METHOD VALIDATION NFI Scientific Report

Forensic float-glass analysis using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS)

METHOD VALIDATION NFI Scientific Report

Shirly Montero 07 February 2005 version 8

Contents: 1. Purpose ……………………………………………………… 2 2. Introduction ……………………………………………………… 2 3. Experimental Section ……………………………………….…… 3 3.1. Sampling ………………………………………………. 3 3.2.1. Glaverbel single pane………………............... 3 3.2.2. Double windows…………………………......... 5 3.2.3. Store window………………………………… 6 3.2. Instrumental parameters ……………………………….. 7 3.3. Analytical method ………………………………………. 7 3.4. Data processing …………………………………………. 8 3.4.1. Descriptive statistics…………………………. 8 3.4.2. Comparison of variances …………………….. 9 3.4.3. Comparison of means ………………………… 9 4. Results …………………………………………..… 11 4.1. Ruggedness Test……………………………………......... 11 4.2. Limits of detection………………………………….......... 13 4.3. Linearity ……………………………………….….… 14 4.4. Specificity ………………………………………….…. 15 4.5. Single laboratory precision and bias study ……….… 16 4.5.1. Standard reference materials ………………… 16 4.5.2. Small fragments ……………………………… 20 4.5.3.Glaverbel single pane…………………………. 22 4.5.4. Double windows…………………………….... 23 4.5.5. Store window……………………………….… 24 4.6 Uncertainty statement …. ………………………………… 25 5. Conclusions ……………………………………………………..… 25 6. References …………………………………………………………… 27 Appendix A…………………………………………………………….. 30 Appendix B ……………………………………………………………. 39

1

1. PURPOSE The purpose of this study was to develop and validate a method for Forensic float-glass investigations using laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS). The study was executed in two parts. In the first part a procedure was developed and validated for the elemental analysis of float glass by LA-ICP-MS, for the comparison of fragments from a known source to the recovered fragments from a questioned source. The elemental menu comprises 10 elements: potassium (K), titanium (Ti), manganese (Mn), rubidium (Rb), strontium (Sr), zirconium (Zr), barium (Ba), lanthanum (La), cerium (Ce) and lead (Pb). Most validation studies in this report were made using homogeneous international glass reference materials that may be seen as more ideal when compared to actual glass panes as encountered in daily forensic investigations, for example with respect to the homogeneity of the material. In the second part the heterogeneity of some actual float glass panes was measured using the method developed in the first part. On the basis of these results matching criteria were set for the forensic comparison of float glass samples. This procedure is applicable for other glass types and other elements if the appropriate modifications are validated. Other methods for the elemental analysis of float glass are listed in the references. 2. INTRODUCTION Characterization of a glass sample by its elemental composition has been documented previously. Both classification of glasses and discrimination of glasses of the same type are possible in this way.1-5 A wide range of methods has been used for these purposes, among them are atomic absorption spectroscopy (AAS),6-8 spark source mass spectrometry,9 X-ray methods such as SEM-EDS/WDS,10,11 radiochemical methods such as neutron activation (NAA)12,13 and ICP techniques such as atomic emission spectroscopy (ICP-AES) and mass spectrometry (ICP-MS).14-22 For forensic purposes these techniques offer advantages and limitations.14,15,23 When the analyst is confronted with the choice of the technique, several factors need to be considered. Most criminal cases involve comparison of samples so that techniques that render good discrimination14 (i.e. good precision) are more relevant. Procedures based on ICP techniques are fast, have multielemental capability, high sensitivity (ICP-MS more sensitive than ICP-AES) and provide a large dynamic range. In addition, when using ICP-MS, isotopic information can be acquired and if LA-ICP-MS is available, sample destruction is negligible. The method validated here is based on the use of LA-ICP-MS. When comparing two samples for forensic purposes one of the most important aspects is defining the appropriate “matching criteria” for the property being used in such a comparison. The matching criteria define how different the properties can be for the samples and still to be consistent with the samples originating from the same source (“to match”). These criteria depend on the homogeneity of the materials being compared as well as the precision of the method (or uncertainty) used for such a comparison. Glass is an ideal material for comparisons due to its internal homogeneity. However, although the manufacturing of float glass produces a very homogeneous material, small differences within (along and across) a glass pane can be measured by LA-ICP-MS and complicate interpretation of the results. Part of the study below was designed to determine the variations in elemental concentrations

2

within a single pane of float glass. For a single pane (same source), the magnitude of such variations (due to the method and the material) will be related to the “matching criteria”. By investigating a number of different glass panes from various sources we are able to define more general matching criteria as to be used for forensic glass investigations in the Netherlands. 3. EXPERIMENTAL SECTION 3.1. Sampling A variety of glass samples was used for our experiments, both standard glass reference and certified materials as well as samples from ‘real life’ glass panes. All samples were washed in methanol, (reagent grade, Merck, Darmstadt, Germany) for 10 minutes, followed by HNO3 3% (diluted from reagent grade 65%, Merck, Darmstadt, Germany) for 30 minutes, rinsed with deionized water (> 18 Ωcm-1) and air dried before use. Standard glass reference and certified materials were obtained from NIST (National Institute of Standards and Technology, Gaithersburg, MD, USA) and SCHOTT (SCHOTT Glas, Germany, obtained through NITECRIME Network). These samples since delivered in small sizes were directly crushed with a protected pestle and mortar. 1. 2. 3. 4. 5. 6.

Standard Reference Material NIST 614 (trace elements in glass matrix, 1 µgg-1) Standard Reference Material NIST 612 (trace elements in glass matrix, 50 µgg-1) Standard Reference Material NIST 610 (trace elements in glass matrix, 500 µgg-1) Standard Reference Material 1831 (soda-lime sheet glass) Standard Reference Material 1830 (soda-lime float glass) Certified Reference Material FGS02

Glass panes were obtained from three sources: 7. Production samples from Glaverbel Nederland BV, the only producer of float glass in the Netherlands. 8. Double windows were obtained through a local supplier of double windows. 9. As a street sample a glass pane from a door in a store was used. These pane samples were crushed first using a hammer following the schemes as given in the following sections. Then the fragments were further crushed with a protected pestle and mortar. 3.1.1. Glaverbel Single Pane Glass production samples from Glaverbel Nederland BV were used as samples that were obtained directly from a major relevant glass manufacturer. The glass samples were produced 21st July 2002 and supplied by courtesy of Glaverbel. The Glaverbel group has 13 float-glass plants in Europe that are dedicated to the production of flat glass. Glaverbel Nederland BV is the only manufacturer of flat glass (using the float process) in the Netherlands. We do not know the market share of Glaverbel Nederland BV in the forensically relevant markets

3

(mostly house, office and shop windows as well as car windows) in the Netherlands but expect that this producer will be relevant for us, especially for house, office and shop windows. The glass samples consist of 10 rectangles (H1 through H10) from two 10 cm wide strips of glass taken both along (590 cm) and across (323 cm) a single pane during production. The sampling scheme for this study is shown in Figure 1. From each rectangle, 5 fragments were selected, mounted exposing a non-float side (checked by the absence of Sn fluorescence under an UV lamp) and measured once (single spot). Since our method24 limits the number of spots per analysis sequence (a single series of measurements without exchanging sample stubs or trays in the LA sample chamber), 4 sequences (trays) per experiment were required. In total 5 experiments were made and 20 (5*4) sequences were used. Experiments 1, 2 and 3 used the same order of samples but this order was different from the order as used in experiments 4 and 5. Table 1 shows the order in which analyses were performed. Fresh glass fragments were analyzed in each new experiment. H3 10 cm

1

2 2

3 3

4 4

5 5

6 6

7 7

8

9 9

10

10

11

11

H2

H1

12

12

H5 H4

50 cm

40 cm

(a) H7 10 cm

A

B B

H8 C C

D D

E E

F F

G

H6

G H10

H9 50 cm

23 cm

(b) Figure 1. Sampling scheme for the Glaverbel single glass pane a) along the pane and b) across the pane. Red dots indicate the sampling location (samples H1 through H10). Table 1. Measuring scheme for the 4 sequences in each of 5 experiments. Experiments

Sequence

Samples

1, 2, 3

1 2 3 4

H1, H2, H3 H4, H5, H6, H7, H8, H9 H10

4, 5

1 2 3 4

H4 ,H9, H2 H1, H6, H10 H3, H5, H8 H7 4

3.1.2. Double windows For this part of the study glass panes from two double windows (A and B) were analyzed. Both windows were provided by a Dutch supplier, Spliet & De Waal, and were both produced within a 24 hours period as confirmed by this supplier. The glass thickness for the inner pane was 4 mm and for the outer pane 5 mm for both windows. The sampling scheme for window A is shown in Figure 2. From each location 3 fragments were analyzed and one LA-ICP-MS analysis was performed for each fragment. Sample A8 was not collected by mistake. For each pane, all samples were analyzed in a single sequence.

75cm

Z

W

T

Y

V

S

X

U

R

(a) 9

6

3

8

5

2

7

4

1

(b) Figure 2. Sampling scheme for double windows from a single supplier (a) inner pane (letters) and (b) outer pane (numbers). Blue dots indicate the sampling location (samples R through Z and 1 through 9).

5

3.1.3. Store window For this part of the study, glass pane fragments from a front door in a store were used. This pane was broken during an attempted burglary of the store. Large glass sherds still present in the door frame were collected at 5 main locations and numbered 1 through 5 (see Figure 3). Because of the difference in sizes of these (large) glass sherds, in total 18 different locations were sampled on the sherds. At each position 6 glass fragments were collected and each fragment was measured once. In total 6 analysis sequences were used with the samples ordered as listed in Table 2.

