JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 1991 VOL. 6 45 1 Comparison of Refractive Index Energy Dispersive X-Ray Fluorescence and Inductively Coupled Plasma Atomic Emission Spectrometry for Forensic Characterization of Sheet Glass Fragments Robert D. Koons Forensic Science Research Unit FBI Laboratory FBI Academy Quantico VA 22135 USA Charles A. Peters and Pamela S. Rebbert Elemental and Metals Analysis Unit FBI Laboratory Washington DC 20535 USA Fragments (in the milligram size range) from 81 tempered sheet glasses were used in order to evaluate the source discrimination capabilities of refractive indices (RI) and elemental composition by using energy dispersive X-ray fluorescence (EDXRF) and inductively coupled plasma atomic emission spectrometry (ICP- AES). The X-ray intensities of five elements were determined by EDXRF with precisions of between 1 and 25%.The concentrations of nine elements were determined using ICP-AES and precisions of from less than 1 to about 10% were obtained. Both methods offer improved discrimination capability over RI measurements alone. The technique of EDXRF provides rapid non-destructive testing and is widely available in forensic laboratories. The ICP-AES method offers the advantages of providing quantitative data on the concentration of elements applicability to a greater number of elements and improved discrimination. Keywords X-ray fluorescence; inductively coupled plasma atomic emission spectrometry; elements in glass; forensic examination Methods for determining the concentration of elements in glass fragments both quantitative and qualitative have been explored in the forensic characterization of glass for more than 20 years.Two approaches have been suggested depending upon whether the point in question is one of classification i.e. the placing of a glass fragment into a product-use category or discrimination i. e. the distinction among sources of glass within a product class. Methods used for classification must provide good accuracy but not necessarily good precision since the composition of a sample is compared with a database consisting of fairly broad class definitions. Methods used for discrimination among similar glasses must provide a high degree of precision for as many measured parameters as possible however a high degree of accuracy is not required as long as the samples in question are compared under the same analytical conditions.The long-standing interest in elemen- tal characterization of forensic glass samples is indicated by the number of methods that have been suggested for this purpose including neutron activation analysis l v 2 atomic absorption spe~trometry,~ d.c. arc emission spectro- g r a p h ~ ~ spark source mass spectr~rnetry,~-~ scanning elec- tron microscopy/)<-ray fluore~cence,~~~ energy dispersive X- ray fluorescence (EDXRF),8-10 inductively coupled plasma atomic emission spectrometry (ICP-AES)' '-l3 and induc- tively coupled plasma mass spectrometry.14 Of these ICP- AES and EDXRF are the only methods that have been applied to a significant number of samples and are currently utilized in forensic laboratories.Several recent papers provide a good overview of the use of ICP-AES and EDXRF for forensic glass analysis.8~11.12~15~16 Most procedures for elemental determination were devel- oped for the purpose of classification either as a means of eliminating alibi sources8J2J5J6 or in investigations involv- ing the contamination of or tampering with products.11 However the majority of criminal cases involve the comparison of glass fragments i. e. discrimination is required. Methods developed for classification may not be directly applicable to source discrimination because of the different analytical requirements of the two questions. Currently most examiners of forensic glass rely predomi- nantly upon physical and optical properties when compar- ing glass fragments found at the scene of a crime with those associated with a Recently there has been renewed interest in elemental analysis as a means of improving discrimination because there is some evidence of tighter manufacturing controls on the physical and optical properties of glass.Clearly properties such as refractive index (RI) and dispersion will remain the principal methods of comparing glass samples because they can be determined non-destructively with commonly avail- able instrumentation offer