Spectrochimica Acta Part B 62 (2007) 1419 – 1425 www.elsevier.com/locate/sab
Forensic comparative glass analysis by laser-induced breakdown spectroscopy ☆ Candice M. Bridge a , Joseph Powell b , Katie L. Steele a , Michael E. Sigman a,⁎ a
National Center for Forensic Science and Department of Chemistry, University of Central Florida, PO Box 162367, Orlando, FL 32816-2367, United States b South Carolina State Law Enforcement Department (SLED), 4400 Broad River Road, Columbia, SC 29210, United States Received 15 December 2006; accepted 11 October 2007 Available online 18 October 2007
Abstract Glass samples of four types commonly encountered in forensic examinations have been analyzed by laser-induced breakdown spectroscopy (LIBS) for the purpose of discriminating between samples originating from different sources. Some of the glass sets were also examined by laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS). Refractive index (RI) measurements were also made on all glass samples and the refractive index data was combined with the LIBS and with the LA-ICP-MS data to enhance discrimination. The glass types examined included float glass taken from front and side automobile windows (examined on the non-float side), automobile headlamp glass, automobile sidemirror glass and brown beverage container glass. The largest overall discrimination was obtained by employing RI data in combination with LAICP-MS (98.8% discrimination of 666 pairwise comparisons at 95% confidence), while LIBS in combination with RI provided a somewhat lower discrimination (87.2% discrimination of 1122 pairwise comparisons at 95% confidence). Samples of side-mirror glass were less discriminated by LIBS due to a larger variance in emission intensities, while discrimination of side-mirror glass by LA-ICP-MS remained high. © 2007 Elsevier B.V. All rights reserved. Keywords: LIBS comparative glass analysis; LA-ICP-MS; Forensic analysis
1. Introduction Recent advances in instrumentation have produced commercial LIBS spectrometers that are inexpensive, thereby opening new opportunities for industrial applications. One industry that stands to gain significant advantage for the advent of inexpensive LIBS instrumentation is that of forensic science, where state and local crime labs often operate on somewhat limited budgets. The relatively non-destructive nature of LIBS, rapid analysis time, minimal sample preparation, and potential
☆
This paper was presented at the 4th International Conference on Laser Induced Plasma Spectroscopy and Applications (LIBS 2006) held in Montreal, Canada, 5–8 September 2006, and is published in the Special Issue of Spectrochimica Acta Part B, dedicated to that conference. ⁎ Corresponding author. Tel.: +1 407 823 3420. E-mail address:
[email protected] (M.E. Sigman). 0584-8547/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.sab.2007.10.015
for field portability also make this spectroscopic method attractive to forensic science. In a previous study, LIBS was evaluated as a method for the forensic analysis of glass, which is an evidentiary item where elemental composition can provide valuable discriminating information [1]. In this paper, LIBS is further examined for the discrimination of multiple sets of glass samples, including samples of automobile window glass, drink container glass, automobile side-mirror glass and automobile headlamp glass. A summary description of the sample sets and the analytical methods employed are given in Table 1. Several reviews of LIBS applications have recently appeared [2–5]. Challenges in the application of LIBS spectroscopy to quantitative analytical problems, including a high background continuum [6], line-broadening, and self-absorption do not prohibit the use of LIBS for the forensic discrimination between two samples. Comparison of samples from questioned and known origins by LIBS can be made based on the relative
1420
C.M. Bridge et al. / Spectrochimica Acta Part B 62 (2007) 1419–1425
Table 1 Glass samples analyzed Sample set description A 27 Automobile glass samples (side windows and rear windows); analyzed on
the non-float side by LIBS. Refractive index measurements were also used in combination with LIBS. B 15 Automobile headlamp samples, analyzed by LIBS and LA-ICP-MS. Refractive index measurements were used in combination with both LIBS and LA-ICP-MS. C 15 Brown beverage glass samples; analyzed by LIBS. Refractive index measurements were also used in combination with LIBS. D 34 Automobile side-mirror samples (analysis of non-mirrored side) by LIBS and LA-ICP-MS. Refractive index measurements were used in combination with both LIBS and LA-ICP-MS.
