a

TLTE (K)a

Area × 10−7 (a.u.)a

5120 (390) 4640 (270) 4920 (780) 10 000 (0)b 5380 (310) 4740 (660)

1.731 (0.244) 2.097 (0.117) 1.287 (0.055) 5.091 (0.485) 1.987 (0.091) 1.285 (0.028)

Numbers in parentheses are standard deviation. The fitting algorithm reached a boundary condition for the LTE for this material.

simulations or from small amounts of impurities in the air, pellet press, or other materials that contacted the powdered samples in the process of creating the pellets. Regardless, the fitting of the contoured spectra at lower resolutions and even poorer matching to theory than obtained here was already shown to enable the extracting of acceptable relative abundances for boron compounds in earlier LAMIS results [16]. Though better matching to theory compared to previous LAMIS work was available, it was noted before data were collected that the LAMIS approach would probably not be able to predict the concentration better than about 1% with the calibration-free approach, as confirmed here. It was technically impossible to predict the correct percentage of isotopologues to better than 0.25% due to omission of isotopes of oxygen in the BO simulations. More importantly, the uncertainty in the relative abundance of 11B in natural sources of boron was known to vary by more than 1% around the accepted value of 80.1%; a similar error was expected for 10B, since it makes up most of the remainder of the natural abundance of boron at 19.9% [65]. Furthermore, data for this work was not performed on calibrated reference standards because it was assumed early on that the accuracy of the calibration-free models was only on the order of 1 to 2% at best due to fluctuations in temperature and shot-to-shot variation that generally occurs with LIBS, as demonstrated by Table 2 results. These results also suggested that the matrix was a minimal factor in the determination of the relative abundance of boron in a sample in most cases, except for LaBO3, which will be discussed later. The calculated relative abundance of 10B was approximately what was expected from the known natural abundance, though the H3BO3 and Na2B4O7·10H2O showed higher than the expected variability. The reason for this variability was unclear. One potential issue could have been sampling effects, attributed to the density of the pellets deviating

209

between compounds, which was a topic set aside for possible inclusion in a forthcoming paper. It should be noted that some depth profiling, in which a location was previously ablated with shots not included in the accumulated LIBS spectrum at a given spot, was performed on some of the samples. Accumulations of 30 shots, each at a different spot on the pellet, were also acquired on pellets of H3BO3 and BN. No obvious differences in the BO emission in these spectra were observed for either sampling method that was outside the typical variation of the datasets using the experimental conditions of Table 2. Strangely, LAMIS of LaBO3 had very poor accuracy, even to the point that the mean was not within at least one standard deviation of the known 10B abundance. The mean was grotesquely off because of what appeared to be little BO emission, as well as an interfering transition near the 10BO band head that caused a much higher than expected %10B. The rovibrational band heads carried most of the weight for the predictive capabilities, since they were the regions highest above the noise with the purest contributions from the individual isotopologues. There appeared to be several atomic emission lines in the spectrum as well, though it was difficult to identify the majority of them. The appearance of atomic transitions at the gate delays used in this work was intriguing, especially since they did not show up in the other samples to any appreciable degree. Only ionized species of La were known to emit in the wavelength range used here. As a result, it was concluded that these were not atomic emission from La, but perhaps emissions from unknown impurities in the sample. The lack of BO emission and these additional unknown transitions prompted an early exit of the fitting routine, as boundary conditions for the concentrations – initially set to 20% 10B but allowed to vary by 20% – and a temperature of 10 000 K were reached during the fit. Moreover, picking a different chemometric algorithm would likely not have yielded significantly better results for LaBO3, albeit a larger signal-to-noise in the non-interfering portions of the spectrum, as well as exclusion of the BO-interfering transitions, may have produced slightly more accurate isotopic ratios. To further understand the matrix effects and investigate possible spectral variation from the amount of laser energy, an analysis was performed on all the samples using a laser energy of 100 mJ. Under these conditions, it was expected that BO would possibly form later in the plasma lifetime, since the plasma should stay warmer over a longer period as the laser energy is raised. Though a warmer plasma may have favored the appearance of more non-molecular species later in the plasma lifetime, it was believed that the larger amount of ablated material would increase the total emission of BO. As shown in Fig. 2, earlier

Fig. 5. The intensity versus the wavelength for a fit of the simulated theoretical spectra of 11B16O and 10B16O to a typical experimental BO spectrum obtained from H3BO3. The simulated fit represents a mixture of 79.9% 11B16O and 20.1% 10B16O.