0.63 m

4 A 1.58 m 4 B 2 B

4 C

2 A

3

5 1

Figure 3. Sampling scheme for the store window. Table 2. Measuring scheme with the sequences followed for the analysis of the store window. Sequence

Samples

1 2 3 4 5 6

1.1,1.2, 1.3 2A1, 2A2, 2A3 3.1, 3.2, 3.3 2B1, 2B2, 2B3 4A, 4B1, 4B2 4C1, 4C2, 5

6

3.2. Instrumental Parameters Exact values for the instrumental parameters can only be given for the laser ablation system. The ICP-MS system needs to be optimized on a day-to-day basis according to the protocol designed for its use25. However, a typical set of parameters including those for the laser is shown in Table 3. Table 3. Typical instrumental parameters LA Ablation Cell

New Wave Research UP 213nm aperture imaged Nd:YAG laser Standard NWR cell, 30 cm3

Crater Diameter

60 µm

Repetition Rate

10 Hz

Fluence

12 Jcm-2 (0.4 mJ)

ICPMS

ELAN DRC plus operating in standard mode

Carrier Gas

0.889 Lmin-1 He

Nebulizer Gas Flow

0.95 Lmin-1 Ar

Auxiliary Gas Flow

0.90 Lmin-1 Ar

Plasma Gas Flow

15 Lmin-1 Ar

RF Power

1350 W

Tubing Length

323 cm

These conditions reflect a doubly-charged ratio Ba++/Ba+ and an oxide ratio CeO+/Ce+ less than 3 %. The tuning is usually performed using NIST 612 and FGS02 and according to the laboratory protocol for the use of the ICP-MS and LA-ICP-MS.24,25 3.3. Analytical method The analytical method used for this study is described in Appendix A,24 unless stated otherwise. The glass fragment size used was normally in the range of 0.1 to 0.5 mm. The limit of size to a minimum of 0.1 mm in each dimension was set to avoid breakage or ablating through the fragment. In addition, at least one crater per fragment is guaranteed this way.

7

3.4. Data Processing The software package used to reduce the data was GLITTER (McQuarie, Australia), using SRM NIST 612 as the calibrator and 29Si as internal standard (concentration almost constant in float glasses . All the results obtained using our method are elemental concentrations in µgg-1 in the glass samples. A personal computer equipped with SYSTAT (SPSS Corp., v10) and Excel 2003 (Microsoft Corp., v11.6113.5703) was used for all statistical analyses of the data. The descriptive statistics for the data (e.g. arithmetic mean <x>, the standard deviation s and relative standard deviation rsd) and the graphs were calculated using Excel, while the multiple pair comparisons (comparison of multiple means) and analysis of variance (ANOVA) calculations were performed using SYSTAT (General Linear Model with fixed Mean Squared Error). 3.4.1. Descriptive Statistics Sample sets or subsets are described by the arithmetic mean, the standard deviation and relative standard deviation (percentage). The arithmetic mean for a set of data {x1, x2, …, xn} is calculated by26 n <x> = 1/n ∑ xi (1) i=1 The standard deviation for the entire set or population is n σ = [1/n ∑ (xi - <x>)2]1/2 i=1

(2)

and for a smaller set or sample of the population the standard deviation (s) can be calculated as n s = [1/(n-1) ∑ (xi - <x>)2]1/2 i=1

(3)

The variance, a measure of the dispersion of the data about the mean, is the square of the sample standard deviation: n 2 s = 1/n ∑ (xi - <x>)2 (4) i=1 and the relative standard deviation (rsd) in percentage is calculated as %rsd= s/ <x> *100

(5)

8

When a set of specific measurements is normally distributed, approximately 68 % of the data is expected to be in the interval <x> ± s, 95 % of the data is expected to be in the interval <x> ± 2s, and 99.7% of the data is expected to be in the interval <x> ± 3s.27 3.4.2. Comparison of Variances The F statistic was employed to compare variances according to equation 6.27 Whenever the calculated F value was larger than the critical value, the null hypothesis that states the equivalence of the variances (s12 = s22) was rejected. The F statistic is well known and the critical values are found tabulated in almost every introductory statistical book. F ν1, ν2 = s12/s22

for s12 > s22

(6)

where F ν1, ν2 is the calculated F value, s12 and s22 are the variances for samples 1 and 2, and ν1 and ν2 are the degrees of freedom associated with s12 and s22, respectively. 3.4.3. Comparison of Means Pair-wise comparisons (multiple pairs) in the selected sets were carried out using analysis of variance (ANOVA) in the General Linear Model (GLM) tool from SYSTAT. The raw data were analyzed without averaging since the individual replicates are required in the calculations. The mean squared errors for within the samples (MSE) and mean squared treatments between the samples (MST) are provided by the output of the ANOVA28 calculations. The null hypothesis of equal means for within and between the samples was tested using an F-test for the ratio MST/MSE. The results of an ANOVA calculation served to indicate whether these multiple means differ significantly26 but without identifying which of the means were significantly different. For that purpose, Tukey’s post hoc test was selected to determine which pairs of means differed significantly (p< 0.05). This test is a powerful post hoc test for a large number of pair comparisons and uses the studentized range statistic. Tukey’s test defines confidence values based on the mean square error (MSE) within k groups of n replicates as shown in equation 7.29 (<x>i - <x>j ) ± T k,nk-k,1-p (MSE/n)1/2

(7)

where T is the 100(1-p)% point of distribution of the studentized range with k and nk-k degrees of freedom. For each set of samples, the comparison of means was performed by analyzing each element individually, since a difference in one element will be reported as a difference in the samples. The output of Tukey’s test included a matrix with the probabilities associated with the acceptance or rejection of the null hypothesis. When this probability was less than 0.05 (pvalue), the null hypothesis was rejected and the pair in question was considered distinguishable. Tukey’s approach for multi comparisons minimized the rejection of a true hypothesis or Type I errors that would be generated if the multiple t-tests performed were not protected.29 For Type II error or the acceptance of a false hypothesis, the confidence level was fixed to 95% (p=0.05). 9

In order to evaluate the variation of the concentrations and to be able to set appropriate matching criteria the data for each set of samples was analyzed separately. Based on preliminary experiences, the values used as the MSE in the GLM model were a- the averaged squared standard deviation of the results per location (1sd)2 b- the average of the squares of double the standard deviation for these values (2sd)2 The objective of this procedure (as mentioned in the introduction) was to obtain the values (matching criteria) related to the dispersion of the data (including the uncertainty of the measurement), that provide an optimal balance between Type I and Type II errors, e.g. small enough to minimize Type II errors and large enough to minimize Type I errors.

10

4. RESULTS 4.1. Ruggedness Test A ruggedness test on a wafer of SRM NIST 612 (ideal glass reference material with trace elements at a 50 µgg-1 level) was made following the standard guide ASTM E1169-8930 in order to evaluate the significance of the effect produced on the concentrations when the experimental parameters are varied slightly. A Plackett–Burman design (Table 4) of 8 experiments was used to evaluate such effects. The four parameters studied were frequency of the laser, energy output, diameter of the beam and the nebulizer flow. The changes in the levels of the factors were chosen to be relatively small, so that the ruggedness test have less interactions or unimportant ones compared with the main effects. To complete the design three dummy parameters were used (and allowed for by the guide for this design). The designs were conducted in duplicate and repeated on a different day to include possible dayto-day variation (four full designs in total). Table 4. Placket-Burman design for the ruggedness test (The ‘-’ sign represent the lower value and the ‘+’ the higher value) experiment

frequency, Hz

energy, %

diameter, µm

1 2 3 4 5 6 7 8

12 8 8 12 8 12 12 8

100 100 90 90 100 90 100 90

70 70 70 50 50 70 50 50

nebulizer flow, Lmin-1 0.96 1.00 1.00 1.00 0.96 0.96 1.00 1.00

a

b

c

+ + + + -

+ + + + -

+ + + + -

The ruggedness of the method was tested for all the elements and isotopes selected for the original menu (based on a glass round robin experiment of the NITECRIME Network31): 7Li, 11 B, 23Na, 24Mg, 25Mg, 27Al, 28Si, 39K, 42Ca, 49Ti, 55Mn, 57Fe, 59Co, 65Cu, 66Zn, 71Ga, 75As, 85 Rb, 88Sr, 90Zr, 93Nb, 118Sn, 123Sb, 133Cs, 137Ba, 139La, 140Ce, 147Sm, 151Eu, 159Tb, 165Ho, 169Tm, 175 Lu, 178Hf, 181Ta, 182W, 197Au, 206+207+208Pb, 209Bi, 232Th, 238U. For the results given below, the specific isotope of the element of interest will be omitted unless this might lead to confusion. Each of the experiments described below was comprised of 6 different spots ablated at the conditions set for that particular experiment. The first, second and the sixth spot were used as standards and the third, fourth and fifth spots were used as samples. All the calculations used are the ones as described by the standard guide.30 The data were separated into two sets. Set A were data measured in two different days at the start of the day and Set B measured in two different days after using the instrument a couple of hours. The Li results from set A are listed in Table 5. According to the calculations, the effects on the concentration of Li when varying the parameters between the two previously set levels are not significant (p=0.05 or 0.01). Table 6 lists the calculated t values. The critical values are 3.18 (p=0.01) and 5.38 (p=0.05).

11

Table 5. Results for the measurement of Li concentrations (µgg-1) on two different dates at the start of the day (set A). date

experiment 1 1 1 2 2 2 3 3 3 4 4 4 5 5 5 6 6 6 7 7 7 8 8 8

20030507 40.26 40.15 39.72 44.04 44.98 43.28 46.51 35.78 49.58 38.61 47.25 40.34 36.22 42.99 36.00 42.29 42.87 41.98 42.21 38.08 36.08 47.12 46.89 48.44

20030508 42.79 35.75 38.50 37.89 37.53 43.79 47.58 42.00 40.78 40.43 39.11 36.84 47.98 53.90 40.61 38.64 42.18 37.62 38.61 41.77 44.69 35.98 34.68 39.70

40.71 44.10 43.96 42.07 38.40 42.38 38.79 47.48

39.01 39.74 43.45 38.79 47.50 39.48 41.69 36.79

average 1 2 3 4 5 6 7 8

Table 6. Significance of the effects measured (µgg-1) in the ruggedness test for Li in set A. Factor freq freq ener ener diam diam n.flow n.flow a a b b c c

Level 12 8 100 90 70 50 1.00 0.96 + + + -

average 40.99 43.49 40.50 43.97 42.79 41.69 42.23 42.24 41.28 43.19 41.74 42.74 40.88 43.59