elemental emission intensities from each sample without relating emission intensities to elemental concentrations [7,8]; however, signal reproducibility is an important issue for sample discrimination and proper statistical analysis of spectral data is required to insure that Type I and Type II errors are held at reasonable levels. Questioned and known sample comparisons are also exempt from influence by fractionation, wherein the interaction of the laser with the material can produce atomic populations in the plasma that do not reflect the relative concentrations in the matrix [9]. If two samples originate from the same source, fractionation effects from both samples may reasonably be expected to be the same. Confinement effect on emission intensities, which result from plasmas formed within craters [10], suggests that strict protocols for laser focusing and spectral averaging may play an important role in sample comparison methodologies. A review of the current status of forensic glass comparison has been given elsewhere [11]. Forensic analysis of glass has been based on refractive index, dispersion and density analyses [12]. The statistical discrimination of flat glass by neutron activation analysis, and by the determination of up to 70 elements by inductively coupled plasma-mass spectrometry (ICPMS) has been demonstrated [13,14]. The forensic significance of glass composition and refractive index has been assessed, using inductively coupled plasma-atomic emission spectroscopy (ICP-AES) to measure the concentrations of ten elements (Ca, Fe, Al, Mn, Sr, Mg, Ba, Ti, Zr, Na) [15]. The probability of unrelated glass specimens having indistinguishable elemental compositions and refractive indices was calculated to lie between extremes of 10− 5–10− 13. Similar results were later shown to hold for determination of the same ten elements by ICP-AES and ICP-MS [16,17]. Elemental analysis of glass evidence in forensic casework has recently been reported [18]. Laser ablation-inductively coupled plasma-mass spectrometry (LAICP-MS) has also been used for the forensic comparison of glass samples [19,20]. There have been previous reports of glass analysis by LIBS [21–29]; however, those reports have not focused on forensic comparative analysis of glass. Research by Almirall [18–20], Koons [11,16] and Duckworth and Bayne [17] have established a statistical framework for interpreting forensic glass evidence. The data analysis approach taken in this paper for the analysis of LIBS data follows methods that have previously been published and are gaining
acceptance for the analysis of glass by the forensic community [11,16–20]. Alternative approaches to the analysis of LIBS data have also been reported, notably including spectral correlation [30,31], and multivariate statistical approaches such as principal components analysis (PCA) [31,32]. Those methods have not been applied to the analysis of the data presented here. 2. Experimental 2.1. Instrumentation The LIBS instrument used in this research was an Ocean Optics (Dunedin, Fl, USA), model LIBS2000+. This instrument was equipped with a Q-switched Nd-YAG, pulsed laser (Big Sky Lasers, model CFR200, Bozeman, Montana, USA), with a 1064 nm output and a pulse width of 9 ns. Spectra reported here were collected with laser output energy of 98 mJ/pulse, and detector delay that ranged from 2.0 to 6.5 μs depending on the sample matrix. The plasma-generated emission intensities (200–900 nm) were collected by a fiber optic bundle connected to seven sequential CCD spectrometers. The LIBS sample chamber was comprised of a plastic box fitted with an x,yadjustable sample stage, as previously reported [1]. Data acquisition and some data analysis were performed using the Ocean Optics OOILIBS software. The LA-ICP-MS system used in this study was equipped with a laser ablation unit (New Wave Research/Mechantek Fremont, CA, USA) model LUV 213 and an ICP-MS (Agilent Technologies, Palo Alto, CA, USA) model 7500s [1]. Laser ablation was accomplished with 213-nm output from a Qswitched Nd-YAG laser, pulse width of 3–5 ns), and energy of 3 mJ/pulse. Refractive index measurements were made with a glass refractive index measurement (GRIM3) instrument (Foster and Freeman, Evesham, Worcestershire, UK), employing a 589 nm lamp and Mettler Toledo (Columbus, OH, USA) hot stage. Glass samples from side window and side-mirror glass were collected from automobiles at a local salvage yard. Headlamp glass was taken from lamps purchased from after-market vendors and from a local salvage yard. Drinking glass samples were from beverage containers collected from local bars. Glass sample preparation for both the LIBS and the LA-ICP-MS measurements was minimal and involved only wiping the surface with a clean lint-free paper [1]. The first layer of surface ablation data was discarded to ensure a clean surface for LAICP-MS analysis (vide infra), but a similar surface-cleaning laser pulse was not used prior to LIBS analysis. Previously reported studies indicated that a surface-cleaning ablation did not significantly effect LIBS discrimination.