relatively high levels of discrim- ination and are well established in the forensic laboratory and judicial systems. Elemental composition is considered by glass examiners in those instances where additional discrimination is desired The results of a study to evaluate the relative capabilities of RI EDXRF and ICP-AES to discriminate between sheet glasses are reported.Samples for this study consist of 81 tempered side window glasses taken from automobiles produced in the years 1974-1987. Fragments used for comparison of the methods were in the ( 5 mg size range in order to be typical of those occurring in materials trans- ferred from the scene of a crime. Experimental Sample Collection and Description Samples of tempered glass were collected from the side windows of automobiles from automobile salvage yards in the Washington DC USA area. Each sample consisted of several hundred grams of broken glass. The make model year and vehicle identification number of the automobile were recorded for each sample.The 8 1 samples collected for this study represent 19 makes and 60 models from the period 1974-1 987. There are eight instances of two samples from the same make model and year of automobile. Refractive index determinations n (656.3 nm) n (589.3 nm) and nF (486.1 nm) were made on single fragments from each sample using a procedure similar to the Emmons double variation method adopted by the Association of Official Analytical Chemists.17-19 Sheet thickness visual estimate of colour and floathon-float characteristics were also determined using large pieces of glass. Only one sample (1978 Dodge Colt) was non-float glass.452 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 199 1 VOL. 6 Sample Handling and Preparation for Analysis Samples were broken and individual pieces (1-3 g) of the diced glass were selected for analysis.Prior to elemental analysis individual pieces of glass were washed for 30 min with concentrated HN03 and then rinsed three times each with de-ionized water in order to remove surface contami- nation. After drying each piece of glass was placed between polyethylene sheets and crushed so as to provide fragments of appropriate size ( ( 5 mg). Replicate samples for analysis were selected from three separate pieces which had been carried through the washing and crushing procedure. Reagents and Standards Multi-element standard sclutions for ICP-AES were pre- pared by dilution of commercially available single element stock solutions with concentrations of 1000 ,ug rnl-l using distilled de-ionized water.High-purity HCl and HF (J. T. Baker or Fisher) were used for dissolution of the samples. National Institute of Standards and Technology glass Standard Reference Materials (SRMs) 62 1 (Glass Con- tainer) 1830 (Soda-Lime Float Glass) and 1831 (Soda- Lime Sheet Glass) were analysed by ICP-AES as checks on the accuracy of results. No standards were required for EDXRF analysis. However SRMs were used in order to provide multiple fragments of a homogeneous sample needed for optimization of the conditions used for the acquisition of X-ray data. EDXRF Instrumentation and Analytical Conditions X-ray fluorescence measurements were made on individual fragments of glass using a collimated X-ray source and energy dispersive X-ray analyser. Instrument operating conditions were as listed in Table 1.Samples were posi- tioned in the X-ray beam by suspending them upon an approximately 0.4 mm wide strip of adhesive tape contain- ing negligible levels of all the elements of interest. By using the conditions given in Table 1 reproducible results were obtained for a sample positioned within a region 1 mm on either side of the centre of the beam. The accelerating voltage size of the collimator and composition and thick- ness of the filter were optimized in order to obtain the highest peak-to-background ratios for Fe K a X-rays from glass fragments in the milligram size range and allow measurement of X-ray intensities in the energy range up to 20 keV. X-ray spectra were collected until the intensity of the Ca K a X-ray plus background equalled 40000 counts.For a 1 mg sample with the X-ray tube current set to obtain 50% dead time the counting time was about 150 s. Acquisition time was a function of sample size shape and orientation since for the limited range of glasses in this study all concentrations of Ca were about the same. Generally larger