Effect -2.50 -3.47 1.10 -0.02 -1.90 -1.00 -2.71

average 39.74 41.87 41.98 39.63 40.42 41.19 40.92 40.69 41.19 39.42 41.38 40.24 43.03 38.58

effect

difference(d)

average effect

s average

tcalculated

-2.12

-0.38

-2.31

2.03

1.14

2.36

-5.83

-0.56

2.03

0.27

-0.77

1.87

0.17

2.03

0.08

0.22

-0.24

0.10

2.03

0.05

2.77

-4.67

0.43

2.06

0.21

1.14

-2.14

0.07

2.03

0.04

4.45

-7.16

0.87

2.03

0.43

12

The results can be summarized as follows: For set A (morning), 24Mg and Sm results were significantly affected by the variation of frequency (p=0.01), K results were significantly affected by the variation of energy and W results were significantly affected by the variation of the nebulizer flow (p=0.01). For set B (afternoon), several elements were significantly affected by the variation in frequency (Ca, Zn, Ta and Th p=0.01), in energy (Zr and Eu p=0.01), in diameter of the beam (Si, Sr, Zr, Eu and Th p=0.01) or in the nebulizer flow (Cu, Sr, Ta and U p=0.01). For the series of measurements taken on 7th May 2003, several elements were significantly affected by the variation in frequency (Zn, Nb, Sb, Lu, Ta and Th p=0.01), in energy (24Mg, K and La p=0.01), in diameter of the beam (Al, Nb, Sb and Ho p=0.01) or nebulizer flow (Sb p=0.01). For the series of measurements taken on 8th May 2003, two elements were significantly affected by the variation in frequency (Na and Zn p=0.01) and one by the diameter of the beam (Zr p=0.01). Only Zn was significantly affected by the variation of the frequency in three of the combinations studied (comparison of morning/afternoon on two different days and comparison of two dates in morning data). In general, the effects were very small and similar to the variation obtained from the three spots in the same experiment (within the experiment variation). If the confidence level is decreased to 95% (p=0.05), then the influence of the variation of the parameters studied in this ruggedness test on the elemental concentrations is insignificant. Our final menu was selected on the basis of accuracy ( |bias| < 5 µgg-1 and |relative bias| < 10%) and precision ( |standard deviation| < 3 µgg-1 and |relative standard deviation| < 10%) using SRM NIST 612. The selected isotopes are 39K, 49Ti, 55Mn, 85Rb, 88Sr, 90Zr, 137Ba, 139La, 140 Ce and 206+207+208Pb (or sumPb). In addition, 118Sn was included in the method in order to check independently for float glass samples that the ablated side is not the float side. 4.2. Limits of detection The actual detection limits of the method depend on the operational conditions of the instrument. However, the values presented in Table 7, were used as reference since they were obtained as the average of several measurements. The values were calculated by Glitter as three times the standard deviation of the noise measured while the laser was firing but not ablating the sample. The limits of quantification were calculated by multiplying the limits of detection by 3.3 (10/3).

13

Table 7. Typical limits of detection and quantification for the method. element K Ti Mn Rb Sr Zr Ba La Ce Pb

limit of detection LOD, µg g-1 15 17 3 1 0.2 0.4 2 0.2 0.2 1.5

limit of quantification LOQ, µg g-1 50 57 10 3 0.7 1.3 7 0.7 0.7 5

4.3. Linearity There are only a small number of glass standards that are certified for the elements of interest at the trace level. This section only considered the SRMs 614, 612 and 610 from NIST. Whenever the concentrations were not certified, the values used were obtained from concentrations reported in the literature.32,33 As an example, Table 8 and Figure 4 show the data for Sr using the Si concentration (~34 x 104 µgg-1 Si for all samples discussed in this report21,32) as an internal standard. The data pair (0,0) was included in all plots. Appendix B contains the results for all other elements. Table 8. Linearity of the method for the measurement of the concentration of Strontium Glass, NIST 614 612 610

Sr/Si 0 0.022 0.041 0.281

Sr/Si standard deviation, s

Sr/Si uncertaintya

0.00008 0.00066 0.00044

0.0002 0.0016 0.0011

reported concentration Sr in µgg-1, [Sr]b 0 45.8 78.4 497.4

[Sr] uncertaintyb µgg-1

0.1 0.2 0.5

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b certified by NIST

14

600

y = 1756x + 4 2 R = 0.9998

Concentration, µgg

-1

500 400 300 200 100 0 0.00

0.05

0.10 88

0.15

0.20

0.25

0.30

29

Sr signal ratio to Si

Figure 4. Linearity for the determination of the concentration of Sr using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610. 4.4. Specificity Because of the nature of the technique, the specificity should require the preparation of solid glass standards with only the elements of interest. There is no such validated standard at the moment. However, in order to check for specificity a high-resolution (HR) instrument was used to measure SRM NIST 612 for all elements except Rb (Rb became part of the menu at a later time). The results can be compared in Table 9 with those measured with a quadrupole (Q) ICP-MS. These biases are les than the 10% limit defined when designing the method and therefore they are accepted. Table 9. Biases in elemental concentrations for NIST 612 compared with LA-HR-ICP-MS. SRM NIST 612 element K Ti Mn Sr Zr Ba La Ce Pb a

LA-Q-ICP-MS, -1

(µgg ) 70.9 50.2 40.4 80.8 37.8 41.1 38.6 41.4 42.5

LA-HR-ICP-MSa, (µgg-1) 69.9 47.7 37.9 76.4 37.1 37.6 36.3 38.2 38.6

bias, % 1.5 5.3 6.6 5.7 1.8 9.3 6.3 8.4 10

measured in OCAS facilities (Zelzate, BE)

15

4.5. Single laboratory precision and bias study. The terms of repeatability, reproducibility and bias used below are those defined by the standard practice ASTM E 177-90a34 and the standard ASTM E 456-96.35 Since there were no between laboratory measurements, the repeatability and reproducibility of the test method cannot be calculated. However, with the results presented below, estimates of withinlaboratory precision can be made (uncertainty of the measurement and heterogeneity of the material) according to the standard practice ASTM E 177-90a.34 4.5.1 Standard reference materials This study was done parallel to a population variability study (paper in progress), for which 203 glass samples representative of forensic casework glass materials were embedded in a two component epoxy stubs (total of 47 stubs, approximately 7 glass fragments per stub). Table 10. Elemental concentrations in µg g-1 for different glass SRM’s(n: number of experiments) NIST 1830 (n=47)

NIST 1831(n=23)

NIST 612 (n=47)

element

average

s

rsd, %

average

s

rsd, %

average

s

rsd, %

K

301.2

23.6

7.8

3005

305

10

70.9

5.4

7.6

Ti

74.2

7.0

9.5

146.1

8.7

6.0

50.2

6.0

12

Mn

10.1

0.6

6.3

13.6

0.9

6.3

40.4

1.9

4.6

Rb

<1

6.4

0.4

5.7

33.9

2.1

6.1

Sr

52.0

2.7

5.1

96.2

5.8

6.0

80.8

3.3

4.1

Zr

83.4

4.5

5.4

43.6

4.1

9.4

37.8

1.7

4.5

Ba

12.1

1.0

8.5

33.4

2.3

7.0

41.1

2.4

5.9

La

1.6

0.1

8.1

2.6

0.2

6.3

38.6

2.2

5.7

Ce

2.7

0.3

9.5

4.9

0.4

7.2

41.4

2.3

5.5

Pb

1.6

0.03

2.2

1.9

0.2

11

42.5

3.2

7.5

yellow: above limit of detection but below limit of quantification red: below limit of detection (if a number was given is calculated with the values above the limit)

16

One fragment of glass SRM NIST 612 and one fragment of glass NIST 1830 were included in all stubs (n=47), and in 23 stubs a fragment of glass NIST 1831 (n=23) was also included. Each stub was measured using a separate sequence. The measurements of the NIST 612 were comprised of 5 different spots ablated at the conditions set by the protocol.24 The first, second and the fifth spot were used as calibration standards for the sequence and the third and fourth spots were used as samples. For the NIST 1830 and NIST 1831 samples three spots were ablated in each stub. The within-laboratory precision as defined in section 23.1.2 of the standard practice ASTM E 177-90a34, can then be determined using the values obtained for each of the reference glasses measured in different sequences (several per day), on different days and by different operators. The concentration values obtained are listed in Table 10. The variability of Sr, Ti and Zr concentrations is shown in Figures 5 through 7. NIST 612 120.0

Concentration, µ gg

-1

100.0

80.0 Ti49 60.0

Sr88 Zr90

40.0

20.0

0.0 1

4

7

10

13

16

19

22

25

28

31

34

37

40

43

46

Expe rime nt

Figure 5. Variability in the concentration of Ti, Sr, and Zr over time for SRM NIST 612.

17

NIST 1830 120.0

concentration, µ gg -1

100.0 80.0 Ti49 Sr88

60.0

Zr90 40.0 20.0 0.0 1

4

7

10

13

16

19

22

25

28

31

34

37

40

43

46

Experiment

Figure 6. Variability in the concentration of Ti, Sr, and Zr over time for SRM NIST 1830.

NIST 1831 200.0 180.0 Concentration, µ gg-1

160.0 140.0 120.0

Ti49

100.0

Sr88

80.0

Zr90

60.0 40.0 20.0 0.0 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 45 46 47 Experiment

Figure 7. Variability in the concentration of Ti, Sr, and Zr over time for SRM NIST 1831.

18

From Table 10 it can be seen that the variation in the concentrations measured (rsd, %) is mostly < 10 %. Exceptions are some measurement results that are above the detection limits but below the quantitation limits (Pb with all 23 replicates for NIST 1831 and Ti with 43 replicates for NIST 612). The variation in measurement results that are below detection limits (Rb with 47 replicates and Pb with 36 replicates for NIST 1830) is not taken into account since no significance can be attributed to the actual values. These values, therefore, are acceptable. It should be mentioned that, in general, the analysis of fragments in all 47 stubs took more than one sequence per day, more than one day and were performed by at least two operators. With the measurements reported here, only the within-laboratory variation is considered. If required, however, the measurements can also be characterized using the index of precision following ASTM E 177-90a.34 That is, the measurements have a coefficient of variation in the concentration of ± ts/n½ s (2.5s for 3 replicates an 95 % confidence level), where the s values are listed per element in Table 10. Since most of the target samples are float glass, the results for NIST 1830 can be used for them. Only a few values are available in the literature for the trace elemental concentrations in glasses NIST 1830 and NIST 1831.32 The biases measured for the specific elemental concentrations in the different glasses are listed in Table 11. The large variation in the concentrations of Ti for NIST 1830 and NIST1831 may be partly due to the use of NIST 612 as a calibrator since its values are close to the limit of quantification (57 µgg-1). The last three columns contain the data for NIST 612 measured as samples. It is shown here to illustrate the reliability of the procedure and software once the calibration is made. When the concentrations of FGS02 are available (certified), this glass reference material can be used as a calibrator and to evaluate the bias on the NIST 612 measurements.