[1] The LIBS spectra were collected under an argon flow at a rate of approximately 120 mL/min. Ten spectra were taken at a frequency of 1 Hz, and averaged to give a single “average spectrum” from a single spot on the glass surface. Average spectra were collected at five different positions on the glass surface for each sample. The total time required to analyze each glass sample was approximately 2 min. Table 2 lists the emission wavelengths and assigned emitting species for each category of
C.M. Bridge et al. / Spectrochimica Acta Part B 62 (2007) 1419–1425
1421
Table 2 LIBS emission line ratios and associated wavelengths used for glass discrimination on all samples (A) Float glass (non-float side) Si (221.174)/Nb (243.538) Si (252.851)/Mn II (257.61) Si (288.18)/Ca II (317.93) Fe (373.713)/U (383.146) Cr (391.568)/Fe (393.591) Fe (393.591)/Ca (422.67) Ca II (527.018)/Mn (534.94) Na (819.479)/Cl (822.174)
(B) Headlamp glass B (208.957)/Si (221.089) B (249.772)/Si (250.69) Mg II (279.553)/Pb (280.2) Mg II (280.271)/Si (288.158) Th II (339.203)/Zr II (343.823) U (394.382)/Al (396.152) Al (396.152)/Ca II (396.847) Na (819.479)/Cl (821.204) Na (819.479)/Cl (822.174)
(C) Brown container glass Si (221.806)/Fe II (238.204) Mg II (279.553)/Na (285.281) Pb (280.200)/Mg (285.213) Na (285.281)/Si (288.158) Mg (518.36)/Ca (558.876) Ca (558.876)/Na (588.995) Ca (616.217)/Cd (643.847) Na (818.326)/Cl (822.174) Na (819.479)/Cl (821.204)
(D) Side-mirror glass Na (285.281)/Si (288.158) Mg II (280.271)/Sn (283.998) Si (288.158)/Ca II (317.93) Fe (373.713)/U (383.146) Mg (383.829)/Ca II (396.847) Cr (390.568)/Fe (393.591) Fe (393.591)/Ca (422.67) Mg (518.36)/Ca (558.875) Fe (559.47)/Na (568.864)
glass. Emitting species are assigned based on elements commonly found in glass with transitions falling within ± 0.06 nm of the observed line. Spectral resolution on the instrument used in this research was 0.06 nm/channel. Representative LIBS spectra from float glass, beverage glass and headlamp glass are shown in Fig. 1, and serve to demonstrate the differences between glass samples of different types. Three representative LIBS float glass spectra from different sources are shown in Fig. 2. The similarity of these samples are apparent from the spectra and the difficulty in discriminating between samples from different sources. The LA-ICP-MS analysis was conducted under argon and data from the first ablated layer was discarded, while data from the
Fig. 2. LIBS automobile side window float glass spectra (float side of the glass) taken from (A) a 1990 Chevrolet Caprice, (B) a 2000 Pontiac Grand AM and (C) a 1993 Mazda 626. Each spectrum is the average of 10 single-pulse spectra taken at one location on the glass surface.
following three layers of ablation were averaged. The data from the first ablated layer was discarded to reduce the possibility of carryover contamination between samples. Each ablation was approximately 5 μm deep and data was collected by rastering over the sample to produce an ablated area of approximately 1.5 mm × 1.5 mm. The isotopes utilized for the discrimination of each set of glass samples is given in Table 3. The sets of ions was chosen based on previous work by Su [33], Almirall [19], and previous experience by the South Carolina Law Enforcement Department (SLED). The final selection of a set of isotopes for a given sample set was based on an analysis of variance and maximizing the information content, as described below under “Statistical Analysis”. The intensities measured by LA-ICP-MS were optimized for the analyte isotopes of interest in the NIST SRM sample 612 and held to a day-to-day variation of ±15%, as previously reported. [1]. Refractive index (RI) measurements on glass samples were made according to a previously reported methodology. [1] The refractive index (RI) was obtained by placing glass fragments in standard immersion oils B and C from Locke Scientific (Hants, UK). The reported RI values are averages taken from 14 fragments of each glass studied.