210

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212

Table 3 Least-squares fitted parameters from LIBS using 100 mJ laser energy. Boron material

10

B (elemental) BN H3BO3 LaBO3 Na2B4O7 Na2B4O7·10H2O

23.8 (2.4) 20.1 (1.4) 19.6 (3.6) 52.6 (8.9) 22.3 (1.3) 20.4 (2.2)

a

B abundance (%)a

TLTE (K)a

Area × 10−7 (a.u.)a

5530 (330) 4670 (200) 4550 (540) 1980 (1830) 5310 (160) 5090 (250)

5.666 (1.460) 9.298 (2.629) 4.811 (1.454) 39.35 (11.61) 19.80 (8.72) 6.736 (1.964)

Numbers in parentheses are standard deviation.

time-resolved studies on BN showed strong BO intensities over a longer period relative to the area-normalized intensity at the same iCCD timings as when 20 mJ of energy was used. As well, the gate width was sufficiently large to obtain most of the molecular emissions for both laser energies, so no changes were made to the gate delay or gate width of the iCCD for the 100 mJ setting to facilitate comparison to the 20 mJ setting. Table 3 shows the results of the least-squares fitting when the higher laser energy was used. Comparing the area under the fitted portion to Table 2, it was obvious that the total emission was generally higher when 100 mJ was used, as also seen in Fig. 6. Additionally, the BO emission was still clearly visible and generally free of atomic emission in the region of interest for all spectra except LaBO3, which again produced rather unique spectra. Using this higher energy, the mean %10B for all materials was within the experimental errors from previous LAMIS work [16], except for B and LaBO3. Elemental B was not in a pellet form, and it was likely the irregular surface of the material (i.e., sampling effects) causing varying alignment that led to the large deviation from the natural abundances. Even so, the variances in the abundances were lower or the same for all materials, ignoring LaBO3, which suggests that increasing the laser energy may produce more repeatable analyses. Consideration of the data in both Tables 2 and 3 also suggested that there was a slight matrix effect that appeared when the laser energy was not necessarily optimal for BO emission for the chosen spectrometer settings. It was understandable that every possible boron-containing material would not produce exactly the same volume of plasma with exactly the same composition for a given set of laser parameters. Some materials were likely not ablated as much per laser pulse, and

6

5

16

others possibly caused the plasma to be quenched slightly faster, leading to errors in the relative abundance. Here, the effect was subtle for these pure materials, barring LaBO3, so that as long as BO was visually obvious in the spectrum, a good relative abundance determination was obtained for pure materials. However, some compounds such as LaBO3 contained features that appeared to be atomic or ionic rather than molecular even for the reduced energy setting, so that caution should be exercised when applying LAMIS to new materials. The sampling effects that caused some deviation from the expected relative abundances for elemental boron must also be addressed when performing LAMIS. Finally, spectra of boron-containing materials mixed with those not containing boron were also collected to see if BO was formed in observable amounts under the same experimental conditions for pure materials at the 100 mJ energy setting. For this effort, LIBS data were collected from single pellets of a 50% w/w mixture of boric acid with cellulose powder to simulate boron compounds on organics or from a 50% w/w Na2B4O7 mixed with aluminum silicate (2Al2O3·3SiO2) to simulate boron-containing rocks; LIBS data of Pyrex glass (~13% w/w boric oxide, B2O3) was also collected. As seen in Fig. 7, only the mixture of boric acid with cellulose produced any BO, which was surprising considering that it has about the same weight percent of boron as Pyrex and the lowest mole fraction of boron of all the materials studied. In fact, very little BO was produced on the silicate materials even at long gate delays, though strong emissions near 257 nm – later verified as aluminum (Al) and to a lesser extent, silicon (Si) by comparison with spectra of pure Al and Si respectively – were readily observable. An isotopic analysis on the silicate materials was not performed, but analysis on the mixture of boric acid with cellulose yielded a %10B nearly identical to pure boric acid within experimental error. It should be noted that the lack of BO emission for these mixtures was not due to timings of the spectrometer. In an effort to see emission from BO on the silicate materials, a time-resolved study was performed by changing the gate delay of the iCCD in stepwise until the emission peaks of the strong aluminum lines were just visible. Throughout this process, any BO that was produced was buried under noise. Some depth profiling was also performed. At no point was the emission from BO visible for reasons that are not understood at the moment. Explaining the missing BO spectra from the silicate mixtures will require a much more rigorous analysis, but it is obvious that there are

2.5

x 10

x 10

14 2 12 LaBO

3

10

1.5

Intensity (a.u.)