Table 11. Relative biases in elemental concentrations for different glass SRMs. (Elemental concentrations are in µgg-1). NIST 1830

a

NIST 1831

element

average

rep.a

bias, %

K Ti Mn Rb Sr Zr Ba La Ce Pb

301.2 74.2

300 66

0.4 12

NIST 612

average

rep.a

bias, %

average

rep.a

bias, %

3002 146.1 13.6 6.4 96.1 43.4 33.4 2.6 4.9 1.9

2700 110 15.0 6.11 89.1 43.36 31.52 2.3 4.5 1.99

11 33 -9.4 0.3 7.9 0.02 5.9 10 8.0 -4.3

70.9 50.2 40.4 33.9 80.8 37.8 41.1 38.6 41.4 42.5

66.26 48.11 38.43 31.63 78.4 35.99 37.74 35.77 38.35 38.96

7.0 4.3 5.1 7.3 3.1 4.9 9.0 7.9 8.0 9.1

rep.= reported value

19

4.5.2. Small fragments The previous results were obtained with fragments embedded in epoxy stubs as mentioned in section 4.5.1. After polishing, these fragments exposed a surface of at least 10 mm2. The procedure of embedding is very laborious and not always possible with recovered fragments submitted as evidence. An alternative sample preparation procedure to analyze particles of smaller size > 1 mm3 was proposed (Appendix A) and the LA ICPMS results for these small fragments were compared with the results for the larger fragments. Figure 8 shows the comparison of the results obtained for the SRM NIST 612. This standard comes in an oval wafer form of approximately ~1.35 cm2 by 3 mm thickness. For the comparison this material was analyzed in the wafer form, embedded in an epoxy stub and finally two larger fragments were broken into smaller fragments (~1 mm2) and mounted as described in Appendix A. The error bars shown correspond to one standard deviation.

NIST612

90.0

wafer stub fragments1 fragments2 reported values

Concentration, µgg-1

80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 K39

Ti49

Mn55

Rb85

Sr88

Zr90

Sn118

Ba137

La139

Ce140

Pb208

isotope

Figure 8. Comparison of results using samples of SRM NIST 612 of different sizes. The reported values were also added in the plot for accuracy evaluation. In addition to the comparison of ranges defined by 1s (plots), the results were compared using ANOVA as described in section 3.4.3. Both methods for comparison support that there are no significant differences of the means for the different sizes for all the elements included in the method. However, precision of the results was best for the embedded samples for most elements. In order to evaluate the effect of sample size and preparation on a less homogeneous glass, the study included also SRM NIST 1830 (float glass). The results shown in Figure 9 include samples of 1830 embedded in epoxy stubs (47) and polished, as well as 2 sets of smaller fragments. The data for K, Ti , Sr and Zr were divided by a factor (25, 10, 10 and 10, respectively) 20

to include all the isotopes in the same plot with a convenient concentration scale. As for the SRM NIST 612 glass, the results from small fragments of float glass following the method described in Appendix A lead to results that do not significantly differ from the results obtained for fragments embedded in epoxy stubs. However, the spread of the data could not be compared directly since for the 47 stubs the results presented were obtained on different days and by different operators while the smaller fragments were analyzed within one day by the same operator.

NIST1830

14.0

Concentration, µgg-1

12.0

stub

x25

fragments1 fragments2

10.0 x10

x10 8.0 x10 6.0 4.0 2.0 0.0 K39

Ti49

Mn55

Rb85

Sr88

Zr90

Ba137

La139

Ce140

Pb208

isotope

Figure 9. Comparison of results using samples of SRM NIST 1830 (float glass) of different sizes. We find that embedding of the glass samples and polishing is the method of choice for sample preparation since it provides higher precision of the LA ICPMS results. However, if the fragments are too small for this method, the sample preparation method described in Appendix A proved to be a good alternative method with results that do not significantly differ from the results obtained using the larger samples.

21

4.5.3 Glaverbel single pane All measured elemental concentrations were compared using multiple pair comparisons as explained in section 3.4.3. The results can be divided according to options a- and b- from that section a- the averaged squared standard deviation of the results per location (1sd)2 b- the average of the squares of double the standard deviation for these values (2sd)2 When using option a- (smaller variation) differences were noted among the sequences and the samples. For example, using the titanium (Ti) concentration in experiment 1, results for sequences 1 and 3 were indistinguishable but results for sequences 2 and 4 were different from results for all other sequences. In addition, significant differences were observed for some samples within a single analysis sequence. For example, within a single experiment glass from locations H1 and H2 are indistinguishable but glass from location H3 is differentiated from these two (see values with the red color in Figure 10a). However, no effect was observed due to the position in the analysis sequence. For a summary of the Ti results see Table 12. When using option b- (larger variation), the samples were not differentiated in any of the experiments (see overlap of ranges in Figure 10b). Upon inspection of the data showed no clear correlation between the position on the pane and the elemental concentrations (for example a consistent increase of an elemental concentration along the length of the pane, going from H1 to H5). We therefore conclude that the variation over the glass pane will be contained in an interval defined by two standard deviations (2s) around the mean and the matching criterium for this experiment should be set as ±2s. Table 12. Concentrations of Ti in µgg-1 in Experiment 1. The different colors represent different sequences of analysis. Ti

average, µgg-1 sd , µgg-1 rsd, %

H1 315.1 307.4 302.3 293.4 292.5

H2 310.3 279.3 301.4 291.4 285.4

H3 279.5 264.7 277.5 268.8 263.7

H4 302.6 329.7 305.9 319.0 283.4

H5 323.4 306.0 321.2 315.4 303.1

H6 302.1 318.5 312.1 326.8 296.6

H7 289.7 282.0 291.2 282.0 306.8

H8 297.3 296.8 308.3 307.0 272.8

H9 291.5 303.7 298.9 288.2 274.3

H10 279.2 252.5 245.1 265.3 255.3

302.1 9.6 3.2

293.6 12.4 4.2

270.9 7.3 2.7

308.1 17.6 5.7

313.8 9.0 2.9

311.2 12.2 3.9

290.3 10.1 3.5

296.4 14.2 4.8

291.3 11.3 3.9

259.4 13.2 5.1

22

Ti, µ gg

-1

450.00 400.00

sequence 1

350.00

sequence 2

300.00

sequence 3

250.00

sequence 4 200.00 150.00 100.00 0.00

0.50

1.00

1.50

2.00 Rb, µgg

2.50

3.00

3.50

4.00

-1

a) 450.00 400.00 sequence 1

Ti, µgg-1

350.00

sequence 2

300.00

sequence 3

250.00 sequence 4

200.00 150.00 100.00 0.00

0.50

1.00

1.50

2.00 Rb, µgg

2.50

3.00

3.50

4.00

-1

b) Figure 10. Concentrations of titanium (Ti) and rubidium (Rb) in µgg-1 in a single pane with the error bars set equal to a) s (standard deviation) and b) 2s. 4.5.3. Double windows The results for the double windows were different to the results for the Glaverbel single pane as described above in that whenever the value of the average of the squared standard deviations s2 was used as the MSE, there were only a few significant differences among the samples for some elements (eg. zirconium in outer pane of A). The variation in results for these experiments (using smaller window sizes as compared to the Glaverbel pane) appears therefore to be smaller. When the MSE value was set to the average of the square of twice the standard deviations 4s2, no Type I errors at all were encountered. Using this criterium panes of a single window could be distinguished for both windows. No significant differences were observed, however, when the inner panes of both windows were compared as well as when the outer panes of both windows were compared. This may be interpreted by being attributed

23

to similar elemental profiles for glass from a single glass manufacturer within a few hours of production. For the results observed for Ti and Rb see Table 13 and Figure 11. Table 13. Concentrations of Ti in µgg-1 for the outer pane of the double window A A1 207.9 213.8 198.4 206.7 7.8 3.8

Ti

average, µgg-1 sd, µgg-1 rsd, %

A2 212.4 193.6 204.4 203.4 9.4 4.6

A3 211.1 198.9 217.0 209.0 9.3 4.4

A4 199.9 224.5 218.7 214.4 12.9 6.0

A5 208.1 203.3 216.0 209.1 6.4 3.1

A6 216.3 206.8 213.6 212.2 4.9 2.3

A7 222.4 210.3 212.0 214.9 6.6 3.0

A9 223.1 212.5 203.0 212.8 10.0 4.7

600 A.R A.T A.V A.X A.Z A.2 A.4 A.6 A.9 B.S B.U B.W B.Y B.1 B.3 B.5 B.7 B.9

500 inner pane Ti, µ gg -1

400

300 outer pane 200

100

A.S A.U A.W A.Y A.1 A.3 A.5 A.7 B.R B.T B.V B.X B.Z B.2 B.4 B.6 B.8

0 0

2

4

6

8

Rb, µ gg -1

Figure 11. Concentrations of titanium (Ti) and rubidium (Rb) in µgg-1 in a 4 panes from 2 double windows. (Error bars equal to 1 standard deviation).

4.5.4 Store window For the store window the variation in the elemental composition across the window was larger than for the two other glass sample sets in 4.5.2 and 4.5.3. However, no effect was observed due to the position in the window (no tendency for higher or lower values across the window). When the MSE value was changed to the average of the square of twice the standard deviations 4s2, no Type I errors were encountered. For the results observed for Ti and Rb see Figure 12.