Table 3 Isotopic abundance ratios used to discriminate glass samples by LA-ICP-MS (B) Headlamp glass
(D) Side-mirror glass
7
7
Li/ Cr Na/24Mg 27 Al/47Ti 57 Fe/55Mn 66 Zn/60Ni 85 Rb/232Th 88 Sr/90Zr 118 Sn/44Ca 139 La/178Hf 27 Al/29Si 23
Fig. 1. LIBS spectra taken from (A) automobile headlamp glass, (B) brown beverage container glass and (C) automobile side window float glass, spectrum taken on the float side of the glass. Each spectrum is the average of 10 singlepulse spectra taken at one location on the glass surface.
53
Li/53Cr Na/24Mg 27 Al/47Ti 29 Si/39K 55 Mn/238U 57 Fe/55Mn 66 Zn/60Ni 118 Sn/44Ca 232 Th/238U 27 Al/29Si 23
1422
C.M. Bridge et al. / Spectrochimica Acta Part B 62 (2007) 1419–1425
2.2. Statistical analysis The Tukey honestly significant difference (HSD) test was used to analyze the LIBS, LA-ICP-MS and RI data in order to determine the discriminating power of these techniques for each set of glass samples. The discriminating power was determined for each ratio of emission intensities (LIBS), isotopic abundance ratios (LA-ICP-MS), and refractive index. Refractive index was combined with both LIBS and LA-ICP-MS as a means of possibly enhancing discrimination by each of these two techniques. The Tukey HSD test was utilized to ensure that the probability of a Type I error was held constant during the multiple pairwise comparisons within each data set [17]. An analysis of variance (ANOVA) was performed to evaluate the variance in the isotope ratios, refractive index measurements and emission intensity ratios, both within a set of replicate measurements on a single glass sample and between sets of replicate measurements on different samples (i.e. the statistical F values were calculated). The isotopic ratios from LA-ICP-MS and emission ratios from LIBS analysis with the largest F values were chosen from each sample set to be used as possible discriminating measurements. When two ratios were found to be highly correlated, one of the ratios was abandoned in favor of a more poorly correlated ratio with a high F value, so as to minimize the correlation within a data set and maximize the information content. The Pearson productmoment correlation matrix was calculated for each set of ratios to ascertain linear independence within the set [15]. The ratios were deemed to be sufficiently independent to be retained for the purpose of discrimination if the correlation coefficient (r) was less than a cutoff-criterion of 0.8. The ANOVA and ratio selection was followed by the Tukey HSD post test. The average of an elemental or emission ratio for two glasses is significantly different at a chosen experimentwise error rate (α) if Tukey's HSD comparison, Eq. (1), holds. P P SW jX i X i Vjz pffiffiffi TQða; h; df Þ n
ð1Þ
In Eq. (1), X¯i and X¯i′ are the average value of the parameter for samples i and i′, and n is the number of replicates per sample, SW is the within-group standard deviation, and Q(α,h, df) is the critical value of a studentized range distribution at the α percentile point for h samples and df degrees of freedom used to estimate the standard deviation SW [17]. The standard deviation within groups, SW, is calculated by Eq. (2);
SW ¼
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi uP n P 2 uh P u X Xi ti¼1 j¼1 i;j N h
ð2Þ
where Xi,j is the jth replicate measurement of a parameter (emission ratio, refractive index, etc.) of the ith sample. The total number of replicates, N, may be calculated as h * n if each of the h samples contains n replicates. Refractive index values were analyzed by the same method. Statistical tests were
performed in Mathematica (Wolfram Research) and Microsoft Excel (Microsoft Corp.) software packages. The “discrimination percentage” for each ratio was calculated as the fraction of the total number of pairwise comparisons that were “distinguishable” based on that specific ratio. The overall discrimination of glass samples was achieved by considering either the LIBS or the LA-ICP-MS data alone and in combination with the RI data. 3. Results and discussion 3.1. Precision of LIBS and LA-ICP-MS measurements Although the statistical techniques used in the pairwise comparisons of questioned and known glass samples take into account the within and between sample variances (ANOVA and Tukey HSD discussed above), it is instructive to examine the precision of the LIBS and LA-ICP-MS measurements and how the precision varied between analyses conducted on the same day and on different days, following changes in optical alignment and re-optimization. In order to determine the precision of the LIBS measurements, the following experiments were performed on three separate days. Each day, the optical alignment of the LIBS instrument was optimized by adjusting the collection optic and height of the focusing lens to achieve maximum emission intensity with a borosilicate glass sample. To determine the same-day (back-toback) reproducibility of LIBS spectra, a sample of NIST SRM 610 standard glass was analyzed by collecting average LIBS spectra (each comprised of an average of 10 single-shot spectra), at five locations on the sample. The alignment of the collection optic and the height of the laser focusing lens were