samples required longer acquisition times because of the need to reduce the X-ray tube current in order to maintain 50% dead time. Samples requiring acquisition times of more than 300 s resulted in relatively poor precision so were further crushed in order to obtain smaller fragments and shorter counting times. Glass frag- ments of a size that is useful forensically are incompletely penetrated by both the primary and fluorescent X-rays. As a result quantitative elemental concentrations could not be determined with sufficient accuracy and precision for samples of this small size and irregular shape. However in comparing fragments of varying mass and shape from a homogeneous glass standard the intensity ratios of two X- ray lines of similar energies were relatively constant.Therefore the discrimination capability of EDXRF for glass samples was evaluated using X-ray peak intensity ratios after correction for Si escape peaks and continuum background subtraction. In all samples the elements Si K Ca Fe Sr and Zr were determined. Additionally Mn and Rb were determined in some samples. Triplicate determi- nations were made for each sheet glass sample using fragments obtained from separate pieces of glass. Dissolution Procedure for ICP-AES Dissolution of samples was similar to procedures reported previou~ly.~~J~ Fragments of approximately 5 mg in size were weighed on a microbalance to the nearest 0.01 mg and placed in 15 ml polyethylene screw top tubes for dissolu- tion.To each sample tube 500 pl of HF were added. The tubes were placed in an oven at 80 "C for 1 h removed and individually placed briefly in a sonic bath then returned to the oven and dried. After drying 500 p1 of concentrated HC1 were added to each tube. The samples were again dried overnight in the oven at 80 "C. Upon removal from the oven and cooling 500 pl of concentrated HCl 500 pl of a 1000 ppm Sc solution and 9.00 ml of water were added to each tube. The tubes were capped mixed on a vortex mixer and returned to the oven for 1 h.The samples were cooled to room temperature and analysed by ICP-AES using multi- element standards prepared from stock standard solutions. ICP-AES Instrumentation and Analytical Conditions Instrument and operating conditions for the ICP-AES instrument are shown in Table 2. Calculation of the concentration of the elements was carried out using multi- Table 1 Instrumental and analytical conditions for EDXRF Instrumentation- Spectrometer model X-ray source Multichannel analyser Computer Signal processing Analytical conditions- Operating voltagelkv Operating current Source collimator diametedmm Source filter Path Spectral Ijnes/keV Kevex 0700 1200 W Rh anode in direct mode 1024 channels 0-20 keV at 2 eV per channel DEC LSI 11/23 Quantex version 4x21 35 As needed to give 50% dead time 3 0.15 mm aluminium Vacuum Si 1.740 K 3.312 Ca 3.690 Mn 5.895 Fe 6.400 Rb 13.375 Sr.14.142 Zr 15.746JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 199 1 VOL. 6 453 Table 2 Instrumental and analytical conditions for ICP-AES Instrumentation- Spectrometer model Dispersing system Torch Nebulizer Spray chamber R.f. generator Argon flow rate/l min-' Analytical conditions- Sample uptake rate/ml min-' Observation height/mm Spectral line/nm Background correction Signal compensation Integration time/ms Perkin-Elmer Plasma IT Monochromator A 3600 grooves mm-* resolution ~ 0 . 0 0 9 nm 160-400 nm Monochromator B 1800 grooves mm-I resolution ~ 0 . 0 18 nm 160-800 nm Fassel-type Perkin-Elmer high solids Scott design 27.12 MHz 1.2 kW forward power Plasma 15 Auxiliary 1 .O Nebulizer 1.2 1 .o 15 above load coil Monochromator A Fe 238.204 Monochromator B Ca 393.366 Mn 257.610 Mg 279.553 Ti 334.941 A1 396.152 Sr 407.77 1 Ba 455.403 Na 589.592 Auto On all except Na 100 Table 3 Frequency of indistinguishability of 8 1 sheet glass samples taken in pairs (3240 comparisons) Comparison parameter and criteria No.of indistinguishable pairs Frequency n +- 0.0002 n +- 0.000 1 (1) and n,+-0.0004 and nF+0.0004 (2) and n,-k0.0002 and nFk0.0002 EDXRF (see text for criteria) ( 5 ) and (3) ( 5 ) and (4) (8) and (3) (8) and (4) TCP-AES (see text for criteria) 648 418 487 178 305 81 33 3 3 2 1:5.0 1 :7.8 1:6.7 1:18.2 1:10.6 1 :40 1 :98 1:1080 1:1080 1:1620 element external standard solutions for all elements and Sc as an internal standard for all of the elements except for Na.A standard response