24

450.0 '1.1'

'1.2'

'1.3'

2A1

2A2

2A3

2B1

2B2

2B3

3.1

3.2

3.3

4A

4B1

4B2

4C1

4C2

5

400.0 350.0

Ti, µ gg

-1

300.0 250.0 200.0 150.0 100.0 50.0 0.0 0.00

1.00

2.00

3.00

4.00 Rb, µ gg

5.00

6.00

7.00

8.00

-1

Figure 12. Concentrations of titanium (Ti) and rubidium (Rb) in µgg-1 at 18 locations of a store windows (Error bars equal to 1 standard deviation). 4.6. Uncertainty statement Summarizing the results obtained in the previous sections, individual concentrations will be reported considering a coefficient of variation in percent concentration of ± ts/n½ times the standard deviations. This is, for triplicates and a 95% confidence level, the concentration will be given as the mean of the replicates ± 2.5 s. If the objective is to compare samples analyzed within a small time frame, then it was demonstrated that variations in the results due to the precision of the method or the homogeneity of the sample could be expressed as twice the standard deviation. This is, discrimination will be defined as the absence of an overlap of ranges defined for each samples as the mean of replicates ± 2 s. 5. CONCLUSIONS A method based on the LA-ICP-MS technique was developed and validated for the forensic investigation of float glass samples. The validation was done through the quality characterization of the method such as ruggedness, limits of detection, specificity, linearity, precision and bias. Then the method was used to analyze three different sets of common float glass. With a technique with high lateral and depth resolution such as LA-ICP-MS, small differences in the elemental composition within (along and across) a glass pane can be measured. It was found that the variation within a single pane of float glass was measurable by the method developed here. This is, the results obtained by this method could erroneously be interpreted as different sources of glass if the variation within a single pane of float glass is not taken into account. When defining an appropriate “matching criterion” for the comparison of two glass samples for forensic purposes, there is always the balance between defining a criterion that is too small, increasing the chance of a false exclusion (Type I), or too large and therefore in-

25

creasing the chance of a false inclusion (Type II error). Both errors have serious consequences associated with them. While a Type I error may result in letting the guilty go free, a Type II error may lead to an incarceration of an innocent. The measurable variation of elemental concentrations across glass panes means that for an optimal forensic investigation more information is required on what happened during the breaking of the window, e.g. was a single or double window broken, what was the size of the window and the size of the hole? More detailed information should also be collected on the elemental variation across the pane such as by collecting and measuring a wider range of glass fragments from the broken pane (as e.g. retrieved from the window frame). In the absence of this information from the above experiments it appears reasonable to use a variation of ±2s around the mean as a matching criterion. With this criterion no Type I errors were observed. This matching criterion applies for every element for all pairs of samples compared. Special care was taken to ensure that the standard deviations were always under control (≤10% usually ≤ 5%). Further studies will be appropriate for the evaluation of Type II error when these criteria are used. However, this is an iterative process only possible by the accumulation of data through the analysis of samples known to originate from different sources and interpreting the results with the accepted matching criteria. The experience accumulated so far allows for the confident use of the matching criteria proposed here as derived from the results of the three different sets. Finally, we propose to define guidelines on how to report the results of the comparisons. Possible hypotheses on the occurrence of the evidence are, 1. 2. 3. 4.

The samples originate from the same pane (window) The samples originate from different panes but the same manufacturer and any correspondences in observed results are based on chance The samples originate from different panes and different manufacturers but they are the same type of glass, any correspondences in observed results are based on chance The samples originate from different panes and different manufacturers of different types of glass (comparison with the world glass population), any correspondences in observed results are based on chance

If the elemental profiles of the glass samples examined are distinguishable (the concentration ranges do not overlap for at least two elements), then one can state that there is not evidence favoring hypothesis one over the others. However, when the elemental profiles are undistinguishable (the concentration ranges do overlap for at least 9 of the 10 elements), we have a “match”. Then, based on literature values32 and the experience gained so far within NFI, this researcher feels confident to report that the weight of the evidence favoring hypothesis one over hypothesis two is weak but it is strong (very strong if 10 elements overlap) favoring hypothesis one over three and very strong (extremely strong if 10 elements overlap) favoring hypothesis one over four. The weight scale used here is in increasing order: weak, strong, very strong and extremely strong.

26

5. REFERENCES 1. Hickman, D.A. Forensic Sci. Int. 1981, 17, 265-81. 2. Ryland, S. J. Forensic Sci. 1986, 31, 1314-29. 3. Hickman, D.A. Forensic Sci. Int. 1987, 33, 23-46. 4. Koons, R. D., Fiedler, C., Rawalt, R.C. J. Forensic Sci. 1988, 33, 49-67. 5. Hickman, D.A. Anal. Chem. 1984, 56, 844A-852A. 6. Hughes, J.C.; Catterick, T.; Southeard, G. Forensic Sci. 1976, 8, 217-27. 7. Catterick, T.; Wall, C.D. Talanta. 1978, 25, 573-7. 8. Hickman, D.A.; Harbottle, G.; Sayre, E.V. Forensic Sci. Int., 1983, 23, 189-??? 9. Haney, M. J. Forensic Sci. 1977, 22, 534-44. 10. Reeve, B.; Mathiesen, J.; Fong, W. J. Forensic Sci. 1976, 21, 291-306. 11. Andrasko, J.; Maehly, A. J. Forensic Sci. 1978, 22, 250-62. 12. Coleman, R.F.; Goode, G.C. J. Radioanal. Chem. 1973, 15, 367-88. 13. Pitts, S.J.; Kratochvil, B. J. Forensic. Sci. 1991, 36, 122-37. 14. Buscaglia, J. Anal. Chim. Acta. 1994, 288, 17-24. 15. Almirall J.R. Glass as Evidence of Association. In: Mute Witness. When Trace Evidence Makes the Case. Houck, M., Ed.; San Diego: Academic Press, 2001; 139-55. 16. Zurhaar, A.; Mullings, L. J. Anal. At. Spectrom. 1990, 5, 611-7. 17. Parouchais, T.; Warner, I. M.; Palmer, L.T.; Kobus, H. J. Forensic Sci. 1996, 41, 351-60. 18. Suzuki, Y.; Sugita, R.; Suzuki, S.; Kishi, T. Proceedings of the American Academy of Forensic Sciences Meeting, New York, NY. February 17-22. 1997, 3, 43. 19. Suzuki, Y.; Sugita, R.; Suzuki, S.; Marumo, Y. Anal. Sci. 2000, 16, 1195-8. 20. Duckworth, D.C.; Bayne, C.K.; Morton, S.J.; Almirall, J.R. J. Anal. Atom. Spectrom. 2000, 15, 821-8.

27

21. Pearce, N.J.G.; Perkins, W.T.; Westgate, J.A.; Gorton, M.P.; Jackson, S.E., Neal, C.R.; Chenery, S.P. Geostand. Newsl. 1997, 21, 115-146. 22. Duckworth, D.C.; Morton, S.J.; Bayne, C.K.; Koons, R.D.; Montero, S.; Almirall, J.R. J. Anal. Atom. Spectrom. 2002, 17, 662-8. 23. Curran, J. M.; Hicks, T.N.; Buckleton, J.S. Forensic Interpretation of Glass Evidence. CRC Press: Boca Raton, 2000. 24. NFI SOP document 222104. “Kwantitatieve bepaling van elementen in glasfragmenten met behulp van LA-ICP-MS”.(werkvoorschrift) Quantitative determination of elements in glass fragments using LA-ICPMS 25. NFI SOP document 222103. “Bediening van de ICP MS ELAN 6100 DRCPLUS. (handleiding)”. Operation of the ICPMS ELAN 6100 DRCPLUS. 26. Miller J.C. and Miller J.N. Statistics for Analytical Chemistry. 3rd ed.; West Sussex: Prentice Hall, 1993. pp 233. 27. Mendenhall, W. Introduction to Probability and Statistics. 3rd ed.; Duxbury: Belmont, California, 1971. pp 466 28. Anderson, T.W.; Sclove, S.L. Introductory Statistical Analysis. Houghton-Mifflin: Boston, 1974. pp 499. 29. Kleinbaum, D.G.; Kupper, L.L. Applied Regression Analysis and Other Multivariable Methods. Duxbury: Duxbury: Belmont, California, 1978, pp556. 30. American Society for Testing and Materials. Standard Guide for Conducting Ruggedness Tests. In: American Society for Testing and Materials. Annual Book of ASTM Standards. West Conshohocken: ASTM, 2001; 14.04. pp 416-21.. 31. Latkoczy, C. et al. “Development and evaluation of a standard method for the quantitative determination of elements in float glass samples by LA-ICP-MS.” NITECRIME internal report. 32. Montero, S. Trace elemental analysis of glass by inductively coupled plasma-mass spectrometry (ICP-MS) and laser ablation-inductively coupled plasma-mass spectrometry (LAICP-MS). FIU: Miami, 2002. 33. Sylvester, P.J.; Eggins, S.M.. Geostand. Newsl. 1997, 21, 215-229. 34. American Society for Testing and Materials. Standard Practice for Conducting an Interlaboratory Study to Determine the Precision of a Test Method. In: American Society for Testing and Materials. Annual Book of ASTM Standards. West Conshohocken: ASTM, 2001; 14.04. pp 200-21.

28

35. American Society for Testing and Materials. Standard Practice Terminology for relating to quality and statistics In: American Society for Testing and Materials. Annual Book of ASTM Standards. West Conshohocken: ASTM, 2001; 14.04. pp 132-9.

29

APPENDIX A “VERGELIJKINGSONDERZOEK VAN GLASMONSTERS OP ELEMENTEN MET BEHULP VAN LA-ICP-MS (werkvoorschrift)” A1.

DOEL EN TOEPASSINGSGEBIED

Dit werkvoorschrift heeft als doel de vergelijking van glasmonsters ten behoeve van forensisch onderzoek. Daartoe is het noodzakelijk om de concentraties te bepalen van de elementen kalium (K), titanium (Ti), mangaan (Mn), rubidium (Rb), strontium (Sr), zirconium (Zr), barium (Ba), lanthaan (La), cerium (Ce) and lood (Pb) in floatglass fragmenten met behulp van laser ablatie inductief gekoppeld plasma massa spectrometrie (LA-ICP-MS). De resultaten worden gebruikt voor het vergelijken van float-glass fragmenten, gebaseerd op de overlap van de bereiken die worden gedefinieerd door de concentraties van de elementen ± tweemaal de standaarddeviatie voor die metingen. A2.

PRINCIPE

De LA-ICP-MS opstelling bestaat uit een laser, ICP-bron en een massaspectrometer. Met een gefocusseerde, gepulste laser straal worden op discrete posities vaste materialen geableerd (door voornamelijk kinetische energie verkleind glas van enige micrometers groot) waarna dit geableerde materiaal via een helium-argon stroom getransporteerd wordt naar het plasma van de ICP-MS. Het plasma verdampt, atomiseert en ioniseert de aanwezige elementen. Een in hoog vacuüm gepositioneerd quadrupool massa filter, scheidt de elementen op basis van gewicht en lading. Tenslotte wordt de intensiteit van een massa/lading, uitgedrukt in “counts per second”, gemeten door de detector. Deze intensiteit is een maat voor de concentratie van het te meten element. Dit voorschrift betreft een "eigen" methode. A3.

CHEMICALIËN EN REAGENTIA

A3.1 A3.2 A3.3

Standard Reference Material 612 (trace elements in glass matrix, 50 mg/kg), NIST Certified Reference Material FGS02 Standard Reference Material 1830 (soda-lime float glass), NIST

A4.