then altered and the process was repeated on subsequent days. The average %RSD for a set of 11 emission peak intensity ratios was calculated from individual spectra and the average peak ratio %RSD for any given day was determined to be 6.5 ± 1.4%. The ratios of emission intensities were chosen such that they approximate the range of ratios observed for the glass comparisons described below. While the peak ratios on any given day were fairly consistent in back-to-back average spectra, the ratios showed considerably larger change from day to day. The average %RSD for the peak ratios determined over the three days was 24.5 ± 29.2%. The %RSD for individual peak ratios determined over a three day period were as large as 105% in one case. The results emphasize the benefit of limiting glass comparisons to same-day back-to-back analyses of questioned and known samples. All comparisons within a set of glass samples discussed in the following paragraphs are based on emission intensity ratios from measurements made back-to-back on a single day. The precision of LA-ICP-MS measurements over a series of days was determined by analyzing a sample of SRM 612 on six days and constructing a set of 36 ratios from 72 isotopic intensities. The average %RSD for the 36 ratios determined over the six days was only 9.1 ± 5.8% with the largest between-day % RSD for any one ratio being 24%. While the single-day average %RSD of the ratios was not determined for LA-ICP-MS, the
C.M. Bridge et al. / Spectrochimica Acta Part B 62 (2007) 1419–1425 Table 4 Discrimination for each sample set (A–D) by LIBS, LA-ICP-MS and RI LIBS
LA-ICP-MS
(A) Float glass (27 samples, 351 comparisons, analysis of non-float side) 10 emission ratios + RI 98.9 10 emission ratios 74.4
RI 94.6
(B) Automobile headlamps (15 samples, 105 comparisons, analysis of outside surface only) 10 emission ratios + RI 100.0 10 isotopic ratios + RI a 92.4 79.0 10 emission ratios 98.1 10 isotopic ratios a 70.48 (C) Brown drinking glass (15 samples, 105 comparisons) 10 emission ratios + RI 99.0 10 emission ratios 99.0
80.0
(D) Automobile side-mirror glass (34 samples, 561 comparisons, analysis of non-coated side only) 10 isotopic ratios + RI b 100.0 45.3 10 emission ratios + RI 75.2 10 emission ratios 56.2 10 isotopic ratios b 100.0 a b
Sample analysis by “Drill-Down” technique. Sample analysis by “Rastering” technique.
between-day precision is seen to be significantly better than observed for LIBS. All discrimination of glass samples by LIBS was based on single-day back-to-back measurements. 3.2. Pairwise discrimination of glass samples Table 4 gives the tabulated discrimination for each dataset. The table lists the number of emission or isotopic ratios used in the comparison for each data set and the percent discrimination obtained by each method at the 95% confidence level. Percent discrimination was determined by LA-ICP-MS, LIBS and RI. Percent discrimination was also determined from LA-ICP-MS and LIBS data combined with RI measurements. Some comments are required concerning interpretation of the RI data. A standard deviation of 2.2 × 10− 4 has been reported for repetitive automated RI measurements within a single pane of float glass [11]. This value for the standard deviation was adopted for determining the discriminating power of RI for float glass in previous studies which combined RI and LA-ICP-AES [17]. If the previously used standard deviation for RI measurements is applied to the float glass samples in this study, the discrimination is lower than reported in Table 4. The RI discrimination values for automobile side window glass (set A) in Table 4 are based on RI measurements made using a GRIM3 instrument, which gave a smaller standard deviation (1 × 10− 5 ), thereby affording a somewhat higher discrimination, 94.6%. The percent discrimination based on RI decreased for sample sets B, C and D, due to a larger variance of the refractive index within a glass sample. This change is reflected in the refractive index ANOVA F values of 4,054 and 36, for sample sets A and D respectively. The lower F value for the side-mirror glass (set D) is the result of an increase of the within-group variance by three orders of magnitude, while the between-group variance increased by only one order of magnitude. The refractive index measurements for some, but not all, of the side-mirror glass samples exhibited large variations. Approximately 11 of the 34 samples gave excessively large
1423
Table 5 Total discrimination across all sample sets (A–E) by LIBS and LA-ICP-MS at the 95% confidence level without the use of refractive index data Total for LIBS
Total for LA-ICP-MS
1122 comparisons
666 comparisons
783 discriminations = 69.8% discrimination
635 discriminations = 95.3% discrimination
