1.1 Two analytical X-ray techniques were investigated and compared against an atomic absorption spectrometry (AAS) procedure for analyzing V.

1.1.1 X-ray diffraction (XRD) was performed using custom OSHA Laboratory software and two Model 3500 Automated Powder Diffractometers (APDs) (Phillips Electronics Instruments Co., Mahwah, NJ) sharing the same generator. Results for these APDs are labeled APD-A or APD-B in this report.

1.1.2 X-ray fluorescence (XRF) was performed using a Model 77-800 (upgraded to Model 77-900A) Energy Dispersive X-ray Fluorescence (EDXRF) Spectrometer (Finnigan Corporation, Sunnyvale, CA) consisting of: a an X-ray generator with a direct-beam Rh end-window X-ray tube and an X-ray spectrometer console using a Computer Automation Alpha 16 computer.

More recently, field samples submitted to the Laboratory for V₂O₅ analyses have been analyzed using a Kevex 770/8000 Delta EDXRF system (Kevex Instruments Inc., San Carlos, CA) consisting of: Kevex 770 X-ray generator, its associated satellite box, vacuum system, helium flush system, firmware-based 8000 keyboard console, computer monitor, Digital Equipment Corporation (DEC) 11/73 computer, graphics memory, Kevex spectrum analyzer, and Toolbox II software. This latter system uses an Fe secondary target for this analysis and offers improved sensitivity, lower background and greater resolution of interferences (9.1). The system parameters for both XRF systems are given in the experimental section of the method (9.1).

1.1.3 For the purpose of comparison, the V₂O₅ samples on Ag membranes were re-analyzed for V by AAS procedure (9.3) using a Model 603 Atomic Absorption Spectrometer (Perkin-Elmer Corp., Norwald, CT).

1.2 Three particle-size distributions were investigated:

1. "M" samples simulating respirable particle size range (0 to 10 µm with median of approximately 3 µm). The "M" samples are referred to by the label "Respirable".
2. "W" samples simulating fine-respirable particle size range (0 to 3 µm). The "W" samples are referred to by the label "Fine respirable".
3. "Fine W" samples simulating fume-like particles.

1.2.1 Quantitation of V₂O₅ dust (approximating respirable dust) was assessed using spikes at nominal levels of 233, 467 and 700 µg V₂O₅ ("M" samples).

1.2.2 Quantitation of V₂O₅ dust (approximating fine-respirable particle size) was assessed using spikes at levels of 237, 474, and 710 µg V₂O₅ ("W" samples). The spiked levels indicated in the NIOSH study were duplicated and are not the usual validation levels of 0.5, 1, and 2 times the OSHA PEL. The respirable characterization and spiking levels are further described in Reference 9.4.

1.2.3 To assess analyte sensitivities to very small fume-like particles, aliquots of an unstirred acetonitrile suspension of the finest particles in the "W" material were analyzed (See Section 2.4.2).

1.2.4 All samples were analyzed by XRD, XRF, and AAS techniques.

2.1 Filter Membranes

2.1.1 FWS-D (0.5 µm pore size) membranes were spiked with sonicated acetonitrile suspensions of V₂O₅. Small pore size membranes were used during the evaluation to prevent the loss of small particles during the spiking with liquid suspensions. During sample collection of particles suspended in air, the 5.0-µm pore size PVC membrane should sufficiently retain the smaller particles due to static charge and collection characteristics. Collection efficiencies exceeding 99% have been reported of 0.3-µm Dioctyl phthalate aerosol collection on PVC filters (9.5).

2.1.2 Silver metal membranes (25-mm diameter, 0.45-µm pore size) were used to support the prepared thin films for presentation in X-ray analyses.

2.2 Preparation of Standard Materials

Procedure:

2.2.1 Reagent grade V₂O₅ (99.8%, J.T. Baker, Phillipsburg, NJ) was used as the starting material for the X-ray methods. For the AAS comparison method, a 1,000 µg/mL V standard in a dilute HCl matrix (Lot #J141, RICCA Chemical Company, Arlington, TX) was used for preparing AAS standards.

2.2.2 Respirable stock material ("M") was prepared by hand grinding reagent grade V₂O₅ in an alumina mortar and pestle at room temperature. The ground material was added to 50 to 75 mL of tetrahydrofuran (THF) in a glass beaker. The sonicated suspension was west-sieved through 10-µm nylon mesh using a sieving bottle as shown and described in Section 6.4. of the method (9.1). The dust was isolated from the suspension by filtering it onto a Ag membrane. This material was used in preparing only spikes.

2.2.3 For a finer particle-size distribution, additional stock material ("W") was prepared by a different grinding technique. Reagent grade V₂O₅ was ground in a freezer mill operated with liquid nitrogen for 10 minutes. This stock material was also wet-sieved through 10-µm nylon mesh as described in Section 2.2.2. In addition to spiked sample preparation, this material was used to prepare calibration standards because the freezer mill produces more reproducible particle-size distributions than those produced from hand grinding with a mortar and pestle.

2.2.4 Results: The reagent grade material, "W", and "M" materials were analyzed by AAS and gave an assay of 101.3% V₂O₅ as shown below:

Reagent		"W" Material		"M" Material
mg Taken	Assay by AAS	mg Taken	Assay by AAS	mg Taken	Assay by AAS
1.307	104.59%	2.263	101.59%	2.069	102.66%
1.633	101.84	1.581	101.39	1.535	103.45
2.348	99.96	2.295	99.48	5.663	97.30
2.457	98.82	1.297	100.93	2.138	100.09
Mean	101.3%		100.8%		100.9%

The manufacturer's assay of 99.8% was used in calculations for all materials derived form the reagent grade V₂O₅.

