EXPERIMENTAL

The Spectrace Model 5000 was used for this analysis (now replaced with the Thermo NORAN QuanX). This EDXRF spectrometer is equipped with a liquid nitrogen cooled Si(Li) detector (0.3 mil Be window), and a 50 KV, 1 mA (maximum) Rh anode X-ray tube. Spectrace Model 5000 software  automates spectrometer parameters and sample tray position. All data analysis software described comes standard on Spectrace Model 5000 and QuanX installations.

Standard Mylar films and unknown Teflon filters were analyzed as received. Films and filters were held on a ten position automated sample tray that accommodates either 32 or 47 mm diameter circles.

STANDARDIZATION

Standard samples used were provided by Micromatter, Inc., Deer Harbor, WA. The eight standard thin films are listed in Table 1.

Element/ Compound	Loading (mg/cm2)	Element/ Compound	Loading (mg/cm2)
NaCl	45.8	V	45.1
Al	38.8	Fe	51.1
CuS	54.9	Pb	46.3
KCl	47.9	Sn	46.6

Table 1. Standards used for EDXRF analysis of air filter samples.

Initially, five sets of acquisition conditions, listed in Table 2, were applied to each air filter.

Acquisition parameter	Condition 1	Condition 2	Condition 3	Condition 4	Condition 5
Tube voltage (KV)	11	30	5	35	50
Tube current (mA)	0.80	0.99	0.99	0.99	0.99
Filter	Cellulose Whitman 41 (6 sheets)	Thin Pd 0.05 mm	None	Thick Pd 0.13 mm	Copper 0.63 mm
Chamber atmosphere	Vacuum	Air	Vacuum	Air	Air
Counting time (sec)	400	400	400	400	400
Elements analyzed (italics indicate detected elements)	Al, Si, P, S, Cl, K, CaAl, Si, P, S, Cl, K, Ca	Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn	Na, MgNa, Mg	As, Se, Br, Rb, Sr, Y, Zr, Mo, Pb, Au, Hg, U	Pd, Ag, Cd, In, Sn, Sb,Ba

Table 2. Spectral acquisition conditions selected for air filter analysis.

Figures 1 through 5 illustrate the spectra acquired from each of the five X-ray analysis conditions. Those elements in Table 2 listed in italics were detected in at least one of the eight submitted air filter samples. Total analysis time was 37 minutes per sample.

Figure 1.  Sample spectrum from Analysis Condition Set 1.Click to see a high resolution version of this image.	Figure 2.  Sample spectrum from Analysis Condition Set 2.Click to see a high resolution version of this image.
Figure 3.  Sample spectrum from Analysis Condition Set 3.Click to see a high resolution version of this image.	Figure 4.  Sample spectrum from Analysis Condition Set 4.Click to see a high resolution version of this image.
Figure 5.  Sample spectrum from Analysis Condition Set 5.Click to see a high resolution version of this image.

Since using condition 5 did not result in detection of an element in the spectral range optimized by condition 5, for subsequent determinations only conditions 1 through 4 were employed. This reduced the total analysis time to 29 minutes per sample.

The analysis method used for this work was a Fundamental Parameters (FP) thin film routine that is capable of determining 19 elements and an intermediate thin film thickness, simultaneously. The FP software option can also determine, from first principles, the sensitivities for elements that may not have a physical standard in the standards file. For example, since only those elements listed in Table 1 were used to quantitative the detected elements in Table 3, the sensitivities for Mg, Si, Ca, Mn, Ni, Zn, and Br were determined with this method.

BLANK DETERMINATION

For optimum lower limits of detection, the absence of stray lines in blank spectra is critical. Figure 6 shows four spectra illustrating the effect of stray lines in blank spectra. The upper two spectra are a blank and a sample spectrum from the Spectrace 5000. Comparing this blank spectrum to a blank spectrum acquired from a secondary target (Zr- lower spectrum in Figure 6) the Cu and Fe stray lines are clearly indicated in the secondary target spectrum. Stray lines in the secondary target excited spectrum have the deleterious effect of elevating the achievable lower limit of detection of those elements, in addition to any elements with emission lines that would overlap the stray lines, such as Mn and Zn.

Sample 010 (blank) Thick Pd filter 35KV	Sample 031 (97H) Thick Pd filter, 35KV
Sample 009 (blank) Zr secondary, 35 KV	Sample 030 (97H) Zr secondary, 35KV
Figure 6. Comparison of Spectrace 5000 spectra (top two) and Zr secondary target spectra (bottom two)

Four of the eight samples exhibited extremely low loading, they were 011, 014, 018 and sample 022. Only those analytes found to have greater than 3 times the counting statistics error are reported. Table 3 lists four samples that could be considered blanks. The total loading does not exceed 120 ng/cm² on any of the filters. For the blank correction discussed later, the values reported for sample 014 were used.

Element	Sample 011	Sample 014	Sample 018	Sample 022
Na	94.5 + 7.6	75.9 + 7.4	81.8 + 7.9	91.5 + 8.0
Mg	20.2 + 3.6	28.3 + 3.6	25.2 + 3.8	23.1 + 3.6
Ca				4.3 + 1.4
Cu	2.3 + 0.6		2.0 + 0.6
Loading	117.0	104.2	109.0	118.9

Table 3. Blank determinations. All values in ng/cm².