graph was constructed and SRMs were analysed as a check on accuracy before each set of approximately 25 samples. Triplicate measurements were made on each solution and on triplicate samples from each sheet glass. Triplicate samples were not run sequentially so that a good estimate of total precision over the course of a set of samples would be obtained. Results and Discussion Optical and Physical Properties Discrimination among sources based on thickness colour and floathon-float characteristics provides some sub-divi- sion of automobile sheet glass samples. However since these parameters cannot be measured on small fragments they are not considered in this study.Refractive indices are generally considered to provide the best discrimination capability of the commonly measured optical and physical properties of glass. The RI values for the 81 sheet glasses in this study occur within the ranges of n (1.51 11-1.5201) n D (1.5138-1.5231) and nF (1.5 193- 1.5287). Although these ranges suggest that good discrimination of sources is possible 5 1 of the 8 1 samples had n D values in the range 1.5 180- 1.5 194 and the 65 glasses produced in the US had n values in the range 1.5 1 72- 1.5 1 99. Examiners of forensic glass samples have generally considered glass fragments to originate from different sources when n values differ by more than 0.0002 or n or nF values differ by more than 0.0004.17 These values were derived from interlaboratory studies and are greater than the combined heterogeneity of modern sheet glasses and short-term analytical precision. Using auto- mated imaging systems for match point determination it has been shown that RI variation across most sheet glasses is less than 0.00003 for all three indices.*O For the measurement procedure used in this study more appropri- ate criteria for source discrimination based on within- laboratory precision are 0.0001 for and 0.0002 for n and nF.The worst instance of discrimination using n criteria is 1.5 185 k 0.0002 which includes 32 samples or 1.5 187 k 0.000 1 which includes 22 samples. For 81 samples taken in pairs there are 3240 possible comparisons. Of these the number of indistinguishable pairs of sheet glasses using several RI criteria are shown in the first four lines of Table 3.As shown the discrimination capability increases to some extent when nc and nF are included in the comparison in addition to n,. This conclusion supports the observations made in other foren- sic laboratories in the US and contrasts with those from the UK where dispersion is generally considered to offer little added discrimination. The discrimination capability im-454 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 199 1 VOL. 6 Table 4 Summary of EDXRF results for sheet glass fragments. Results shown are range and mean X-ray intensities or intensity ratios of 243 fragments and mean relative standard deviation (RSD) of triplicate measurements of 8 1 samples Element Range Mean RSD (Yo) Si 3 41 1-5 407 4 338 4.3 Ca 36 155-38717 37482 1 .O Fe 4 989-38 262 27 099 4.3 Sr 534-10791 3080 20.1 Zr 1226-20055 6910 22.4 Si:Ca 0.090-0.144 0.1 16 4.3 Fe:Ca 0.134-1.01 8 0.723 4.8 Sr:Zr 0.102-1.533 0.514 11.5 proves only moderately as the overlap criteria are made more stringent.There are no consistent relationships between automobile make and model and the likelihood of indistinguishable RI values. Of the eight pairs with match- ing make model and year of automobile three pairs have indistinguishable TZ values at the kO.0002 level and two pairs are indistinguishable at all RI levels. The only trend related to place of manufacture is that all glasses made in Japan are at the low end of the RI ranges and the highest three RI values are for glasses made in Europe. Also the non-float glass in this study has the lowest RI values of any sample.EDXRF The discrimination capability of any variable is determined in large part by the precision of the measurement (random error plus sample heterogeneity) and the range of values observed across sources. As a measure of the combined precision of the X-ray intensities obtained and the within- sample variability the standard deviation of the measure- ments from the triplicate fragments of each glass sample relative to the sample mean relative standard deviations (RSDs) were calculated and then averaged over the 81 glass samples. The RSD measurements for the intensity of each element and selected intensity ratios are shown in Table 4. The results are generally similar to those reported by Ryland* using different instrument operating condition^.