APPARATUUR EN HULPMIDDELEN

A4.1 A4.2 A4.3 A4.4 A4.5

Perkin Elmer SCIEX ICP-MS ELAN DRCPLUS, de metingen zijn in “standard mode” NEW WAVE research Merchantek Products UP-213 Laser Ablation System, aperture imaged Nd:YAG laser met standard ablatie NWR cell, 30 cm3 Elan computer: Dell Optiplex GX270 Printer: HP deskjet 930C Laser computer: Dell Dimension 4100

A5.

WERKWIJZE VAN DE ANALYSE

Voorzorgsmaatregelen/veiligheid De algemene laboratoriumveiligheidsregels zijn van toepassing. De afzuiging van de ICP-MS en boven het plasma dient ten alle tijde aan te staan als het plasma aan is. Zorg bij het werken met glasfragmenten voor goede lichaamsbescherming, zoals bril en hand-schoenen.

A5.1

Het aansluiten en opstarten van de Laser-ICP-MS combinatie

De instructies hieronder gaan ervan uit dat de ICP-MS eerst is opgewarmd en geoptimaliseerd voor vloeistof analyse (dual calibratie, autolens en daily performance, zie hiervoor het bedieningsvoorschrift: BEDIENING VAN DE ICP MS ELAN 6100 DRCPLUS). Na deze optimalisatie wordt de ICP-MS uitgeschakeld en de gastoevoer naar de ELAN gesloten.

30

• • • • • • • • • • • • •

• • • • • • • • •

Start de laser energie aanvoer door de sleutel te draaien van de 0 positie naar de 1 positie Koppel de ‘Sync-out’-kabel, de ’S-Video’-kabel en de computerkabel (alvast) aan de laser Start de laser ablatie systeem (MEO laser) op de laser computer Open de ablatiekamer en breng NIST 612 [3.1] en FGS02 [3.2] standaarden in Klik op ‘purge’ Stel het licht in (ring en transmitted) en stel het oppervlak van NIST 612 [3.1] zo scherp mogelijk in de monitor (Z ≈ 13 mm) Klik op het rastericoon en creëer een ablatie-raster oppervlak van 7 x 105 µm2 Controleer de ‘properties’ van het raster (Passes: 1, Depth/Pass: 0 µm, rastering spacing 25 µm, spot size 60 µm, scan speed 25 µm/s, laser output 90%, repetition rate 10 Hz, pre-ablation pass uit) Sluit het properties venster. Wacht totdat de status ‘online’ is Koppel de slangetjes van de verstuiver en verstuifkamer, die in de slangen van de peristaltische pomp zitten, voorzichtig los Ontkoppel de connectie tussen het draaggas en de verstuiver (easy-fit connectie) Draai de koppeling tussen de verstuifkamer en de basis van de toorts mondstuk los en plaats deze in de doos die daarvoor bestemd is Er is een ‘3-slang-connectie’ na de laser output. Eén slang komt van de ablatiekamer output vandaan en de andere twee komen van de ICP-MS. Eén slang heeft een easy-fit connectie en moet aangesloten worden op het draaggas. De andere slang heeft een glazen fitting die aangesloten moet worden op de toorts mondstuk Open de workspace C:\ELANDATA\workspace\NFIoptimisation workspaces\ LA Daily Performance.wrk Ga op de Elan computer naar het sample venster Klik rechts op de het tabblad report. De ‘Send to File’ optie moet aangevinkt zijn, net als de opties ‘Include Title’, ‘Use Separator’ en ‘Overwrite’. De ‘Report Option Template’ moet daily3.rop zijn. Verander de rapportnaam naar de datum (jjmmdd), bijv. LA_041102 Sla de methode op (C:\ELANDATA\method\NFIoptimisation\LA Daily Performance.mth) Ga naar het optimization venster en open de default.dac file die voor vloeistofanalyses gebruikt wordt. Verander de parameters voor het ‘Nebulizer Gas Flow’ naar ongeveer 0,90 L/min en de ICP RF Power naar 1200 Watt. Sla de file op als default laser.dac De ‘Mass Flow Controller’ voor Helium moet in ‘flow’-positie en in ‘on’-positie staan en zover mogelijk zijn terugdraait (vergeet niet de helium-kraan open te draaien). De He-flow moet nu op minimum staan (≈ 0,017 L/min) Open het instrument venster, klik het Front Panel tabblad en ontsteek het plasma Wanneer het plasma aan is, draai langzaam de Helium flow op tot ≈ 0,75 L/min (wanneer dit te snel word gedaan kan het plasma uit gaan) Klik op ‘run scans’ in de laser computer. Een nieuw venster opent zich waarin moet staan: ‘Pattern to run selected patterns only, preablation bypass, ablation on line, wash out delay 0 seconds, laser enabled en warm up 0 seconds’. Wanneer de laser op online staat kan ‘run scan’ worden geklikt. De opwarmtijd van de laser is ongeveer 20 minuten.

31

Tabel 1A. Samenvatting van de instellingen Item

eenheid

waarde

Crater Diameter

µm

60

Repetition Rate

Hz

10

Laser output

%

90

Fluence Carrier Gas Nebulizer Gas Flow

-2

Jcm

12 (0.4 mJ)

Lmin-1 He

0.75 *)

-1

0.90 *)

-1

Lmin Ar

Auxiliary Gas Flow

Lmin Ar

0.90

Plasma Gas Flow

Lmin-1 Ar

15

W

1200 *)

RF Power Sweeps/Reading

1

Readings/Replicate

400

Replicates

1

Dwell time ms

10

Settling time

ns

Scan mode

200 Peak Hopping

Si

Massa (amu)

29

K

Massa (amu)

39

Ti

Massa (amu)

49

Mn

Massa (amu)

55

Rb

Massa (amu)

85

Sr

Massa (amu)

88

Zr

Massa (amu)

90

Sn

Massa (amu)

118

Ba

Massa (amu)

137

La

Massa (amu)

139

Ce

Massa (amu)

140

Pb

Massa (amu)

208

*) deze parameters kunnen, afhankelijk van de optimalisatie, iets varieren.

32

A5.2 • • • •

• • • • • •

Het uitvoeren van de Daily Performance. Open het sample venster in de ELAN computer en kies het tabblad manual. Vermeld de naam van het monster als glasnaam DP jjmmdd (bijv. 612 DP 041102). Sla het op als C:/ELANDATA/sample/sample04/la-daily.sam Open het dataset venster. Open de dataset C:/ELANDATA/dataset/daily performance 04/la_daily#.dat (# staat voor het toebehorende nummer) Ga terug naar het sample venster Controleer tijdens het opwarmen het signaal van NIST 612 [3.1] door op ‘analyze sample’ te klikken en bekijk in het ‘real time’ venster het signaal. Er moet aan de specificaties in Tabel 2 voldaan worden (gemeten met de detector in de pulsmode). Dit kan zo vaak herhaald worden als gewenst, aangezien de data elkaar overschrijven Verander de sample naam in FGS02 DP jjmmdd en sla de methode op Ga met de laser naar FGS02 [3.2] (Z ≈ 12 mm) en stel het oppervlak scherp op de monitor. Maak in plaats van een raster een lijn over het glas van ongeveer 3000 µm. Controleer of de ‘properties’ nog hetzelfde zijn Start de scan en klik na ongeveer 7 seconden op ‘analyze sample’ in de Elan computer. Controleer het signaal in het ‘real time’ venster Controleer de resultaten. Wanneer ze binnen de waarden van tabel 1 vallen kan verder worden gemeten aan de monsters. Als de waarden niet aan de eisen voldoen, moet er worden nagegaan wat daarvan de oorzaak is (mogelijke lekkage, vervuilde van cones, detector functioneert niet optimaal enz.). Corrigeer het probleem en meet de daily performance opnieuw Sla de workspace op.

Tabel 2A. Criteria voor Daily Performance. Achtergrond, 8 Criterium en 220 (cps)

29

Si (cps)

139

La (cps)

NIST 612 [3.1]

<3

> 2 000 000

> 20 000

FGS02 [3.2]

<3

> 1 000 000

> 10 000

A5.3 • • •

•

156

CeO/140Ce

<0.03

69

Ba++/138Ba

<0.03

Analyseren van de meetserie. Open de ablatiecel, haal NIST 612 [3.1] en FGS02 [3.2] eruit en vervang dit door het rooster met monsters. Sluit de ablatiecel en klik op ‘purge’ Stel het oppervlak van het rooster (Z ≈ 11 mm) scherp op de monitor De analyse moet beginnen en eindigen met twee metingen van NIST 612 [3.1] als calibrator. Verder moet de meetserie (als deze lang genoeg is) bestaan uit drie andere metingen van NIST 612 [3.1] en drie metingen van NIST 1830 [3.3] als controlemonsters, verspreid door de analyseserie (zorg ervoor dat de glassplinters groot genoeg zijn voor min. 7 (2+2+3 metingen voor NIST 612) respectievelijk 3 shots(3 metingen voor NIST 1830)) Eén meetserie mag uit maximaal 30 metingen bestaan. Als er meer metingen nodig zijn, moet er een nieuwe meetserie worden gemaakt.

33

Tabel 3A. Voorbeeld van een meetserie, waarin SVO A & B onbekende monsters zijn en SVO Z een bekend monster is. Spot Monster Bestandnaam 1 NIST 612 standaard 612#1 2 NIST 612 standaard 612#2 3 SVO B B#1 4 NIST 1830 1830#1 5 SVO A A#1 6 NIST 612 612#3 7 SVO Z Z#1 8 NIST 612 612#4 9 SVO Z Z#2 10 SVOA A#2 11 NIST 1830 1830#2 12 SVO B B#2 13 SVO A A#3 14 NIST 612 612#5 15 SVO Z Z#3 16 SVO B B#3 17 NIST 1830 1830#3 18 NIST 612 standaard 612#6 19 NIST 612 standaard 612#7 • • • • • • • • • •