Intensity (a.u.)

H3BO3

8

6

1

Na B O •10H O 2 4

7

2

4

BN

0.5 2

0

Na B O 2 4

B

7

0 256

258

260

Wavelength (nm)

262

256

258

260

262

Wavelength (nm)

Fig. 6. Mean LIBS spectra for each of the pure compounds studied in this work at 100 mJ laser energy. Spectra have been offset for clarity. As before, the edge pixels of these digitized spectra (λ b 255.1 and λ N 263.9 nm) represent unintensified light and dark current from the CCD rather than signal from LIB.

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212 Al

211

Al

Si

Na B O + 3Al O •2SiO 2 4 7 2 3 2

H3BO3 + cellulose

Pyrex

255

256

257

258

259

260

261

262

263

Wavelength (nm) Fig. 7. Spectra of mixtures of boron-containing compounds with those that do not contain boron. Each mixture spectrum is a sum of 5 transients collected on each material. Spectra are offset for clarity.

effects from the matrices. The silicates used here were very dense; Pyrex was a uniform glass, and the Na2B4O7 with 2Al2O3·3SiO2 pellets were only about a third as thick as the best BN pellets and resembled thin pieces of brittle, but strong ceramic. The intrinsic density, or packing density, may have had some effect on the emission. The presence of silicon and other elements that readily formed oxides, such as SiO or AlO, may also have reduced the emission of BO by consuming more oxygen from the plasma. Another possibility is that these elements could have altered the plasma conditions such that the reactions for producing the BO could not occur. For example, most of the ablated H3BO3 and cellulose likely burned rather than forming a plasma, making the reactions of this mixture much different than those of the relatively incombustible silicates. The matrix may have also affected the amount of ablated material per pulse, changing the delay times at which molecules could be observed. All these theories were under investigation at the time of this work and will be published later as data becomes available. 4. Conclusion In this work, the LAMIS technique for the determination of the relative abundance of isotopes of boron in different materials was tested. A total of 50 LIBS spectra, and 30 averaged transients of elemental B, BN, H3BO3, Na2B4O7, Na2B4O7·10H2O, and LaBO3 were collected for each material at two different laser energies and analyzed for their %10B content relative to total boron using a calibration-free least-squares model that fitted the Δv = −2 band of the BO LIBS emission to corresponding theoretical BO spectra of the 11B16O and 10B16O isotopologues. All spectra were taken under the same experimental conditions designed to maximize the BO signal when BN was used. In addition, spectra were collected over a relatively short period of time to minimize effects from instrumental variation of the LIBS signal. The analysis here revealed some matrix effects on the determination of %10B content when different pure compounds were used. Of the materials tested, BN had results that were closest to the natural abundance with the smallest amount of %10B variation, possibly because the experimental conditions were designed to optimize BO emission for this compound. Materials other than LaBO3 all provided a reasonable average relative abundance for 10B, though the standard deviation around the mean was higher for H3BO3 and Na2B4O7·10H2O when the lower laser energy was used. Results for LaBO3 were not within the known natural abundance within one standard deviation and were considered too poor to be useful for a reliable abundance analysis for the experimental settings used. Using different laser energies revealed a subtle matrix effect for different pure compounds other than LaBO3. The lower energy showed lower overall emission that presumably affected the algorithm for