^ Fragment sizes in the range 0.5-3 mg were used to obtain the data given in Table 4.The RSD values of the intensity of individual elements range up to about 25%. The RSD of the Sr:Zr intensity ratio shows about a 2-fold improvement over either element individually as expected for two X-rays of similar energy from small samples of irregular size and shape. A comparison of the RSD values with the ranges shown in Table 4 indicates that the Fe:Ca and Sr:Zr ratios are the parameters which provide the greatest degree of discrimination using the EDXRF method. Painvise comparison of glass samples to determine the discrimination capability of EDXRF was carried out using a two-step method. Firstly samples containing measurable levels of uncommon elements Mn and Rb and high levels of K were separated as a sub-group of the sheet glasses.Secondly comparison of two samples within a group was made by the method of range overlap ie. for the variables Si:Ca Fe:Ca and Sr:Zr the ranges of triplicate measure- ments for each of two samples were compared. For samples in which a particular ratio exhibited an unusually small range of values the range was increased to a level based on two times the mean RSD before comparison was made. If all variables exhibited overlapping ranges then the samples were considered to be indistinguishable. Use of range overlaps instead of statistically based confidence intervals was selected as being more realistic to the forensic ex- aminer. In forensic comparison of trade evidence the need to limit type I1 errors i.e.incorrectly attributing two samples to a common source is more important than the need to minimize type I errors i.e. incorrectly attributing two samples to different sources since the former tend to associate a suspect with the scene of a crime. Clearly overlapping ranges of all parameters between two samples render them analytically indistinguishable and hence dif- ferent sources are not indicated. As shown in Table 3 of the 3240 paired comparisons 305 indistinguishable pairs were found when the EDXRF criteria were used. In the worst instance one sample (1 984 Ford EXP) was indistinguishable from 26 others. Discrimi- nation of EDXRF is better than that obtained by TZ alone and all three indices using the wider RI discrimination criteria but slightly worse than all three RI using the more stringent criteria.In general the samples that were indistin- guishable by EDXRF were not the same ones that were indistinguishable by RI. This is evidenced by the significant improvement in discrimination when EDXRF and RI results are combined as shown in lines 6 and 7 of Table 3. There were no readily discernable relationships between EDXRF spectra and manufacturers except for the automo- biles made in Japan. All Japanese glasses contained higher amounts of IS Mn and Rb than other automobiles in this study. The eight pairs of samples from the same make model and year were all distinguishable using EDXRF criteria. No other elements such as Mn Ti As or Ba reported by other workers9J0 were observed in the samples of limited composition used in this study.One likely contributing factor to the success of EDXRF in discriminating among similar samples such as those produced within a model year is the heavy reliance upon concentrations of Zr. Zirconium occurs in glass both as a result of erosion of the refractory material from the walls of the melting tank during manufacture and the presence of zircon grains in the sand used in the making of glass. Measurable variations in the concentrations of Zr over the course of large production runs within a manufacturing facility are thus likely to occur and can serve as a disciminator of otherwise indistinguishable glasses. ICP-AES In contrast to EDXRF ICP-AES provides quantitative measurements of the concentrations of elements in glass.In order to avoid biases that might occur between day-to-day analyses of sets of samples SRMs were analysed together with each sample set. For each set of samples the results obtained for the SRMs were compared with their certified values and whenever the results were outside accepted limits recalibration was performed prior to analysis of samples. Table 5 Summary of ICP-AES results