2.3 Preparation and XRD Analysis of Calibration Standards

Procedure:

2.3.1 X-ray calibration standards preparation

Twenty-four calibration standards were prepared from THF suspensions of fine-respirable stock material ("W"). To avoid reduction of V₂O₅ by warm organic agents, fixatives were not used to secure the dust on the membranes. This also facilitated sample and standard re-analysis by AAS. Three standards were prepared on Ag membranes at each of the following levels:
 

"W" Material - Standards

Standard Delivered µg V2O5		Volume (mL)		Reagent Concentration µg/mL
50 100 200 250 499 998 1996 2495		5 10 2 25 5 10 20 25		10.040 10.040 100.02 10.040 100.02 100.02 100.02 100.02

2.3.2. Calibration and analysis (XRD)

These calibration standards were analyzed on two different APDs. The sample analysis order was scrambled to reduce instrumental drift effects. Two-theta calibrations were performed using the primary 38.15° two-theta (2q) Ag calibration line to avoid potential interferences from other V₂O₅ lines. (In practice, the strong secondary 44.33° 2q Ag line is within 0.03° 2q of a low intensity V₂O₅ peak; therefore, no significant error is introduced when using the secondary Ag line in 2q calibrations.) The X-ray generator settings were 40 kV and 40 mA. Integration times of 1 s and 0.02° 2q steps were used throughout the study. The detection limit was determined at both 1- and 10-s integration periods. A custom OSHA computer program (9.6) was used to establish a two-piece calibration curve and calibration coefficients. Standard data is shown in Table 1. A two piece curve fit or a polynomial curve fit is generally performed to optimize recovery over the entire analytical range. A second-order polynomial fit for the low end allows for the correction of losses due to penetration of dust into the Ag membrane and to correct for the conservative heuristic used when establishing integration limits. The second-order fit of the upper end is smoothly spliced onto the upper end of the low range fit. The second-order fit of the upper end allows for partial correction for sample self-absorption effects. This is important because a Leroux correction (9.7) is not performed. In the past, Leroux corrections have been found to over-correct (9.8).

2.3.2 Calibration and analysis (XRD)

These calibration standards were analyzed on two different APDs. The sample analysis order was scrambled to reduce instrumental drift effects. Two-theta calibrations were performed using the primary 38.15° two-theta (2q) Ag calibration line to avoid potential interferences from other V₂O₅ lines. (In practice, the strong secondary 44.33° 2q Ag line is within 0.03° 2q of a low intensity V₂O₅ peak; therefore, no significant error is introduced when using the secondary Ag line in 2q calibrations.) The X-ray generator settings were 40 kV and 40 mA. Integration times of 1 s and 0.02° 2q steps were used throughout the study. The detection limit was determined at both 1- and 10-s integration periods. A custom OSHA computer program (9.6) was used to establish a two-piece calibration curve and calibration coefficients. Standard data is shown in Table 1. A two piece curve fit or a polynomial curve fit is generally performed to optimize recovery over the entire analytical range. A second-order polynomial fit for the low end allows for the correction of losses due to penetration of dust into the Ag membrane and to correct for the conservative heuristic used when establishing integration limits. The second-order fit of the upper end is smoothly spliced onto the upper end of the low range fit. The second-order fit of the upper end allows for partial correction for sample self-absorption effects. This is important because a Leroux correction (9.7) is not performed. In the past, Leroux corrections have been found to over-correct (9.8).

2.4 Preparation and X-ray Analyses of Spiked Samples

Procedure:

2.4.1 Sample preparation

Analyses were performed on a total of 36 samples (six samples at each of three test levels for the two different materials, "W" and "M"). Acetonitrile was used as the vehicle for spiking FWS-D membrane filters. Neither V₂O₅ nor the PVC filter medium dissolves appreciably in acetonitrile. The "W" spiked samples were prepared by filtration of acetonitrile suspensions of the freezer mill material upon FWS-D (0.5-µm pore size) filters supported on a fritted-glass filtering support. The "M" spiked samples were prepared by similar filtration of acetonitrile suspensions of the mortar-and-pestle ground material. Upon drying, the filter membranes were transferred to centrifuge tubes and the 10 mL of THF was added to each to dissolve the membranes. The tubes were placed in an ultrasonic bath and the tube contents were sonicated for approximately 10 minutes. The sonicated suspension was filtered onto 25-mm Ag membranes (0.45-µm pore size) for subsequent analysis. As in the preparation of the calibration standards (See Section 2.3.1), fixatives were not used. The three test levels were produced as follows:
 

	Sample Delivered µg V2O5	Aliquot (mL)	Reagent Concentration µg/mL
"W"	237 474 710	2 4 6	118.62 118.62 118.62
"W"	233 467 700	2 4 6	116.92 116.92 116.92

2.4.2 Fume-like sample preparation:

In order to assess the effect of very fine particles on recovery, three samples were prepared (10-mL aliquots) from the center of the same 118.62 µg V₂O₅/mL acetonitrile suspension of fine-respirable stock material "W" after allowing the coarser material to settle out of the unstirred suspension. After 2.5 h, a significant fraction of the larger particles had settled out leaving a hazy suspension. Aliquots of the supernatant suspension were spiked directly onto 0.45-µm pore size Ag membranes. These samples were referred to as "Fine-W" samples.

2.4.3 The "W", "M", and "Fine-W" Samples and blanks were analyzed by XRD, XRF, and AAS using the procedures which follow:

XRD analytical procedure

The Ag membrane samples and blanks were analyzed in the same manner as the calibration standards described in Section 2.3.2.

2.4.4 XRF analytical procedure

The "W" calibration standards were analyzed by XRF using the program described in Appendix 1. Count data were collected on all channels in air at 20 kV, 0.5 mA, for 100 s with a narrow collimator and without an X-ray filter. After analyzing the standards, a concentration-response curve was prepared to calibrate on the integrated counts in the 17-channel (~0.64 kV) region spanning the V K_a peak at 4.949 kV. Background counts were estimated using a linear background model between 3.4 to 5.9 kV. The equations obtained (Table 1) were used to calculate the amount of V₂O₅ present in spiked ("M" and "W") samples.