determining the natural abundance. This effect was manifested in more variation in the relative abundance determination of the two boron isotopes. While the effect here was small, there may be cases were changes to the laser energy can magnify the effect without precisely locating the temporal window in which BO is optimally formed for a given compound. Finally, LIBS spectra were taken of mixtures of compounds that contained boron with materials that did not contain boron. Although a thorough analysis of these spectra was not performed, it was clear that the LAMIS technique would not have yielded accurate relative abundance information of the BO isotopologues in the silicate mixtures used here. These samples were somewhat dense materials with atomic emissions overlapping the BO emission, though the overlap was not as important as the fact that very little BO was observed. As with some of the previous materials studied, these spectra indicated that matrix effects were a significant issue for the LAMIS technique unless experimental conditions can be optimized to provide strong signal-to-background spectra of BO. Acknowledgments The authors wish to thank the Defense Threat Reduction Administration (DTRA) of the U.S. Department of Defense for their support for this work under contract HDTRA1-12-1-0036. References [1] R.S. Mulliken, The isotope effect in band spectra, II: the spectrum of boron monoxide, Phys. Rev. 25 (1925) 259–294. [2] F. Jenkins, A. McKellar, Mass ratio of the boron isotopes from the spectrum of BO, Phys. Rev. 42 (1932) 464–486. [3] R.S. Mulliken, The isotope effect as a means of identifying the emitters of band spectra: application to the bands of the metal hydrides, Nature 113 (1924) 489–490. [4] R.S. Mulliken, The isotope effect in the spectrum of silicon nitride, Nature 116 (1925) 14. [5] R.S. Mulliken, The isotope effect in band spectra, III. The spectrum of copper iodide as excited by active nitrogen, Phys. Rev. 26 (1925) 1–32. [6] C. Townes, F. Merritt, B. Wright, The pure rotational spectrum of ICl, Phys. Rev. 73 (1948) 1334–1337. [7] O. Gilliam, C. Johnson, W. Gordy, Microwave spectroscopy in the region from two to three millimeters, Phys. Rev. 78 (1950) 140–144. [8] K.P. Huber, G. Herzberg, Constants of Diatomic Molecules, Litton Educational Publishing, New York, 1979. [9] J.K. Böhlke, Isotopic compositions of the elements, 2001, J. Phys. Chem. Ref. Data 34 (2005) 57. [10] G. Herzberg, A. Lagerqvist, A new spectrum associated with diatomic carbon, Can. J. Phys. 46 (1968) 2363–2373. [11] W. Ketterle, H. Figger, H. Walther, Emission spectra of bound helium hydride, Phys. Rev. Lett. 55 (1985) 2941–2944. [12] D.E. Tevault, R.R. Smardzewski, Matrix reactions of chlorine atoms with NO2 molecules, J. Chem. Phys. 67 (1977) 3777.