variations in RI, and the remaining samples exhibited variances which approximated the variance in RI measurements for the other sets of glass samples. Table 5 gives the total discrimination across all sample sets, as determined by LA-ICP-MS and LIBS without the use of RI data to aid in the discrimination. Combining the RI data with either LIBS or LA-ICP-MS data provides higher discrimination. Table 6 gives the total discrimination across all sample sets when LA-ICP-MS and LIBS data are used in conjunction with RI data. Comparison of Tables 5 and 6 shows LA-ICP-MS to have an overall higher discrimination across the four sample sets examined. LA-ICP-MS, when used in combination with RI provided greater than 98% discrimination across all sample sets, Table 6. LIBS, used in combination with RI gave 87.2% discrimination across all data sets, Table 6. The overall performance of LIBS was decreased by the low discrimination percentage observed for the side-mirror data set, where LIBS in combination with RI gave 75.2% discrimination, Table 4. The automobile side-mirror glass samples (discussed above) had the largest within-group variance for the refractive index; however, LA-ICP-MS provided very good discrimination for this data set. The two “optical” methods (LIBS and RI) gave lower discrimination percentages for the side-mirror glass. As noted above, 11 of the 34 glass samples exhibited the large variance in RI. When the 11 high-variance glass samples were removed from the data set, the discrimination by RI increased to 88.5%. However, after removing the results for the highvariance glasses, the discrimination by LIBS without RI data changed only slightly, increasing from 56.2% to 60.9%. The high-variance in RI was not directly correlated to the reduced discrimination by LIBS. The reason for the reduced discrimination of side-mirror glass by LIBS is not known at this time. An important parameter that was not examined for any of the sample sets is the potential of wavelength dependence for LIBS discrimination. A modified sampling methodology that employs a shorter wavelength laser that would be optically absorbed by the glass and result in ablation of a larger amount of sample might possibly increase the discrimination for this set of Table 6 Total discrimination across all sample sets (A–E) by LIBS and LA-ICP-MS at the 95% confidence level with the use of refractive index data Total for LIBS
Total for LA-ICP-MS
Total for RI
1122 comparisons
666 comparisons
1122 comparisons
978 discriminations = 87.2% discrimination
658 discriminations = 98.8% discrimination
753 discriminations = 66.1% discrimination
1424
C.M. Bridge et al. / Spectrochimica Acta Part B 62 (2007) 1419–1425
glasses. Alternatively, the discriminating power may be increased by selection of an alternative set of emission ratios. 4. Conclusion In this work, we have shown that the discriminating power of LIBS is noticeably less than that of LA-ICP-MS for the analysis of glass samples drawn from four types commonly encountered in forensic casework. This conclusion is based on the analysis methodology that utilized a comparison of emission line ratios within each spectrum. The use of refractive index measurements in addition to the LIBS spectra significantly increases the overall discriminating power of LIBS. The combined discriminating power of LIBS and RI exceeds that of either individual technique. The discriminating power of LIBS in this study was lowered significantly by the results from a single set of glass samples, those from side-mirrors. A subset of the side-mirror samples exhibited high variance and low discrimination by RI, which was not correlated with the reduced discrimination by LIBS for the same data set. Based on the results of this investigation, LIBS analysis of glass performed with 1064 nm light offers a significant discriminating capability when combined with refractive index measurements. The method is nearly non-destructive, requires little or no sample preparation and the analysis time is only a few minutes. LIBS offers an inexpensive glass analysis method affordable to most state and local crime laboratories, but care must be taken when the technique fails to discriminate between a questioned and known sample. Failure to discriminate in the sample sets examined here constitutes a Type II error, since the glasses are known to come from different sources. In cases were two samples are not discriminated by LIBS or LIBS in combination with RI, a secondary analysis is recommended, preferably by LA-ICP-MS in combination with RI. Acknowledgments This work was supported under Award number 2004-IJ-CXK031 from the Office of Justice Programs, National Institute of Justice, Department of Justice. Points of view in this document are those of the authors and do not necessarily represent the official position of the U.S. Department of Justice. The work was done at the National Center for Forensic Science, a National Institute of Justice program hosted by the University of Central Florida, and a member of the Forensic Science Resource Network. References [1] C.M. Bridge, J. Powell, K.L. Steele, M. Williams, J.M. MacInnis, M.E. Sigman, Characterization of automobile float glass with laser-induced breakdown spectroscopy and laser ablation inductively coupled plasma mass spectrometry, Appl. Spectrosc. 60 (2006) 1181–1187. [2] Y.-I. Lee, J. Sneddon, Recent developments in laser-induced breakdown spectroscopy, ISIJ Int. 42 (2002) S129–S136. [3] W. Lee, J. Wu, Y. Lee, J. Sneddon, Recent applications of laser-induced breakdown spectroscopy: a review of material approaches, Appl. Spectrosc. Rev. 39 (2004) 27–97.