For the "Fine-W" spiked samples, a separate regression was performed using single representative "W" standards at each of the loadings.

2.4.2 Fume-like sample preparation:

In order to assess the effect of very fine particles on recovery, three samples were prepared (10-mL aliquots) from the center of the same 118.62 µg V₂O₅/mL acetonitrile suspension of fine-respirable stock material "W" after allowing the coarser material to settle out of the unstirred suspension. After 2.5 h, a significant fraction of the larger particles had settled out leaving a hazy suspension. Aliquots of the supernatant suspension were spiked directly onto 0.45-µm pore size Ag membranes. These samples were referred to as "Fine-W" samples.

2.4.3 The "W", "M", and "Fine-W" Samples and blanks were analyzed by XRD, XRF, and AAS using the procedures which follow:

XRD analytical procedure

The Ag membrane samples and blanks were analyzed in the same manner as the calibration standards described in Section 2.3.2.

2.4.4 XRF analytical procedure

The "W" calibration standards were analyzed by XRF using the program described in Appendix 1. Count data were collected on all channels in air at 20 kV, 0.5 mA, for 100 s with a narrow collimator and without an X-ray filter. After analyzing the standards, a concentration-response curve was prepared to calibrate on the integrated counts in the 17-channel (~0.64 kV) region spanning the V K_a peak at 4.949 kV. Background counts were estimated using a linear background model between 3.4 to 5.9 kV. The equations obtained (Table 1) were used to calculate the amount of V₂O₅ present in spiked ("M" and "W") samples.

For the "Fine-W" spiked samples, a separate regression was performed using single representative "W" standards at each of the loadings.

2.5 Calibration and Analysis (AAS) - Stock Materials and Spiked Samples

Procedure:

2.5.1 Preparation of atomic absorption standards:

Eight standards were prepared in a 4% HCl matrix by serial dilutions of the 1,000 µg/mL V standard. Concentration of the final standards ranged from 100 to 0.1 µg/mL.

2.5.2 Analytical procedure (AAS) - stock materials:

To check the purity of the stock materials used in the evaluation of the X-ray methods, four samples of each of the reagent, "W", and "M" materials were weighed out and transferred to 100 mL volumetric flasks using a total of 4-mL HCl and approximately 5 mL deionized water (DI H₂O). These were brought just to a boil on a hot plate. After cooling, the samples were diluted to volume with DI H₂O to give a 4% HCl matrix. Analysis was performed according to reference 9.3 The results are shown in Section 2.2.4.

2.5.3 Analytical procedure (AAS) - silver membrane samples:

The blanks and the spiked "W", "M", and "Fine-W" samples were re-analyzed by AAS after the XRD and XRF analyses. The edges of the membranes were bent before acid extraction to encourage the free flow of acid above and below the membrane. The membranes were agitated and sonicated for 10 to 15 s in 250-mL Phillips beakers containing 10-mL DI H₂O and 1 mL concentrated HCl. After the dust was visually released from the membranes, the beakers were placed on a hot plate and brought just to a boil. They were then removed to cool to ambient temperature (20 to 25 °C) while being agitated. The solutions were then quantitatively transferred to 25-mL volumetric flasks using 4 to 5 small rinses of DI H₂O. The volumetric flasks were then diluted to volume with DI H₂O to give a 4% HCl matrix. Analysis was performed according to Reference 9.3.

2.6 Results: The results for the three different analytical techniques and sample materials are presented as follows:
 

Table	Results
2 3 4 5 6 7 8 9	XRD results for "W" material XRD results for "M" material EDXRF results for "W" material EDXRF results for "M" material AAS results for "W" material AAS results for "M" material Summary - analyses of "Fine-W" material Summary - analyses of "W" and "M" material

Tables 8 and 9 contain summary results of the three analytical techniques performed on each sample.

3.1 Outlier and Bartlett's Tests (XRD and EDXRF)

The calibration data (Table 1) and all of the "W" and "M" spiked-sample data passed the ASTM test for outliers at the 99% confidence level (9.9). All the spiked-sample data passed Bartlett's test (9.10), so the results were pooled as appropriate. Statistical test results are shown below:

 

Bartlett's Test Results "W" Reference Material (also used for calibration standards)

Bartlett's variance homogeneity tests:
Critical Chi-squared value  =  9.21 (99% Confidence Level and N  =  3)
XRD (APD-A) XRD (APD-B) EDXRF	Chi-squared  =  1.11 Chi-squared  =  2.11 Chi-squared  =  9.21	N  =  3 N  =  3 N  =  3

 

"M" Reference material (coarser than calibration standards)

 

Bartlett's variance homogeneity tests:
Critical Chi-squared value  =  9.21 (99% Confidence Level and N  =  3)
XRD (APD-A) XRD (APD-B) EDXRF	Chi-squared  =  0.26 Chi-squared  =  0.49 Chi-squared  =  2.04	N  =  3 N  =  3 N  =  3

3.2 The precision and accuracy (9.11.) for the XRD method:

Recoveries, precision, and overall errors are shown below. X-ray diffraction results for "W" stock material (AAS results are shown in parentheses) are:
 

Recovery:	APD-A Ave. Recovery APD-B Ave. Recovery Combined XRD Recovery	=  0.894  =  0.880  =  0.887 (0.900)
Precision:	APD-A CV1(Pooled) APD-B CV1(Pooled) Combined XRD CV1(Pooled)	=  0.117    =  0.125    =  0.121 (0.031)
Overall Error:		=  ± 36%

X-ray diffraction results for "M" stock material
(AAS results are shown in parentheses) are:
 

Recovery:	APD-A Ave. Recovery APD-B Ave. Recovery Combined XRD Recovery	=  1.680  =  1.867  =  1.774 (0.933)
Precision:	APD-A CV1(Pooled) APD-B CV1(Pooled) Combined XRD CV1(Pooled)	=  0.062    =  0.073    =  0.068 (0.015)
Overall Error:		=  ± 91%