•

Open de workspace glass_laser (C:/ELANDATA/workspace/NFI Quantitative workspaces/glass/glass_laser.wrk) Op het method venster, de methode moet la_glass_final zijn Klik rechts op het tabblad ‘report’, verander de naam van het rapport in de datum (bijv. 041102_glass.rep, data en rapport word opgeslagen in C:/ELANDATA/Report Output) Sla de methode op (C:/ELANDATA/methods/NFIquantitative/la/glass/ glass_laser_final.mth) Open het sample venster en klik op het ‘batch’ tabblad. Schrijf de monsterlijst en sla het op als datum (jjmmdd), techniek en materiaal (bijv. 041102-LA-glass) Schrijf in het vakje ‘description’ de zaaknummer en sla de monsterlijst op (C:/elandata/ samples/samples04/041102-la-glass.sam) De monsterlijst kan zo lang zijn als gewenst (ook al zijn er meer dan 30 analyses), aangezien de analyse één voor één wordt uitgevoerd Open het dataset venster en creëer een nieuwe dataset met dezelfde naam als die van de monsterlijst (C:/ELANDATA/data/04/0411/041102-la-glass.dat) Zoek de glasdeeltjes en geef elk deeltje een spot (of meerdere spots (bijv. NIST 612)) voor de ablatie in de volgorde van de monsterlijst Selecteer alle spots en ga naar de ‘properties’. Nu moet het vakje met ‘preablation’ worden aangevinkt en controleer de rest van de gegevens. Deze moeten zijn: o ablatie: 1 pass, depth/pass 0 µm, output 90%, repetition rate 10 Hz, spot size 60 µm, dwell time 25s; o preablatie: 1 pass, depth/pass 0 µm, output 90%, repetition rate 3 Hz, spot size 140 µm, dwell time 3s en intersite pause 0s. Vink de vakjes ‘default’ en ‘apply to all selected patterns’ aan zodat deze instellingen worden opgeslagen voor elke meting en klik OK Ga naar spot 1, stel het beeld scherp en klik ‘run scan’. Een nieuw venster opent zich waarin staat aangevinkt: ‘selected patterns only, preablation bypass, ablation on line, wash out delay 25 seconds, laser enabled en warm up 15 seconds’

34

• • • • • • • •

• • •

A5.4 • • • • • A5.5 • • •

• • •

Klik ‘run scan’ Stop de meting wanneer de preablatie is afgelopen, controleer of door de preablatie een vlak oppervlak is ontstaan. Controleer ook of het beeld nog scherp is (wanneer er opnieuw word scherp gesteld, vergeet dan niet de Z te veranderen bij ‘Edit Endpoint’) Herhaal de preablatie totdat er een vlak oppervlak is ontstaan Open de scanwaarden (properties) en schakel de preablatie uit en klik ‘apply settings to laser’ Selecteer in de ELAN computer de regel van de eerste meting en klik ‘analyze batch’. De meting wacht met starten totdat de laser met het opwarmen start Wanneer er in de laser computer de ‘online’ status word gegeven kan de meting beginnen door weer op ‘run scan’ te klikken (zoals reeds beschreven) Als het glasdeeltje breekt of de laser heeft de glassplinter volledig ‘doorboord’ en er is geen optie om de spot te verplaatsen en de meting opnieuw te doen, laat de meting dan gewoon doorgaan. Er bestaat een kans om 5 seconden te integreren tijdens de dataverwerking in plaats van 10 seconden Ga naar het ‘real time’ venster in de ELAN computer. De counts van Si, K, Sn en Ba moeten altijd gecontroleerd worden (er kunnen maximaal 6 elementen in de grafiek getoond worden). Ba zit in de lijm van het dubbelzijdig plakband en hierdoor zie je of je door het glasdeeltje ‘doorboord’ hebt, Sn wordt gemeten om zeker te weten dat niet de float kant van glas wordt gemeten Het achtergrondsignaal van Si en K is ongeveer 2000 respectievelijk 5000 counts Ga naar spot 2 en ga verder als genoemd hierboven bij “Ga naar spot 1…” Open de report file (.rep) wanneer de hele meetserie voltooid is met Notepad en zet elke regel tussen aanhalingstekens (“ ”) en verwijder alle dubbele metingen. Sla vervolgens het rapport op. Uitzetten van de LA-ICP-MS. Sla de workspace op wanneer alle metingen voltooid zijn Sluit de Laser Ablatie Systeem (MEO Laser) en draai de sleutel van de laser energie aanvoer van de 1 positie naar de 0 positie Open de ablatiekamer, haal het rooster weg en sluit de ablatiekamer Open op de Elan computer het instrument venster en klik op ‘stop’ waardoor het plasma uit zal gaan en wacht totdat de gastoevoer afgesloten is Ontkoppel de laser en sluit de verstuifkamer weer aan. Dataverwerking. Open het programma Glitter op de ELAN computer en klik op ‘element concentrations’, selecteer ELAN 6100, concentrations, IS in ppm en laad het rapport (.rep) Selecteer alle monsters en deselecteer ze weer, behalve 612#1, 612#2, 612#6 en 612#7. Dit zijn de standaarden (de interne standaard is 29Si en denk eraan dat het referentiemateriaal NIST 612 50 ppm glass Pierce 1997 is). Nu moeten alle monsters een Si concentratie hebben Open de ‘review signal selection window’ en integreer alle signalen voor de blanco (achtergrondsignaal) tussen de 20 en 60 (40 data punten), en integreer het meetsignaal tussen de 150 en de 270 (40 data punten*) en sla het op (.gli). Dit hoeft per monster slechts één keer gedaan te worden (dus niet per element). * Indien met het ableren het glas totaal wordt doorboord worden zoveel datapunten als mogelijk geïntegreerd (tot 40 datapunten) totdat het Ba signaal snel begint toe te nemen. Vermeld dit ook bij de rapportage van de meetresultaten. Sla het bestand ook op als ‘export’ door space/tab aan te vinken en het kruisje bij “chrondrite norm conc” weg te halen (.txt) Open de worksheet glass_analysis.xlt (G:\MI\MI_ALG\K2-R&D\LA-ICP-MS (2003-I-3)\Glas\ glass_analysis_files\glass_analysis.xlt) en open via Excel het geëxporteerde Glitter bestand (.txt) door als volgt eerst: gescheiden; volgende; komma; spatie en voltooien aan te klikken Kopieer het blad met meetgegevens naar de worksheet glass_analysis.xlt en verwijder het blad met bestaande (oude) meetgegevens

35

• • • • • • • • • A5.6

Sla de worksheet op in dezelfde folder als waar de Glitter bestanden in staan en met dezelfde naam als de Glitter bestanden (jjmmdd_glass.xls) Kopieer de concentraties van spoorelementen en plak (plakken speciaal) ze door de gegevens te transponeren in de sheet ‘All’ (verwijder overbodige rijen) en zet de gegevens op volgorde door in het eerste vakje onder ‘element’ te gaan staan en op de functie A-Z↓ te klikken Kopieer de corresponderende gegevens voor elk van de standaarden naar de daarvoor bestemde sheet ‘standards’ en plak (plakken speciaal) ze als waarden Doe hetzelfde met de meetgegevens in de sheet ‘samples’ Ga naar de sheet ‘plots_2s’ en vul de verkregen resultaten uit de ‘samples’ sheet in als waarden (plakken speciaal) Controleer rechts de grafieken en voeg/verwijder ontbrekende/overbodige gegevens in/uit de grafiek Sla de worksheet op in dezelfde folder als waar de Glitter bestanden in staan (glass_analysis.xls) Gebruik de verkregen resultaten uit de sheet ‘standards’ om de informatie voor de controle kaarten charts (overzicht van behaalde resultaten van de standaarden) in te vullen (contol_charts.xls) Sla het bestand op in de originele folder (G:\MI\MI_ALG\K2-R&D\ICPMS\Glas\ glass_analysis_files) Rapportage.

Glasmonsters worden bij dit onderzoek paarsgewijs onderscheiden indien de element-concentraties van de glasmonsters duidelijk verschillen (de concentratie bereiken overlappen niet voor tenminste twee van de tien elementen waarbij bedacht dient te worden dat Sn alleen kwalitatief wordt gebruikt (duidelijk aan- of afwezig)). De verschillen worden bevestigd met ANOVA en de t-test (p<0.05). Indien de element-concentraties niet zijn te onderscheiden (de concentratie bereiken overlappen voor tenminste negen van de tien elementen) wordt een verband gerapporteerd tussen de beide onderzochte glasmonsters. • • • •

Open het Word document report_template.doc (G:\MI\MI_ALG\K2-R&D\LA-ICP-MS (2003-I3)\Glas\ report_template.doc) Vul de vereiste data in en kopieer de tabellen en grafieken vanuit Excel Sla het rapport op in dezelfde folder als de andere bestanden Gebruik de volgende formulering zoals voorgesteld in het Validatierapport voor de moge-lijke hypotheses en de conclusies. Mogelijke hypotheses met betrekking tot de waargenomen resultaten zijn: 1. De monsters zijn afkomstig van dezelfde ruit (raam) 2. De monsters zijn afkomstig van verschillende ruiten maar van dezelfde producent en elke waargenomen overeenkomst in resultaten berust op louter toeval 3. De monsters zijn afkomstig van verschillende ruiten en van verschillende producenten maar betreffen hetzelfde type glas en elke waargenomen overeenkomst in resultaten berust op louter toeval 4. De monsters zijn afkomstig van verschillende ruiten en van verschillende producenten en betreffen verschillende types glas (vergelijking met de wereldpopulatie van willekeurig glas). Elke waargenomen overeenkomst in resultaten berust op louter toeval Indien de element-concentraties van de glasmonsters duidelijk verschillen (de concentratie bereiken overlappen niet voor tenminste twee van de tien elementen waarbij bedacht dient te worden dat Sn alleen kwalitatief wordt gebruikt (duidelijk aan- of afwezig)) wordt gerapporteerd dat “er geen bewijs is om hypothese een boven de andere te verkiezen ”. Indien de element-concentraties echter niet zijn te onderscheiden (de concentratie bereiken overlappen voor tenminste negen van de tien elementen) wordt gerapporteerd dat “het bewijs om hypothese een boven hypothese twee te verkiezen zwak is ”; “het bewijs om hypothese een boven hypothese drie te verkiezen sterk (zeer sterk indien alletien de elementen overlappen) is ” en “het bewijs om hypothese een boven hypothese vier te verkiezen zeer sterk (extreem sterk indien alletien de elementen overlappen) is ”. Tevens wordt in de rapportage vermeld dat “De hierbij gebruikte schaal van waarschijnlijkheidsconclusies luidt in opklimmende volgorde: zwak, sterk, zeer sterk en extreem sterk.”