[4] D.W. Hahn, A.W. Miziolek, V. Palleschi, Laser-induced breakdown spectroscopy: an introduction to the feature issue, Appl. Opt. 42 (2003) 5937. [5] D.A. Rusak, B.C. Castle, B.W. Smith, J.D. Winefordner, Recent trends and the future of laser-induced plasma spectroscopy, Trends Anal. Chem. 17 (1998) 453–461. [6] M. Kuzuya, M. Murakami, N. Maruyama, Quantitative analysis of ceramics by laser-induced breakdown spectroscopy, Spectrochim. Acta Part B 58 (2003) 957–965. [7] F.C. De Lucia Jr., R.S. Harmon, K.L. McNesby, R.J. Winkel Jr., A.W. Miziolek, Laser-induced breakdown spectroscopy analysis of energetic materials, Appl. Opt. 42 (2003) 6148–6152. [8] V. Hohreiter, D.W. Hahn, Calibration effects for laser-induced breakdown spectroscopy of gaseous sample streams: analyte response of gas-phase species versus solid phase species, Anal. Chem. 77 (2005) 1118–1124. [9] R.E. Russo, X.M. Mao, H. Liu, J. Gonzalez, S.S. Mao, Laser ablation in analytical chemistry — a review, Talanta 57 (2002) 425–451. [10] M. Corsi, G. Cristoforetti, M. Hidalgo, D. Iriarte, S. Legnaioli, V. Palleschi, A. Salvetti, E. Tognoni, Effect of laser-induced crater depth in laserinduced breakdown spectroscopy emission features, Appl. Spectrosc. 59 (2005) 853–860. [11] R.D. Koons, J. Buscaglia, M. Bottrell, E. Miller, Forensic glass comparisons, in: R. Saferstein (Ed.), Forensic Science Handbook, Vol. 1, 2nd ed, vol. 1, Prentice Hall, Upper Saddle River, NJ, 2002, pp. 161–213. [12] J.I. Thornton, D. Crim, The use of k values in the interpretation of glass density and refractive index data, J. Forensic Sci. 34 (1989) 1323–1328. [13] S.J. Pitts, B.J. Kratochvil, Statistical discrimination of flat glass fragments by instrumental neutron activation analysis methods for forensic science applications, J. Forensic Sci. 36 (1991) 122–137. [14] T. Parouchais, I.M. Warner, L.T. Palmer, H. Kobus, The analysis of small glass fragments using inductively coupled plasma mass spectrometry, J. Forensic Sci. 41 (1996) 351–360. [15] R.D. Koons, J. Buscaglia, The forensic significance of glass composition and refractive index measurements, J. Forensic Sci. 44 (1999) 496–503. [16] R.D. Koons, J. Buscaglia, Interpretation of glass composition measurements: the effects of match criteria on discrimination capability, J. Forensic Sci. 47 (2002) 505–512. [17] D.C. Duckworth, S.J. Morton, C.K. Bayne, R.D. Koons, S. Montero, J.R. Almirall, Forensic glass analysis by ICP-MS: a multi-element assessment of discriminating power via analysis of variance and pairwise comparisons, J. Anal. At. Spectrom. 17 (2002) 662–668. [18] S. Montero, A.L. Hobbs, T.A. French, J.R. Almirall, Elemental analysis of glass fragments by ICP-MS as evidence of association: analysis of a case, J. Forensic Sci. 48 (2003) 1101–1107. [19] T. Trejos, S. Montero, J.R. Almirall, Analysis and comparison of glass fragments by laser ablation inductively coupled plasma mass spectrometry, Anal. Bioanal. Chem. 376 (2003) 1255–1264. [20] T. Trejos, J.R. Almirall, Effect of fractionation on the forensic elemental analysis of glass using laser ablation inductively coupled plasma mass