3.3 Precision and Accuracy - XRF method

Recoveries, precision, and overall error are shown below. 
X-ray fluorescence results for "W" material
(AAS results shown in parentheses) are:
 

Ave. Recovery: Precision: CV1(Pooled) Overall Error:

=  0.871 (0.900)  =  0.097 (0.031)    =  ± 32%

X-ray fluorescence results for "M" material
(AAS results shown in parentheses) are:
 

Ave. Recovery: Precision: CV1(Pooled) Overall Error:

=  0.965 (0.933)  =  0.064 (0.015)    =  ± 16%

4.1 Procedure: Blanks were analyzed in order to estimate the microgram detection limits. Blanks were analyzed by XRF using the total analytical times indicated in Appendix 1. The blanks were also analyzed by XRD as described in Section 2.3.2 using total analytical times of 65 and 650 s (corresponding to integration times of 1 and 10 s respectively). The X-ray detection limit estimates were based on the International Union of Pure and Applied Chemistry (IUPAC) definition as three times the standard deviation of the measurements performed on blanks divided by the slope (9.12, 9.13). The AAS detection limit was estimated using three times the minimum AAS reading.

4.2 Results: Detection limit results are summarized below and shown in Table 10. Detection limits determined for the analytical methods used (µg V₂O₅):

X-ray Diffraction			X-ray Fluorescence			Atomic Absorption
DL	Total time		DL	Total time		DL	Total time
25 µg 25 µg	65 s 650 s		14 µg   2 µg	100 s 1,000 s		9 µg -	4 s -

Some XRD blank results were abnormally high. This gave a large detection limit for the XRD method. No V was identified using XRF or AAS analyses on the same blanks; therefore, V₂O₅ contamination was ruled out as a possible cause. A sample of ten Ag membranes from the same lot also did not have the XRD interference. This indicates that the PVC membranes and/or the THF solvent may be responsible. Salt (NaCl) has its primary peak near the V₂O₅ analytical peak. The non-stoichiometric compound, K_0.2Na_0.8Cl, has its primary diffraction line at the V₂O₅ analytical peak. Finger prints which potentially contain salt did not produce significant peaks in the range scanned.

Comparisons were performed on the results for the three different particle-size distributions. Due to the sample preparation method used for the fume-like samples, the amount of V₂O₅ taken was not known beforehand. Using result ratios (Mean Relative Recoveries) allows making a comparison.

5.1 The data used for this comparison study were taken from Tables 8 and 9.

5.2 The mean relative recoveries for the materials studied:

	Mean Relative Recoveries
	XRD/AAS	XRF/AAS
Fume-like particles "Fine-W" Fine-respirable particles "W" Respirable dust particles "M"	0.651 0.986 1.901	0.998 0.968 1.034

Related to the evaluation of accuracy and precision is method (system) comparison which normally employs duplicate sampling (or spiking) to holistically compare the quality of a known analytical system with one or more untested analytical systems. Duplicate spiking (a separate set for each method) was not performed in these comparisons because the non-destructive nature of the XRD and XRF analyses made that unnecessary and counter-productive. Westgard and associates (9.14) have proposed a detailed scheme for method comparison. This evaluation scheme calls for the application of a least-squares linear regression of the results from the candidate method and comparative analytical method (assumed to be dependent and independent variables respectively). The regression is then analyzed by statistical techniques such as the F-test, t-test, least-squares analysis and correlation coefficients. This scheme is based on the assumption that the comparative method gives the true value (9.15). The approach is possibly biased against discovering a better analytical system. In these analyses, however, the AAS technique should give the most accurate value for the amount of V₂O₅ captured on and in the Ag membrane and is considered the reference method. The statistical evaluation is meaningful in that limited context. Comparisons of the XRD and XRF candidate methods with the AAS comparison method are presented below.

6.1 A summary of the AAS versus X-ray comparison data from the computer calculations follows:

	a	Slope	Sslope	Bias	r	r2
APD-A "W" dust APD-A "M" dust	2.11 -28.02	0.9848 1.8826	0.057 0.074	-4.33 340.92	0.9739 0.9886	0.9484 0.9773
APD-B "W" dust APD-B "M" dust	-23.69 -35.38	1.0440 2.0999	0.063 0.088	-4.99 424.44	0.9717 0.9870	0.9441 0.9743
EDXRF "W" dust EDXRF "M" dust	-3.69 -23.34	0.9779 1.0988	0.046 0.041	-13.06  17.95	0.9838 0.9900	0.9679 0.9800

Where:
a (in µg)	=	intercept of regression line
Slope	=	slope of regression line
Sslope	=	standard deviation for the slope
Bias (in µg)	=	mean µg V2O5 found by candidate method less mean µg V2O5 found by reference method
r	=	correlation coefficient
r2	=	Coefficient of determination (fraction of variation in candidate measurements due to variation in reference measurements)

The coefficients of determination are between 0.94 and 1.00. This indicates that, at most, only 6% of the variance in the candidate measurement is not accounted for by variance in the "independent" reference measurement.

The slopes approaching a value of 1 indicate small relative bias between the candidate and comparison methods. This is the case in all but the XRD results for the "M" dust. The slope of approximately 2 indicates an unacceptable relative bias.

6.2 Results (t-tests and F-tests)

	t	t-crit	F	F-crit	df
APD-A "W" dust APD-A "M" dust	-.4557 9.0564	2.110 2.120	1.02257 3.62634	2.30 2.33	17 16
APD-B "W" dust APD-B "M" dust	-.4683 9.0849	2.110 2.120	1.15448 4.52635	2.30 2.33	17 16
EDXRF "W" dust EDXRF "M" dust	-1.659 2.3854	2.120 2.120	1.01211 1.23193	2.33 2.33	16 16

Where:
t	=	calculated Student t-test value (bias indicated by sign)
t-crit	=	two-sided critical t value for 0.05 probability (from Reference 9.16)
F	=	calculated F-test value
F-crit	=	critical F value for 0.05 probability (from Reference 9.16)
df	=	degrees of freedom (no. paired observations - 1)

For the fine respirable dust ("W") samples, no significant difference was detected between the performance of the test and comparative methods.