212

S. Brown et al. / Spectrochimica Acta Part B 101 (2014) 204–212

[13] M. Lorenz, J. Agreiter, A.M. Smith, V.E. Bondybey, Electronic structure of diatomic boron nitride, J. Chem. Phys. 104 (1996) 3143–3146. [14] W. Ling, Y. Xiao-Hua, C. Yang-Qin, Isotopic effect of Cl+ 2 rovibronic spectra in the A–X system, Chin. Phys. B 18 (2009) 2724–2728. [15] R.E. Russo, A.A. Bol'shakov, X. Mao, C.P. McKay, D.L. Perry, O. Sorkhabi, Laser ablation molecular isotopic spectrometry, Spectrochim. Acta Part B 66 (2011) 99–104. [16] X. Mao, A.A. Bol'shakov, D.L. Perry, O. Sorkhabi, R.E. Russo, Laser ablation molecular isotopic spectrometry: parameter influence on boron isotope measurements, Spectrochim. Acta Part B 66 (2011) 604–609. [17] X. Mao, A.A. Bol'shakov, I. Choi, C.P. McKay, D.L. Perry, O. Sorkhabi, et al., Laser ablation molecular isotopic spectrometry: strontium and its isotopes, Spectrochim. Acta Part B 66 (2011) 767–775. [18] B. Yee, K.C. Hartig, P. Ko, J. McNutt, I. Jovanovic, Measurement of boron isotopic ratio with non-gated molecular spectroscopy of femtosecond laser-produced plasma, Spectrochim. Acta Part B 79–80 (2013) 72–76. [19] P. Ko, I. Jovanovic, Boron isotopic measurements from spectrally filtered non-gated molecular spectra induced by laser ablation, Spectrochim. Acta Part B 90 (2013) 68–71. [20] A. Sarkar, X. Mao, R.E. Russo, Advancing the analytical capabilities of laser ablation molecular isotopic spectrometry for boron isotopic analysis, Spectrochim. Acta Part B 92 (2014) 42–50. [21] D.A. Cremers, L.J. Radziemski, History and fundamentals of LIBS, in: A.W. Miziolek, V. Palleschi, I. Schechter (Eds.), Laser Induc. Break. Spectrosc, Cambridge University Press, Cambridge, 2006, pp. 1–16. [22] W.H. King, The estimation of mass shifts in optical isotope shifts, J. Phys. B At. Mol. Phys. 4 (1971) 288–295. [23] D.A. Jackson, Isotope shifts in visible lines of the arc spectrum of krypton, J. Opt. Soc. Am. 70 (1980) 1139. [24] C.G. Parigger, Atomic and molecular emissions in laser-induced breakdown spectroscopy, Spectrochim. Acta Part B 79–80 (2013) 4–16. [25] G. Herzberg, Molecular Spectra and Molecular Structure. I. Spectra of Diatomic Molecules, 2nd ed. Krieger, 1966. [26] R.S. Harmon, F.C. De Lucia, A.W. Miziolek, K.L. McNesby, R.A. Walters, P.D. French, Laser-induced breakdown spectroscopy (LIBS) — an emerging field-portable sensor technology for real-time, in-situ geochemical and environmental analysis, Geochem. Explor. Environ. Anal. 5 (2005) 21–28. [27] F.J. Fortes, J.J. Laserna, The development of fieldable laser-induced breakdown spectrometer: no limits on the horizon, Spectrochim. Acta Part B 65 (2010) 975–990. [28] F.C. De Lucia, R.S. Harmon, K.L. McNesby, J.R.J. Winkel, A.W. Miziolek, {Laserinduced} breakdown spectroscopy analysis of energetic materials, Appl. Opt. 42 (2003) 6148–6152. [29] L.M. Cabalín, J.J. Laserna, Atomic emission spectroscopy of laser-induced plasmas generated with an annular-shaped laser beam, J. Anal. At. Spectrom. 19 (2004) 445–450. [30] J.L. Gottfried, F.C. De Lucia, C.A. Munson, A.W. Miziolek, Laser-induced breakdown spectroscopy for detection of explosives residues: a review of recent advances, challenges, and future prospects, Anal. Bioanal. Chem. 395 (2009) 283–300. [31] R.C. Chinni, D.A. Cremers, L.J. Radziemski, M. Bostian, C. Navarro-Northrup, Detection of uranium using laser-induced breakdown spectroscopy, Appl. Spectrosc. 63 (2009) 1238–1250. [32] F.R. Doucet, P.J. Faustino, M. Sabsabi, R.C. Lyon, Quantitative molecular analysis with molecular bands emission using laser-induced breakdown spectroscopy and chemometrics, J. Anal. At. Spectrom. 23 (2008) 694. [33] D. von der Linde, H. Schüler, Breakdown threshold and plasma formation in femtosecond laser–solid, interaction, 13 (1996) 216–222. [34] B.C. Castle, K. Talabardon, B.W. Smith, J.D. Winefordner, Variables affecting the precision of laser-induced breakdown spectroscopy measurements, Appl. Spectrosc. 52 (1998) 649–657. [35] D.N. Stratis, K.L. Eland, S.M. Angel, Enhancement of aluminum, titanium, and iron in glass using pre-ablation spark dual-pulse LIBS, Appl. Spectrosc. 54 (2000) 1719–1726. [36] V. Nassisi, D. Doria, A. Pedone, L. Torrisi, S. Gammino, L. Ando, Comparison of nanosecond laser ablation at 1064 and 308 nm wavelength, Appl. Surf. Sci. 210 (2003) 262–273. [37] A.J. Effenberger, J.R. Scott, Effect of atmosphere on collinear double-pulse laserinduced breakdown spectroscopy, Anal. Bioanal. Chem. 400 (2011) 3217–3227. [38] B. Zhang, M. He, W. Hang, B. Huang, Minimizing matrix effect by femtosecond laser ablation and ionization in elemental determination, Anal. Chem. 85 (2013) 4507–4511.