for sheet glass fragments. Figures are range and mean of concentrations of 243 analysed fragments and mean RSD (O/O) of triplicate measurements of 81 samples Element Range Mean RSD (O/O) A1 (Yo m/m) Ba/pg g-l Ca (O/o m/m) Fe (Yo m/m) Mg (O/o m/m) Mn/pg g-l Na (Yo m/m) Sr/,ug g-l Tilpg g-I 0.032 1-0.985 7.6-205 4.94-6.76 0.0589-0.484 2.02-2.40 7-262 6.9- 1 3.3 19.0- 1 68 4 1.2-480 0.213 3.2 35.0 5.4 5.97 0.7 0.309 2.1 2.24 1.2 32 8.0 10.1 12.0 46.5 1.1 121 6.5JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 199 1 VOL.6 45 5 Table 6 Data for four sheet glass samples Manufacturer model and year Parameter Optical data- nc nD HF EDXRF- Si (counts) Ca (counts) Fe (counts) Sr (counts) Zr (counts) Si:Ca Fe:Ca Sr:Zr Al/pg g- * Ba/,ug g-I Ca (O/o m/m) Fe (O/o m/m) Mg (O/o m/m) Mnlpg g-I Na (O/o m/m) Sr/,ug g-' Ti/pg g-* ICP-AES- Chevroiet Nova 1978 1.5161 1.5186 1.5246 4291 4 167 4757 37663 37484 37 160 29 033 20 834 28 037 1997 2083 2 312 12 131 13 154 15431 0.114 0.111 0.128 0.77 0.82 0.76 0.16 0.16 0.15 526 515 528 16.1 19.1 17.9 5.96 6.03 6.07 0.330 0.337 0.339 2.17 2.21 2.20 14.5 15.5 11.5 8.7 9.3 10.1 31.0 31.6 31.1 120 112 114 Pontiac Grand Prix 1977 1.5159 1.5187 1.5247 4 279 4045 4 123 36303 38 121 37 146 27 884 28 232 28 987 3261 2122 2118 10 286 5 765.7 338 0.118 0.106 0.111 0.77 0.74 0.78 0.31 0.36 0.29 548 533 582 14.2 16.4 17.8 5.92 6.01 6.1 1 0.302 0.312 0.313 2.26 2.28 2.30 11.4 7.5 11.0 8.7 9.9 9.4 30.0 30.6 30.3 69 49 54 Lincoln Mark VI 1980 1.5160 1.5186 1.5248 3 975 4 142 4 262 36 849 37 248 37 791 31 850 30 576 26 742 1485 1873 1201 6 294 6 976 8 386 0.108 0.1 1 1 0.1 13 0.86 0.82 0.71 0.24 0.27 0.31 540 527 56 1 11.8 11.3 11.4 5.93 5.98 6.02 0.349 0.334 0.335 2.37 2.34 2.34 14.7 15.4 13.9 7.8 10.1 9.9 25.2 26.1 25.6 65 54 53 Chrysler Cordoba 1979 1.5158 1.5186 1.5246 4 460 4 735 4 503 37 986 37 023 37 986 31 706 28679 31,199 3 013 3 687 2 389 7 270 12 063 6 967 0.1 17 0.128 0.1 19 0.83 0.78 0.82 0.41 0.31 0.34 5 12 472 490 22.5 22.5 22.2 5.95 5.97 5.98 0.364 0.355 0.362 2.22 2.16 2.18 13.1 11.7 11.9 8.0 12.6 9.7 49.3 48.8 48.0 107 113 105 The precisions of the concentrations of elements ob- tained by ICP-AES were calculated using the measurements from triplicate fragments in the same manner as discussed previously for EDXRF results.These results are given in Table 5. In comparison with EDXRF measurements the ICP-AES results clearly offer a greater number of par- ameters upon which discrimination can be based and better precision in the determination of these parameters. Most elements were determined with an RSD of 1-5% of the mean concentration. Sodium exhibits greater uncer- tainty than the other elements because of its ubiquitous presence in reagents possible heterogeneity in the glass and the fact that an internal standard could not be accommo- dated by the programme used by the instrument manufac- turer for the calculation of Na concentration.The results for Mn had slightly worse precision than other elements because the concentrations of Mn in the glass sample digests were close to the detection limits. The values for precision for the elemental ranges given in Table 5 indicate that all elements except Na offer good discrimina- tion capability. Pairwise comparison of the 81 sheet glass samples using the ICP-AES results was performed using range overlap criteria similarly to those used with EDXRF data. Two samples were judged to be indistinguishable when for all nine elements measured the concentration ranges of the two samples overlapped.Results of painvise comparison of the 81 samples (Table 3) resulted in discrimination of all but three combinations of the samples. There are three samples that are indistinguishable from one another. These are from a Ford Thunderbird a Lincoln Mark IV and a Lincoln Mark V all produced in the model year 1977. Since these models were all made by the Ford Motor Company in the same year the indistinguishable compositions of the glasses may mean that they represent a single source of production. Of the eight pairs of samples representing the same model and year of production six are readily discriminated and two are similar but distinguishable having only one element in each with non-overlapping concentration ranges.The values of frequency of indistinguishability of pairs of samples using each method singly and in various combina- tions shown in Table 3 must be interpreted