The t-test data above indicate that for the respirable dust ("M") samples there is a significant difference in means between the X-ray test methods and the AAS comparison method.

In the case of the "M" dust samples, the F-test data indicate that there was a significant difference in precision between the XRD and AAS methods, but there was not a significant difference in precision between the XRF and AAS methods.

In order to get the best estimate of the Overall Error, the recoveries and CV₁ Pooled results for the two different APDs used in the validation were averaged. Averaging was not necessary for the XRF results, since only one instrument was used. The ranges shown are for all experiments performed and are therefore somewhat larger than if only a single APD instrument were used.

The results for the accuracy and precision calculations in Section 3 were based on the assumption that the theoretical amount of V₂O₅ delivered to each PVC membrane represented the true amount. The X-ray stock material was verified against the AAS standard giving approximately 100% V₂O₅. The XRD, XRF, and AAS analyses of the "W" dust samples and the XRF and AAS analyses of the "M" dust samples agreed well. A 10% negative bias was noted in recoveries of the "W" dust samples. It was concluded that the negative bias in these cases was probably due to losses incurred in spiking the PVC membrane with a V₂O₅ suspension in acetonitrile. Because the same samples were analyzed by XRD, XRF and AAS, any physical losses incurred in spiking and transfer to the Ag membrane were the same for each sample regardless of the analytical technique. The recoveries for "M" dust samples by XRD disagreed considerably with the other methods investigated and followed the trend expected for larger particles. The process of spiking by means of a suspension does not duplicate aerosol sampling and may not be ideally representative of samples taken with a cyclone.

The ability of each analytical technique to accurately determine each V₂O₅ material was assessed using overall error. The overall error should be within ± 25% and is calculated using the following equation:
 

Overall Error  =  ± ( |mean bias|  +  2[CVT pooled])100%

CV₁ pooled was used instead of CV_T pooled in this study. In Section 3 the low end of the range for the overall error for "M" dust analyzed by XRD was reported as ± 91%. Regardless, all the XRD work exceeds a 25% cutoff for overall error. Only the "M" dust XRF results satisfy an overall error limit of ±25%.

8.1 X-ray diffraction (XRD) is the method of choice in identifying V₂O₅. Due to the significant dependence of XRD analytical sensitivity upon particle-size distribution, XRD is only used as a confirmation technique for this analyte. Analytical lines are available for qualitative verification in addition to the line available for quantitation. In order to quantitate using XRD, the standard material must approximate the analyte particle-size distribution of the samples. This may not be practical for the OSHA dust and fume standards for V₂O₅. If fume is present in an operation, the results in Section 5.2 indicate that recovery may be low due to the reduced XRD analytical sensitivity for finer particles. Therefore, XRD is used for confirmation only.

8.2 X-ray fluorescence (XRF) is the method of choice in quantitating V₂O₅ because particle-size effects are much less severe for XRF compared to XRD. In XRF, the V K_ß peak is also available for qualitative verification of V.

8.3 Due to the common sample preparation technique and superior performance of the XRF methodology, this work suggests a hybrid method incorporating quantitation by XRF and chemical species verification (and semi-quantitative support) by XRD. There is a potential for multi-analyte analyses by such a hybrid approach. If the industrial hygienist desired, both respirable V₂O₅ and respirable quartz [which coexist in certain industrial operations (9.4)] could be determined on the same prepared sample if the quartz sampling procedure is employed (9.17)

8.4 The XRD method was patterned after the NIOSH study (9.4). A discussion of the effects of deviation in V₂O₅ methodology between OSHA and NIOSH can be found in reference 9.18.

8.5 Concluding remarks:

As noted above, there was good agreement between the AAS and XRF results in this study for all three particle-size distributions. The major biases observed for the XRD analyses are most readily explained as due to the change in sensitivity with respect to particle size.

The results in this report support the proposition that the quantitative analysis of V₂O₅ by XRD would require close matching of the particle-size distribution of the standard material to that of air samples collected during industrial operations on PVC filters. As seen in the samples subjected to removal of coarse particles by sedimentation ("Fine-W"), XRD evidenced decreased analytical sensitivity when compared to both XRF and AAS. As shown in Section 5.2, there was a large positive bias in the XRD analyses when the particle-size distribution was biased towards larger particles. This was most clearly shown when the analyses of V₂O₅ finely ground in a freezer mill ("W" samples) are compared to the analyses of V₂O₅ more coarsely ground in a mortar and pestle ("M" samples). The coarse material provided a doubling of recoveries when compared to the recoveries of the fine material. The quantitative analysis of V₂O₅ by XRF was more immune to the particle-size distribution, thus giving improved recoveries and better precision than analysis by the XRD method. Aerosol generation and particle sizing would be advantageous in more fully evaluating these particle-size effects.

9.1 Occupational Safety and Health Administration Technical Center: Confirmation of Vanadium Pentoxide in Workplace Atmospheres by M.C. Rose (USDOL/OSHA-SLTC Method No. ID-185). Salt Lake City, UT. Revised 1991.

9.2 Occupational Safety and Health Administration Technical Center: Metal and Metalloid Particulated in Workplace Atmospheres (ICP Analysis) by J.C. Septon (USDOL/OSHA-SLTC Method No. ID-125G). Salt Lake City, UT. Revised 1991.

9.3 Occupational Safety and Health Administration Analytical Laboratory: OSHA Manual of Analytical Methods edited by R.G. Adler (Method No. I-1). Salt Lake City, UT. 1977.