[39] J.L. Todolí, L. Gras, V. Hernandis, J. Mora, Elemental matrix effects in ICP-AES, J. Anal. At. Spectrom. 17 (2002) 142–169. [40] P.E. Janney, F.M. Richter, R.A. Mendybaev, M. Wadhwa, R.B. Georg, E.B. Watson, et al., Matrix effects in the analysis of Mg and Si isotope ratios in natural and synthetic glasses by laser ablation-multicollector ICPMS: a comparison of single- and double-focusing mass spectrometers, Chem. Geol. 281 (2011) 26–40. [41] J.L. Gottfried, F.C. De Lucia Jr., C.A. Munson, A.W. Miziolek, Laser-induced breakdown spectroscopy for detection of explosives residues: a review of recent advances, challenges, and future prospects, Anal. Bioanal. Chem. (2009) 283–300. [42] M. Dong, J. Lu, S. Yao, Z. Zhong, J. Li, J. Li, et al., Experimental study on the characteristics of molecular emission spectroscopy for the analysis of solid materials containing C and N, Opt. Express 19 (2011) 17021. [43] T. Delgado, J.M. Vadillo, J.J. Laserna, Laser-induced plasma spectroscopy of organic compounds. Understanding fragmentation processes using ion–photon coincidence measurements, J. Anal. At. Spectrom. 28 (2013) 1377. [44] C. Pasquini, J. Cortez, L.M.C. Silva, F.B. Gonzaga, Laser induced breakdown spectroscopy, 18 (2007) 463–512. [45] M.L. da Silva, Guidelines for the Calculation of Bound Molecular Spectra, arXiv:physics/ 05091322005 (17). [46] J. Dunham, The energy levels of a rotating vibrator, Phys. Rev. 41 (1932) 721–731. [47] J.T. Vanderslice, E.A. Mason, W.G. Maisch, E.R. Lippincott, Ground state of hydrogen by the Rydberg–Kelein–Rees method, J. Mol. Spectrosc. 3 (1959) 17–29. [48] N.L. Singh, D.C. Jain, The Rydberg–Klein–Rees method of constructing the true potential energy curves of diatomic molecules, Proc. Phys. Soc. 79 (1962) 274–278. [49] S. Chandra, Relation between spectroscopic constants with limited Dunham coefficients, Pramana 65 (2005) 1133–1137. [50] N. Inostroza, J.R. Letelier, M.L. Senent, On the numerical determination of Dunham's coefficients: an application to X1Σ + HCl isotopomers, J. Mol. Struct. THEOCHEM 947 (2010) 40–44. [51] F. Mélen, I. Dubois, H. Bredohl, The A–X and B–X transitions of BO, J. Phys. B At. Mol. Opt. Phys. 2423 (1985) 2423–2432. [52] V.N. Ochkin, Spectroscopy of Low Temperature Plasmas, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, 2009. [53] H.S. Liszt, W.M.H. Smith, RKR Franck–Condon factors for blue and ultraviolet transitions of some metal oxides, J. Quant. Spectrosc. Radiat. Transf. 11 (1971) 1043–1062. [54] M. Kuzmanović, A. Antić-Jovanović, D. Pešić, V. Bojović, Franck–Condon factors and r-centroids for the B–X bands of 10B18O and 11B18O molecules, J. Serb. Chem. Soc. 70 (2005) 727–733. [55] D.W. Berreman, Unfolding spectrometer slit broadening effects from broad spectra, Appl. Opt. 7 (1968) 1447–1453. [56] D. Bulajic, M. Corsi, G. Cristoforetti, S. Legnaioli, V. Palleschi, A. Salvetti, et al., A procedure for correcting self-absorption in calibration free-laser induced breakdown spectroscopy, Spectrochim. Acta Part B 57 (2002) 339–353. [57] V.N. Rai, S.N. Thakur, Physics of plasma in laser-induced breakdown spectroscopy, in: J.P. Singh, S.N. Thakur (Eds.), Laser-Induced Break. Spectrosc, Elsevier B.V., Oxford, 2007, pp. 84–111. [58] R.M. Herman, Unified impact theory for velocity-changing effects and speed dependencies in neutral species lineshapes, Int. J. Spectrosc. 2010 (2010) 1–9. [59] C. Wang, Cavity ringdown spectroscopy of plasma species, in: P.K. Chu, X. Lu (Eds.), Low Temp. Plasma Technol. Methods Appl. CRC Press, Boca Raton, FL, 2014, pp. 207–260. [60] G. Colonna, A. Casavola, M. Capitelli, Modelling of LIBS plasma expansion, Spectrochim. Acta Part B 56 (2001) 567–586. [61] C.G. Parigger, A. Woods, J. Hornkohl, Analysis of atomic and molecular superposition spectra following laser-induced optical breakdown, Int. Rev. At. Mol. Phys. 2 (2011) 77–83. [62] Z. Machala, M. Janda, K. Hensel, I. Jedlovský, L. Leštinská, V. Foltin, et al., Emission spectroscopy of atmospheric pressure plasmas for bio-medical and environmental applications, J. Mol. Spectrosc. 243 (2007) 194–201. [63] T.F. Coleman, Y. Li, An interior trust region approach for nonlinear minimization subject to bounds, SIAM J. Optim. 6 (1996) 418–445. [64] T.F. Coleman, Y. Li, On the convergence of interior-reflective Newton methods for nonlinear minimization subject to bounds, Math. Program. 67 (1994) 189–224. [65] T.B.B. Coplen, J.K. Bohlke, P. De Bievre, T. Ding, N.E.E. Holden, J.A.A. Hopple, et al., Isotope-abundance variations of selected elements (IUPAC technical report), Pure Appl. Chem. 74 (2002) 1987–2017.