spectrometry, Anal. Chem. 76 (2004) 1236–1242. [21] K. Muller, H. Stege, Evaluation of the analytical potential of laser-induced breakdown spectrometry for the analysis of historical glasses, Archaeometry 45 (2003) 421–433. [22] V. Lazic, R. Fantoni, F. Colao, A. Santagata, A. Morone, V. Spizzichino, Quantitative laser induced breakdown spectroscopy analysis of ancient marbles and corrections for the variability of plasma parameters and of ablation rate, J. Anal. At. Spectrom. 19 (2004) 429–436. [23] M. Ducreux-Zappa, J.-M. Mermet, Analysis of glass by UV laser ablation inductively coupled plasma atomic emission spectrometry, Part 1. Effects of the laser parameters on the amount of ablated material and the temporal behavior of the signal for different types of laser, Spectrochim. Acta Part B 51 (1996) 321–332. [24] M. Ducreux-Zappa, J.-M. Mermet, Analysis of glass by UV laser ablation inductively coupled plasma atomic emission spectrometry. Part 2. Analytical figures of merit, Spectrochim. Acta Part B 51 (1996) 333–341. [25] S. Klein, T. Stratoudaki, V. Zafiropulos, J. Hildenhagen, K. Dickmann, T. Lehmkuhl, Laser-induced breakdown spectroscopy for online control of laser cleaning of sandstone and stained glass, Appl. Phys. A 69 (1999) 441–444.
C.M. Bridge et al. / Spectrochimica Acta Part B 62 (2007) 1419–1425 [26] H. Kurniawan, K. Kagawa, M. Okamoto, M. Ueda, T. Kobayashi, S. Nakajima, Emission spectrochemical analysis of glass containing Li and K in high concentrations using a XeCl excimer laser-induced shock wave plasma, Appl. Spectrosc. 50 (1996) 299–305. [27] Y.I. Lee, J. Sneddon, Direct and rapid determination of potassium in standard solid glasses by excimer laser ablation plasma atomic emission spectrometry, Analyst 119 (1994) 1441–1443. [28] U. Panne, M. Clara, C. Haisch, R. Niessner, Analysis of glass and glass melts during the vitrification of fly and bottom ashes by laser-induced plasma spectroscopy, Spectrochim. Acta Part B 53 (1998) 1969–1981. [29] R. Russo, X.L. Mao, W.T. Chan, M.F. Bryant, W.F. Kinard, Laser ablation sampling with inductively coupled plasma atomic emission spectrometry for the analysis of prototypical glasses, J. Anal. At. Spectrom. 10 (1995) 295–301.
1425
[30] I.B. Gornushkin, B.W. Smith, H. Nasajpour, J.D. Winefordner, Identification of solid materials by correlation analysis using a microscopic laser-induced plasma spectrometer, Anal. Chem. 71 (1999) 5157–5164. [31] C.A. Munson, F.C. De Lucia Jr., T. Pichler, K.L. McNesby, A.W. Miziolek, Investigation of statistics strategies for improving the discriminating power of laser-induced breakdown spectroscopy for chemical and biological warfare agent stimulants, Spectrochim. Acta Part B 60 (2005) 1217–1224. [32] M.Z. Martin, N. Labbe, T.G. Rials, S.D. Wullschleger, Analysis of preservative-treated wood by multivariate analysis of laser-induced breakdown spectroscopy spectra, Spectrochim. Acta Part B 60 (2005) 1179–1185. [33] C.F. Su, S. Feng, J.P. Singh, F.Y. Yueh, J.T. Rigsby III, D.L. Monts, R.L. Cook, Glass Composition measurement using laser-induced breakdown spectrometry, Glass Technol. 41 (2000) 16–21.