9.4 Carsey, T.P.: Quantitation of Vanadium Oxides in Airborne Dusts by X-ray Diffraction. Anal. Chem. 57:2125-2130 (1985)

9.5 Gelman Sciences: The Filter Book. Ann Arbor, MI: Gelman Sciences, 1991

9.6 Occupational Safety and Health Administration Analytical Laboratory: X-ray Diffraction Program Documentation. Salt Lake City, UT. 1981 (unpublished)

9.7 Leroux, J. and C. Powers: Direct X-ray Diffraction Quantitative Analysis of Quartz in Industrial Dust Film Deposited on Silver Membrane Filters. Occup. Health Rev. 21:26-34:26-34 (1970).

9.8 National institute for Occupational Safety and Health: Collaborative Tests of Two Methods for Determining Free Silica in Airborne Dust (DHHS Publication No. 83-124). Cincinnati, OH: National Institute for Occupational Safety and Health, and the Bureau of Mines, 1983.

9.9 Mandel, J.: Accuracy and Precision, Evaluation and Interpretation of Analytical Results, The treatment of Outliers. In Treatise on Analytical Chemistry, 2nd ed., edited by I. M. Kolthoff and P. J. Elving. New York: John Wiley and Sons, 1978. pp. 282-285.

9.10 Youden, W.J.: Statistical Methods for Chemists. New York: John Wiley and Sons, 1964. p 20.

9.11 Occupational Safety and Health Administration Analytical Laboratory: Precision and Accuracy Data Protocol for Laboratory Validations. In OSHA Analytical Methods Manual. Cincinnati, OH: American Conference of Governmental Industrial Hygienists (Pub. No. ISBN: 0-936712-66-X), 1985.

9.12 Long, G.L. and J.D. Winefordner: Limit of Detection -- A Closer Look at the IUPAC Definition. Anal. Chem. 55: 712A-724A (1983).

9.13 Analytical Methods Committee: Recommendations for the Definition, Estimation and Use of the Detection Limit. Analyst 112:199-204:199-204 (1987).

9.14 Westgard, J.O. and M.R. Hunt.: Use and Interpretation of Common Statistical Tests in Method Comparison Studies. Clinical Chemistry 19:49 (1973).

9.15 Ripley, B.D. and M. Thompson: Regression Techniques for the Detection of Analytical Bias. Analyst 112:337-383:337-383 (1987).

9.16 Gore, W.L.: Statistical Methods. New York: Interscience, 1952, pp. 189-191.

9.17 Occupational Safety and Health Administration Analytical Laboratory: OSHA Analytical Methods Manual (USDOL/OSHA-SLCAL Method No. ID-142). Cincinnati, OH: American Conference of Governmental Industrial Hygienists (Pub. No. ISBN: 0-936712-66-X), 1985.

9.18 Occupational Safety and Health Administration Analytical Laboratory: The effects of Deviation in Methodology - OSHA vs. NIOSH Results by M.C. Rose. Salt Lake City, UT. 1987 (unpublished).

Table 1
 
Calibration Data
Freezermill ("W") Material
 
APD-A X-Ray Diffractometer

V2O5 µg Taken	Counts	µg Calculated	Mean	Std Dev	CV1
50   50   50	818   827   860	46.9   47.4   49.3
		(46.4 - 49.4)*	47.9	1.27	0.0264
100 100 100	1522 1634 1718	87.8   94.4   99.3
		(87.2 - 100.4)	93.8	5.77	0.0615
200 200 200	3444 3573 3306	202.2 210.0 193.9
		(192.7 - 211.3)	202.0	8.05	0.0399
250 250 250	4040 4195 4509	238.5 248.1 267.4
		(234.4 - 268.2)	251.3	14.72	0.0586
499 499 499	8054 9131 9270	495.3 568.2 577.7
		(495.3 - 598.9)	547.1	45.08	0.0824
998 998 998	14909   13309   15791	969.3 855.8 1032.7
		(849.5 - 1055.7)	952.6	89.62	0.0941
1996   1996   1996	29359   27312   27943	2096.7   1923.8   1976.5
		(1897.1 - 2100.9)	1999.0	88.62	0.0443
2495   2495   2495	35312   34389   35014	2632.4   2545.7   2604.3
		(2543.2 - 2645.0)	2594.1	44.24	0.0171

* Acceptable ranges from the ASTM test are shown in parentheses.

Calibration fit spliced at 500 µg, 8125 counts:
Low fit :  High fit:	Counts  =   Counts  =	0  +  17.562281   ×   µg  -  0.002626   ×   µg2 399.9865  +  15.962334   ×   µg  -  0.001026   ×   µg2

Table 1 (Continued)
 
Calibration Data
Freezer mill ("W") Material
 
APD-B X-ray Diffractometer

V2O5 µg Taken	Counts	µg Calculated	Mean	Std Dev	CV1
50   50   50	1327   812   974	*   43.5   52.3
		(Test not applicable)	47.9	6.22	0.1299
100 100 100	1736 1593 1888	94.0   86.1 102.4
		(84.8 - 103.6)**	94.2	8.15	0.0865
200 200 200	3691 3720 3922	204.5 206.1 217.9
		(201.1 - 217.9)	209.5	7.32	0.0349
250 250 250	4440 4358 4456	248.2 243.4 249.2
		(243.3 - 250.5)	246.9	3.10	0.0126
499 499 499	8097 9462 9216	475.9 567.9 551.0
		(475.3 - 587.9)	531.6	48.97	0.0921
998 998 998	15331   13357   15996	968.6 831.7 1015.2
		(828.8 - 1048.2)	938.5	95.38	0.1016
1996   1996   1996	31381   29308   29709	2163.3   1999.8   2031.2
		(1965.0 - 2164.6)	2064.8	86.76	0.0420
2495   2495   2495	36411   36811   34880	2573.9   2607.4   2446.8
		(2445.3 - 2640.1)	2542.7	84.72	0.0333

*	One of the 50 µg standards appeared as an outlier. Although the result was acceptable on APD-A, the standard was not used for calibrating APD-B.
**	Acceptable ranges from the ASTM test are shown in parentheses.