36

Wanneer alle data gecontroleerd zijn en de resultaten naar behoren zijn, sla dan alle bestanden (.xl, .rep, .gli, .txt, .xls,.doc) op in één enkele folder en noem deze naar de datum en techniek (bijv. 041101_glass). Kopieer deze folder naar de folder van het betreffende zaakonderzoek op G:\MI\MI_ALG\K1clusters\zaken\20xx\yyyymmddnnn. A6. A6.1

CONTROLE Herhaalbaarheid van standaarden

De standaardedeviatie zoals berekend uit de de concentraties van de herhaalde metingen van de standaarden NIST 612 and NIST 1830 moet kleiner zijn dan 10 % for alle onderzochte elementen behalve Sn (dat alleen voor kwalitatieve doeleinden wordt gebruikt in het onderzoek). Indien dit niet wordt gehaald, ga dan na dat er geen uitbijters zijn met gebruikmaking van de testen van Dixon en Gubbs. Indien uitbijters worden aangetoond, rapporteer dan hierover op de controle-kaart en bereken met weglating van de uitbijters opnieuw de standaardaafwijking. Indien de gestelde precisie dan nog niet wordt gehaald, wordt hierover een opmerking gemaakt op de controle-kaart en wordt het experiment herhaald. A6.2

Controlekaart (controle van de gemiddelde waarden)

Vul de gegevens van het controlemonster per element in op de betreffende controle kaarten, zoals vermeld in 5.5. Onderneem actie (herhaal het experiment of achterhaal waardoor de afwijking wordt veroorzaakt) bij het controlemonster indien een meetwaarde de 3s-grens, tweemaal achtereenvolgens de 2s-grens overschrijdt of de meetwaarde elfmaal achtereenvolgens aan één zijde van het gemiddelde ligt. A7.

ONDERHOUD

Zie hoofdstuk 5 “Maintenance” van de hardware guide van Elan 6100 DRC, raadpleeg de Path-finder onder de helpknop of bekijk de videoclips onder de helpknop. Zie hoofdstuk 6 “Maintenance and Troubleshooting” pagina 88 en 89 van Nd:YAG Laser Ablation Systems Operator’s Manual. A8.

TROUBLESHOOTING

Zie hoofdstuk 6 “Troubleshooting” van de hardware guide van Elan 6100 DRC, raadpleeg de Path-finder onder de helpknop. Zie hoofdstuk 6 “Maintenance and Troubleshooting” pagina 90 en 91 van Nd:YAG Laser Ablation Systems Operator’s Manual.

37

A9. • • • •

LITERATUUR Handleiding “Elan 6100 DRC Optima Hardware guide” van Perkin Elmer Sciex Operater’s Manual “UP Series Nd:YAG Laser Ablation Systems” van New Wave Research Inc. User’s Manual “Glitter, On-line Interactive Data Reduction for the LA-ICP-MS Microprobe” van GEMOC en CSIRO Validation report “Forensic float-glass analysis using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS)”, 7 February 2005, Shirly Montero.

A10. • • •

WIJZIGINGEN TEN OPZICHTE VAN VORIGE VERSIE

De titel is gewijzigd in “Kwantitatieve bepaling van elementen in glasfragmenten met behulp van LAICPMS”. Het uitvoeren van het “vergelijkend deel” van het onderzoek (inclusief criteria) is omschreven. Een criterium voor de herhaalbaarheid van controlemonsters is toegevoegd (was eerder alleen vermeld in het validatierapport).

38

APPENDIX B. LINEARITY

Table 1B. Linearity of the method for the measurement of the concentration of Potassium reported concentration [K] uncertainty, Glass, K/Si standard K/Si K/Si K in µgg-1, [K] in µgg-1 NIST deviation uncertaintya 0 0 614 0.008 0.0004 0.0011 30b 1b c 612 0.016 0.0002 0.0006 66.26 3.3d 610 0.125 0.0024 0.0059 486c 57d a ½ calculated as the 95% confidence value ts/n or in this case 2.5 s b certified by NIST c reported in Ref 21. d calculated as the concentration limit at 95% level

600 y = 3884x + 1 2 R = 0.9998

Concentration, µgg-1

500 400 300 200 100 0 0.00

0.02

0.04

0.06 39

0.08

0.10

0.12

0.14

29

K signal ratio to Si

Figure 1B. Linearity for the determination of the concentration of K using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

39

Table 2B. Linearity of the method for the measurement of the concentration of Titanium Glass, NIST 614 612 610

Ti/Si 0 0.00005 0.00055 0.00643

Ti/Si standard deviation

Ti/Si uncertaintya

reported concentration Ti in µgg-1, [Ti]

[Ti] uncertainty, in µgg-1

0.00001 0.00004 0.00015

0.0000 0.0001 0.0004

0 3.1b 48.11c 434c

0.3b 4.8d 11d

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s reported by NIST c reported in Ref 21. d calculated as the concentration limit at 95% level b

Concentration, µgg

-1

500 y = 66994x + 4 2 R = 0.999

400 300 200 100 0 0

0.001

0.002

0.003 49

0.004

0.005

0.006

0.007

0.008

29

Ti signal ratio to Si

Figure 2B. Linearity for the determination of the concentration of Ti using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

40

Table 3B. Linearity of the method for the measurement of the concentration of Manganese Glass, NIST 614 612 610

Mn/Si 0 0.0004 0.0120 0.1428

Mn/Si standard deviation

Mn/Si uncertaintya

0.00008 0.00004 0.00096

0.0002 0.0001 0.0024

reported concentration Mn in µgg-1, [Mn] 0 4.41b 38.43c 485d

[Mn] uncertainty, in µgg-1 2.5b 0.8e 10e

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b reported in Ref 33. c reported in Ref 21. d certified by NIST e calculated as the concentration limit at 95% level 600 y = 3393x + 0 2 R = 0.9999

Concentration, µgg-1

500 400 300 200 100 0 0.00

0.02

0.04

0.06 55

0.08

0.10

0.12

0.14

0.16

29

Mn signal ratio to Si

Figure 3B. Linearity for the determination of the concentration of Mn using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

41

Table 4B. Linearity of the method for the measurement of the concentration of Rubidium Glass, NIST 614 612 610

Rb/Si 0 0.0004 0.0142 0.1934

Rb/Si standard deviation

Rb/Si uncertaintya

0.00002 0.00028 0.00083

0.0001 0.0007 0.0021

reported concentration [Rb] uncertainty,b Rb in µgg-1, [Rb]b in µgg-1 0 0.855 0.005 31.63 0.4 431.1 0.8

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b certified by NIST 500 y = 22291x + 0 2 R =1

Concentration, µgg-1

400

300

200

100

0 0.00

0.05

0.10 85

0.15

0.20

0.25

29

Rb signal ratio to Si

Figure 4B. Linearity for the determination of the concentration of Rb using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

42

Table 5B. Linearity of the method for the measurement of the concentration of Zirconium Glass, NIST 614 612 610

Zr/Si 0 0.0001 0.0081 0.0974

Zr/Si standard deviation

Zr/Si uncertaintya

0.00001 0.00024 0.00072

0.00004 0.00060 0.00180

reported concentration Zr in µgg-1, [Zr] 0 0.86b 35.99c 439.9c

[Zr] uncertainty, in µgg-1 0.006d 1.6d 7.2d

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s reported in Ref 34. c reported in Ref 21. d calculated as the concentration limit at 95% level b

500 y = 4518x - 0 2 R =1

Concentration, µgg-1

400 300 200 100 0 0.00

0.02

0.04

0.06

0.08

0.10

0.12

-100 90

29

Zr signal ratio to Si

Figure 5B. Linearity for the determination of the concentration of Zr using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

43

Table 6B. Linearity of the method for the measurement of the concentration of Barium Glass, NIST 614 612 610

Ba/Si 0 0.0002 0.0030 0.0364

Ba/Si standard deviation

Ba/Si uncertaintya

0.00001 0.00006 0.00022

0.0000 0.0002 0.0005

reported concentration Ba in µgg-1, [Ba] 0 3.72b 37.74c 424.1c

[Ba] uncertainty, in µgg-1 0.9d 1.2d 27d

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b reported in Ref 33. c reported in Ref 21. d calculated as the concentration limit at 95% level 500 y = 11627x + 1 2 R =1

Concentration, µgg-1

400 300 200 100 0 0.000

0.005

0.010

0.015 137

0.020

0.025

0.030

0.035

0.040

29

Ba signal ratio to Si

Figure 6B. Linearity for the determination of the concentration of Ba using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

44

Table 7B. Linearity of the method for the measurement of the concentration of Lanthanum Glass, NIST 614 612 610

La/Si 0 0.0004 0.0198 0.2540

La/Si standard deviation

La/Si uncertaintya

0.00002 0.00008 0.00157

0.00004 0.00020 0.00393

reported concentration La in µgg-1, [La] 0 0.83b 35.77c 457.4c

[La] uncertainty, in µgg-1 0.02b 1.30d 51.7d

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b reported by NIST c reported in Ref 21. d calculated as the concentration limit at 95% level 600 500 Concentration, µgg

-1

y = 1801x + 0 2 R =1

400 300 200 100 0 0.00

0.05

0.10

0.15 139

0.20

0.25

0.30

29

La signal ratio to Si

Figure 7B. Linearity for the determination of the concentration of La using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

45

Table 8B. Linearity of the method for the measurement of the concentration of Cerium Glass, NIST 614 612 610

Ce/Si 0 0.0004 0.0211 0.2620

Ce/Si standard deviation

Ce/Si uncertaintya

0.00001 0.00043 0.00062

0.00004 0.00106 0.00156

reported concentration Ce in µgg-1, [Ce] 0 0.75b 38.35c 447.8c

[Ce] uncertainty, in µgg-1 0.06d 1.04d 12.9d

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b reported in Ref 33. c reported in Ref 21. d calculated as the concentration limit at 95% level 500 y = 1707x + 1 2 R = 0.99998

Concentration, µgg-1

400 300 200 100 0 0.00

0.05

0.10 140

0.15

0.20

0.25

0.30

29

Ce signal ratio to Si

Figure 8B. Linearity for the determination of the concentration of Ce using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

46

Table 9B. Linearity of the method for the measurement of the concentration of Cerium Glass, NIST 614 612 610

Pb/Si 0 0.002 0.025 0.303

Pb/Si standard deviation

Pb/Si uncertaintya

0.00004 0.00044 0.00130

0.0001 0.0011 0.0033

reported concentration Pb in µgg-1, [Pb]b 0 2.32 38.96 426

[Pb] uncertainty,b in µgg-1 0.04 0.2 1

a

calculated as the 95% confidence value ts/n½ or in this case 2.5 s b certified by NIST

Concentration, µgg-1

500 y = 1401x + 1 2 R = 0.9999

400 300 200 100 0 0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

29

Pb (206+207+208) signal ratio to Si

Figure 9B. Linearity for the determination of the concentration of sumPb using 29Si as internal standard in SRM NIST 614, NIST 612 and NIST 610.

47