carefully. The frequencies listed should not be interpreted as population distribution frequencies since the samples were not ran- domly selected as being representative of the entire popula- tion of tempered automobile glass. However they do provide a measure of the relative discrimination capabili- ties of RI EDXRF and ICP-AES results. Clearly both EDXRF and ICP-AES offer improved discrimination whether used alone or in combination with RI measure- ments. It is also to be noted that the composition ranges for glass used in this study were fairly limited. Compositional analysis particularly by ICP-AES becomes a much more powerful method for discrimination among glasses having a different end use but which may have similar optical and physical properties despite widely different elemental compositions.Example Data for Four Sheet Glass Samples In order to provide an example of the use of EDXRF and ICP-AES for discrimination among glass sources four samples having similar optical properties were selected. Results for these samples are given in Table 6 . Individual EDXRF and ICP-AES results are given in order to illustrate the range overlaps which exist. The four glasses represent different makes models and years of automobile yet they all have indistinguishable RI values. The EDXRF results clearly indicate that the glass from the 1978 Chevrolet Nova is distinguishable from the other three glasses because of its lower Sr:Zr values.The other three samples are not discriminated using the EDXRF criteria. Results for ICP- AES indicate clearly distinguishable compositions among the four glasses.456 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY SEPTEMBER 199 1 VOL. 6 Conclusions Both EDXRF and ICP-AES procedures along with RI measurements and other physical and optical properties have a place in the forensic laboratory for discrimination among sources of sheet glass. Energy dispersive X-ray fluorescence offers the advantages of speed non-destructive testing wide availability reasonably good discriminating capability and provides points of comparison which are relatively independent of optical measures.Inductively coupled plasma atomic emission spectrometry offers the advantages of providing quantitative element concentra- tion data applicability to a greater number of elements and improved discrimination at the cost of destruction of the sample. The quantitative multi-element capability of ICP- AES is also advantageous for classification purposes and for comparison of glasses of a wider range of compositions than shown in the limited product use category of glasses studied here. The authors thank S. Brixey for assistance in sample collection and refractive index measurements and D. Ward C. Fiedler B. Hall and L. Deremer for helpful suggestions concerning this research. 1 2 References Coleman R. F. and Goode G. C. J. Radioanal. Chert?. 1973 15 367.Goode G. C. Ward G. A. Brooke N. M. and Coleman R. F. Atomic Weapons Research Establishment Report 02417 1 Aldermaston UK 197 1. 3 Hughes J. C.. Catterick T. and Southeard G. Foresnsic Sci. 1976 8 217. 4 Blacklock E. C. Rogers A. Wall C. and Wheals B. B. Forensic Sci. 1976 7 12 1. 5 Dabbs M. D. G. German B. Pearson E. F. and Scaplehorn A. W. J. Forensic Sci. Soc. 1973 13 281. 6 Haney M. A. J. Forensic Sci. 1977 22 534. 7 Terry K. W. van Riessen A. and Vowles D. J. Micron 1982 3 293. 8 Ryland S. G. J. Forensic Sci. 1986 31 1314. 9 Reeve V. Mathiesen J. and Fong W.. J. Forensic Sci. 1976 21 291. 10 Dudley R. J. Howden C. R. Taylor T. J. and Smalldon. K. W. X-Ray Spectrum. 1980 9 119. 1 1 Wolnik K. L. Gaston C. M. and Fricke F. L. J. Anal. At. Spectrom. 1989 4 121. 12 Koons R. D. Fiedler C. and Rawalt R. C. J. Forensic Sci. 1988 33 49. 13 Catterick T. and Hickman D. A. Forensic Sci. lnt. 1981 17 253. 14 Zurhaar A. and Mullings L. J. Anal. At. Spectrum. 1990 5 611. 15 Hickman D. A. Harbottle G. and Sayre E. V. Forensic Sci. lnt. 1983 23 189. 16 Hickman D. A. Forensic Sci. lnt. 1983 23 213. 17 Miller E. J. in Forensic Science Handbook ed. Saferstein R. Prentice-Hall Englewood Cliffs NJ 1982 ch. 4. 18 OfJicial Methods of Analysis Association of Official Analytical Chemists Arlington VA 14th edn. 1984 p. 936. 19 McCrone W. C. J. Assoc. Off Anal. Chem. 1973 56 1223. 20 Locke J. and Underhill M. Forensic Sci. Znt. 1985 27 247. Paper 1 /00 71 OF Received February 14th 1991 Accepted May 14th 1991