Calibration fit spliced at 500 µg, 8461 counts:
Low fit :  High fit:	Counts  =   Counts  =	0  +  18.836058   ×   µg  -  0.003830   ×   µg2 773.2365  +  15.743113   ×   µg  -  0.000737   ×   µg2

Table 1 (Continued)
 
Calibration Data
Freezer mill ("W") Material
 
EDXRF

V2O5 µg Taken	Counts	µg Calculated	Mean	Std Dev	CV1
50   50   50	246   206   220	60.7   50.6   54.2
		(49.3 - 61.0)*	55.2	5.12	0.0928
100 100 100	412   359   415	102.4   89.1 103.2
		(89.1 - 107.3)	98.2	7.92	0.0806
200 200 200	869   785   848	217.2 196.1 212.0
		(195.8 - 221.1)	208.4	10.99	0.0527
250 250 250	1005   976 1015	251.4 244.1 253.9
		(244.0 - 255.7)	249.8	5.09	0.0204
499 499 499	2076 2178 2093	520.5 546.2 524.8
		(514.7 - 546.3)	530.5	13.77	0.0259
998 998 998	3924 3579 3811	984.9 898.2 956.5
		(895.7 - 997.3)	946.5	44.20	0.0467
1996   1996   1996	8361 7419 7750	2099.7 1863.0 1946.2
		(1831.6 - 2107.8)	1969.7	120.07	0.0610
2495   2495   2495	9941 9826 10449	2496.8   2467.8   2624.4**
		(2423.9 - 2615.5)	2529.7	83.31	0.0329

*	Acceptable ranges from the ASTM test are shown in parentheses.
**	Although this standard appeared outside of the ASTM test range, it was used in the calibration.

EDXRF curve fit:
Counts  =  4.427507051  +  3.979798615   ×   µg  +  0   ×   µg²

Table 1 (Continued)
 
Calibration Data
Freezer mill ("W") Material*
 
EDXRF

V2O5 µg Taken	Counts	µg Calculated
50 100 200 250 499 998 1996   2495	210   404   756   945 2114 3873 7754 10146	41.1   93.7 188.6 239.3 548.9 1002.4   1956.8   2517.0

*	Standards prepared from this material were used to analyze fume-like "Fine-W" samples.

EDXRF curve fit:
Counts  =  59.00722287  +  3.670421406   ×   µg  +  1.33910E-04   ×   µg²

Table 2
 
Analysis - Spiked Sample Data
Freezer mill ("W") Material
 
APD-A X-ray Diffractometer

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
237	(0.738 - 1.085)	6	0.911	0.0895	0.0982
474	(0.639 - 1.149)	6	0.894	0.1314	0.1470
710	(0.706 - 1.046)	6	0.876	0.0877	0.1001

Average Recovery CV1 (Pooled)

=  0.894   =  0.1173

---

APD-B X-ray Diffractometer

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
237	(0.690 - 1.028)	6	0.859	0.0871	0.1015
474	(0.601 - 1.174)	6	0.887	0.1476	0.1663
710	(0.731 - 1.059)	6	0.895	0.0846	0.0946

Average Recovery CV1 (Pooled)

=  0.880   =  0.1250

---

*   Acceptable ranges from the ASTM test are shown in parentheses.

Table 3
 
Analysis - Spiked Sample Data
Alumina Mortar and Pestle ("M") Material
 
APD-A X-ray Diffractometer

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
233	(1.483 - 1.841)	6	1.662	0.0923	0.0555
467	(1.510 - 1.960)	6	1.735	0.1158	0.0668
700	(1.450 - 1.818)	5**	1.634	0.1052	0.0644

Average Recovery CV1 (Pooled)

=  1.680   =  0.0624

---

APD-B X-ray Diffractometer

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
233	(1.560 - 2.158)	6	1.859	0.1543	0.0830
467	(1.629 - 2.176)	6	1.902	0.1414	0.0743
700	(1.622 - 2.060)	6	1.841	0.1129	0.0613

Average Recovery CV1 (Pooled)

=  1.867   =  0.0734

---

*	Acceptable ranges from the ASTM test are shown in parentheses.
**	One of the 700 µg spiked samples was damaged in transfer from APD-B to APD-A; it was not used in subsequent analyses.

Table 4
 
Analysis - Spiked Sample Data
Freezer-mill ("W") Material
EDXRF

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
237	(0.788 - 1.000)	6	0.8938	0.0547	0.0612
474	(0.632 - 1.064)	5**	0.8481	0.1235	0.1456
710	(0.744 - 0.992)	6	0.8681	0.0640	0.0737

Average Recovery CV1 (Pooled)

=  0.8713   =  0.0966

---

Table 5
 
Analysis - Spiked Sample Data
Alumina Mortand and Pestle ("M") Material
EDXRF

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
233	(0.825 - 1.126)	6	0.9759	0.0776	0.0795
467	(0.839 - 1.076)	6	0.9574	0.0612	0.0639
700	(0.895 - 1.028)	5**	0.9615	0.0377	0.0392

Average Recovery CV1 (Pooled)

=  0.9651   =  0.0642

---

*	Acceptable ranges from the ASTM test are shown in parentheses.
**	One sample was lost in analysis
***	One of the 700-µg spiked samples was damaged in transfer from APD-B to APD-A; it was not used in subsequent analyses.