The error value refers to counting statistics error and is computed with equation 1:

              Equation 1

where,

Cts_b = the background counts in analyte peak fitting region;
Cts_p = the net peak counts for the analyte;
Conc. = concentration predicted by fundamental parameters.Lower Limits of Detection

The lower limits of detection (LLDs) listed in Table 4 were computed using equation 2.

                                   Equation 2

where,

Cts_b = the background counts in analyte peak fitting region measured from the blank;
Cts_p = the net peak counts for the analyte measured from an air filter with significant loading of the element;
L_s = the loading of the analyte determined by the FP method.

Element	LLD (ng/cm2)	Element	LLD (ng/cm2)
Na	12	Mn	3.5
Mg	3.0	Fe	2.6
Al	21	Ni	1.5
Si	12	Cu	1.4
S	6.0	Zn	1.0
Cl	3.2	Br	1.8
K	2.4	Pb	4.0
Ca	2.9

Table 4. Lower limits of detection (LLD) for fifteen elements on air filter media.

DETECTION LIMIT CHART

Figure 7. Lower limit of detection estimates attained with six filter and voltage selections (sixth condition added: 15 KV, 0.99 mA, aluminum filter, 400 s). Click the figure to see a high resolution version of the chart.

AIR FILTER ANALYSIS RESULTS

Table 5 lists the loading determined for the four remaining air filter samples. The reported loading for Na and Mg are corrected for blank 014. Open cells in the table represent non-detected elements (a loading reported at less than 3 times the counting error expressed as loading).

Element	Sample 002	Sample 008	Sample 031	Sample 024
Na	52.0 + 10.6	56.7 + 14.7	152.4 + 11.9	78.8 + 12.0
*Mg			7.7 + 5.4	11.7 + 5.5
Al		58.9 + 18.5		47.2 + 14.5
Si	74.9 + 9.2	177.3 + 12.9	111.3 + 10.2	40.8 + 9.8
S	770.9 + 11.8	2630.1 + 21.8	990.5 + 13.5	1429.6 + 16.1
Cl	7.4 + 2.1	8.8 + 2.9	38.1 + 3.0	7.8 + 2.5
K	35.8 + 1.7	59.3 + 2.0	51.0 + 2.0	25.4 + 1.6
Ca	23.3 + 1.7	58.9 + 2.1	47.1 + 2.0	25.8 + 1.7
Mn		5.0 + 1.6		6.8 + 1.6
Fe	61.0 + 2.3	102.1 + 2.9	94.5 + 2.9	69.5 + 2.6
Ni	4.6 + 0.8	4.9 + 0.8	5.6 + 0.9
Br	2.8 + 0.8	3.1 + 0.8
Zn	9.3 + 0.7	16.5 + 0.9	18.1 + 0.9	22.5 + 1.0
Cu		5.8 + 0.7	3.8 + 0.8	3.6 + 0.7
Pb	5.2 + 1.7	13.0 + 2.0	13.7 + 2.3	7.4 + 2.0
Loading	1047.3	3200.3	1569.3	1776.9

Table 5. Results of four air filters. All values in ng/cm².

*After correcting for Mg in the blank, loading is less than 3 times the counting error loading.

PRECISION AND ACCURACY

Precision of the analysis was measured by replicate analysis of sample 031. Table 6 lists the repeatability for ten analyses that includes the mean loading, standard deviation of ten replicates, and a relative standard deviation (%).

Element	Mean	Std. Dev.	RSD(%)
Na	174.7	15.7	6.26
Mg	13.4	5.5	13.25
Si	94.0	9.9	10.54
S	984.2	12.4	1.26
Cl	43.0	2.8	6.52
K	53.7	2.5	4.69
Ca	44.6	2.4	5.49
Fe	95.5	2.9	3.06
Ni	7.1	1.1	16.08
Zn	18.2	1.1	6.26
Pb	12.7	2.8	22.13
Loading	1597.8	37.20	2.19

Table 6. Precision results of sample 031. All values in ng/cm².

Accuracy of the method was determined by analyzing as unknowns four Micromatter thin film standards that were not used in the FP standardization. Table 7 reports the vendor loading value, loading determined by the FP method, and the relative error.

Element	FP Loading	Given	Relative Error (%)
Co	51.5	50.6	1.8
Cu	39.4	43.1	-8.6
Cr	48.2	45.3	6.4
Ca	24.3	25.2	-3.6

Table 7. Accuracy of FP model on standard thin films. All values in mg/cm².

Note: Units in **mg/cm²**

CONCLUSION

Filtered direct EDXRF spectrometry has proven a valuable tool in the determination of airborne particulate composition on air filters. The absence of stray lines in blank Teflon filter spectra, the accuracy within 10% relative, and detection limits in the range of a few of ng/cm² in a 30 minute analysis duration makes EDXRF superior to several other analytical techniques. Additionally, the non-destructive nature of EDXRF results in the samples being preserved after the analysis.