Table 6
 
Analysis - Spiked Sample Data
Freezer-mill ("W") Material
AAS

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
237	(0.857 - 0.963)	6	0.9096	0.0273	0.0300
474	(0.843 - 0.957)	6	0.9004	0.0294	0.0327
710	(0.838 - 0.942)	6	0.8900	0.0268	0.0301

Average Recovery CV1 (Pooled)

=  0.9000   =  0.0309

---

Table 7
 
Analysis - Spiked Sample Data
Alumina Mortand and Pestle ("M") Material
AAS

V2O5 µg Taken	Recovery Range*	N	Mean	Std Dev	CV1
233	(0.938 - 1.001)	6	0.9692	0.0163	0.0168
467	(0.905 - 0.954)	6	0.9295	0.0124	0.0134
700	(0.875 - 0.924)	5**	0.8994	0.0137	0.0153

Average Recovery CV1 (Pooled)

=  0.9327   =  0.0152

---

*	Acceptable ranges from the ASTM test are shown in parentheses.
**	One of the 700-µg spiked samples was damaged in transfer from APD-B to APD-A; it was not used in subsequent analyses.

Table 8
 
µg V₂O₅ "Fine-W" Recoveries
Spiked on Silver Membranes
Reported by Analysis
 

Sample	APD-A	APD-B	EDXRF	AAS
A B C	190.2 199.3 259.9	188.0 227.2 256.0	309.5 329.0 373.9	297.9 344.9 371.5

Table 9
 
µg V₂O₅ "W" Recoveries
Spiked on PVC Membranes
Reported by Analysis

"True"	APD-A	APD-B	EDXRF	AAS
237 237 237 237 237 237   474 474 474 474 474 474   710 710 710 710 710 710	234.5 220.5 245.9 195.1 204.4 195.5   498.0 403.3 482.0 338.4 375.5 444.4   548.7 548.6 658.3 612.3 669.9 692.2	236.7 202.4 218.8 182.8 188.5 192.0   500.9 434.3 496.5 341.6 346.4 403.2   614.5 542.5 682.3 603.4 662.5 706.4	233.6 206.7 221.3 207.2 202.7 199.7   489.6 403.7 * 338.6 360.2 418.0   555.0 576.8 677.1 607.7 640.9 641.1	225.3 212.3 218.8 209.1 218.8 209.1   430.1 430.1 443.6 410.0 410.0 436.9   598.1 636.6 640.0 622.5 650.6 643.6

---

µg V₂O₅ "M" Recoveries

"True"	APD-A	APD-B	EDXRF	AAS
233 233 233 233 233 233   467 467 467 467 467 467   700 700 700 700 700 700	355.2 412.2 410.8 378.6 382.5 384.5   761.9 886.6 787.7 867.2 797.7 760.2   1079.0 * 1111.3 1230.0 1215.6 1081.9	363.5 443.3 442.7 433.5 467.3 448.5   857.9 990.2 807.5 941.8 864.1 868.9   1173.3 1337.8 1277.4 1382.2 1337.0 1223.3	222.3 255.7 218.5 202.2 228.8 236.8   484.6 482.3 427.3 425.5 437.9 425.0   666.8 * 654.4 709.7 689.4 644.9	228.5 222.0 231.8 225.3 222.0 225.3   426.8 433.5 436.9 443.6 433.5 430.1   615.5 * 636.6 626.0 640.1 629.5

* Sample lost in analysis.

Table 10
 
Detection Limit Determination
Blanks Prepared as Spiked Samples
 
100-s Analysis Time & Normalized Counts
 

Sample	APD-A Counts	APD-B Counts	EDXRF Counts
A B C D E F G	330 135 105 187 196 108 408	434 113 106 165 231 191 545	59 11 22 41 42 8 39
Average	209.86	255.00	31.714
SD	116.46	168.96	18.599
Slope (Count/µg)	17.562	18.836	3.9798
D.L.	19.9	26.9	14.0
Estimated DL	25 µg V2O5 (XRD)		14 µg V2O5 (XRF)

---

Detection Limit Determination
Blanks Prepared as Spiked Samples
 
1,000 s Analysis Time & Normalized to Compare with 100-s Results
 

Sample	APD-A Counts	APD-B Counts	EDXRF Counts
A B C D E F G H I	214 27 47 42 21 15 234 195 97	251 21 94 24 81 84 351 435 106	1.0 4.3 4.3 8.4 0 4.4 4.5 1.2 0
Average	99.11	160.78	3.12
SD	90.10	148.95	2.78
Slope (Count/µg)	17.562	18.836	3.9798
D.L.	15.4	23.7	2.10
Estimated DL	20 µg V2O5 (XRD)		2 µg V2O5 (XRF)

Where DL  =  3(SD) / sensitivity

Appendix 1

The multichannel analyzer was set to 512 channels and the instrument was calibrated using the TiO2-ZnO-Y2O3 calibration standard. The spectrum range was approximately 1.2 to 20.6 kV.

Finnigan EDXRF Program
 

	Changes for analysis on different days
Steps to be entered on instrument console	Analyses at 100 s	Analyses at 1,000 s
(Begin programming.)     LEARN	"Fine-W" and DL	DL
(Program system to set up analytical conditions.)
100 SEC     1ST HALF     CLEAR+CLEAR     MARKER (Z=23)     OUTPUT	100 s	1,000 s
(Program system to acquire spectrum.)     ACQUIRE
(Program the instrument to obtain net counts using background subtraction.)     NET     START/SPAN/SELECT   [Set background low energy region to ~3.4 kV (between Ag La and V Ka peaks).]         START  =  71         SPAN   =  5         INTENS   [Set background high energy region to ~5.9 kV (just beyond the V Kß peak).]
START  =  124         SPAN   =  5         INTENS	127	127
[Set region to span the V Ka analytical peak at 4.949 kV (~0.64 kV span.)]
START  =  91         SPAN   =  17         INTENS	93	93 14
(End of programming.)     EXECUTE

Note: This program was written for a specific instrument (Finnigan 77-900A). Commands are capitalized. Background regions should be adjusted when interferences are present.