INTRODUCTION

Effective assessment and remediation of hazardous waste sites dictates that analytical methodologies be developed which assist in the evaluation of site contamination and simultaneously make efficient use of sampling time and resources (1). Optimally, a technique would provide on-site personnel with immediate and accurate information concerning the identity and concentration of inorganic soil contaminants (2).

Inorganic pollutants can be readily determined in contaminated soils with energy dispersive X-ray fluorescence spectrometry (EDXRF) using a thermoelectrically cooled Si(Li) detector (3). A field mobile laboratory van or trailer can accommodate the EDXRF system because the electrically cooled detector, which provides high resolution EDXRF spectra, does not require cryogenic cooling. Soil sample preparation for EDXRF analysis is minimal, therefore, short turnaround times are realized between sampling and reporting results.

This report will describe an EDXRF method developed to determine four inorganic soil contaminants: lead, arsenic, zinc, and cadmium at four sampling depths. The EDXRF results for approximately one hundred eighty soil samples will be compared to results obtained for sample splits submitted for analysis at an independent laboratory. Evaluation of low concentration arsenic detectability with elevated lead concentrations in these samples will be discussed. Accuracy and precision of the EDXRF method will also be compared to the independent methods using a standard reference material and soil samples submitted in triplicate to both laboratories.

EXPERIMENTAL

The field mobile EDXRF spectrometer used in this work was a Spectrace Model 6000 (now replaced by the Thermo NORAN QuanX). The EDXRF system consists of three modules: the spectrometer, the control/pulse processing electronics, and the data analysis computer. The compact size and weight (90 lbs.) of the modules permits installation of the system in a laboratory trailer or van.

The bench top spectrometer module, which can accommodate a single soil sample, is powered by 110 V line or generator feed. The excitation source used is a low powered Rh anode X-ray tube (50 KV, 0.35 mA (17 W) maximum output) positioned at a 45^o incident angle to the sample. Three primary radiation filters permit optimum spectral acquisition conditions to be computer selected.

The thermoelectrically cooled Si(Li) X-ray detector is mounted at a 45^o take-off angle in an inverted geometry with respect to the sample. The 20 mm2 Si(Li) crystal, which is protected by a 0.5 mil Be window, is cooled to -90^oC for operation using a multi-stage thermoelectric (Peltier effect) cooler. The 300 watts produced at the detector heat sink are dissipated by forced ambient air. Thermoelectrically cooled detectors provide typical resolutions of 185 eV (Mn Ka).

A card cage module is interfaced between the spectrometer and a personal computer. The card cage components include the detector high voltage supply, the pulse processing electronics, and the control circuit board for the EDXRF spectrometer. The data analysis software executed on the PC is capable of either a fundamental parameters or empirical data treatment scheme using a combination of standard reference materials and/or site specific standards.

Sampling of the suspected waste site was performed using EPA approved protocols in a 9500’ x 3500’ rectangular area. Forty-three (43) cores were collected and partitioned into four depth levels: surface to 2"; 2" to 6"; 6" to 12"; and 12" to 18", and designated levels 1 through 4, respectively. At the site, samples were first homogenized and then split into two fractions. One was submitted for EDXRF analysis and the other sent to an independent lab for analysis.

The independent laboratory used EPA SW 846 (methods 3050 and 6010) methodology to determine Cd, Pb, and Zn concentrations in the soil sample splits. Arsenic was determined in those splits using SW 846 method 3050 and EPA method 206.4 (spectrophotometric).

Sample preparation for EDXRF analysis consisted of drying the sample for 4 minutes in a microwave oven followed by sieving the dried sample. Material passing through the 2 mm sieve was collected as sample and was free of large foreign objects such as pebbles and sticks. Drying the sample was required due to the variable moisture content in the submitted soils; some surface samples had the consistency of mud. The sieved soil was then ground in a Spex shatterbox grinder (Spex Ind., Edison, NJ) using tungsten carbide cups for 2 minutes. Grinding cups were subsequently cleaned using soap and tap water. The cleaned cups were rinsed with distilled/deionized water followed by isopropanol. Approximately 5 grams of prepared sample were poured into a disposable 32 mm X-ray sample cup and covered with a 6.3 um polypropylene film. Five grams of dried sample gave the equivalent of a 15 mm sample depth in the cup. Approximately twenty-five samples were prepared and analyzed per day.

STANDARDIZATION METHOD

Two sets of excitation conditions were employed to determine seven elements in the soil samples, four of which are of specific environmental concern: Zn, As, Pb, and Cd. Table 1 lists the two sets of spectral acquisition conditions and which conditions were used to determine each analyte.

SPECTRAL REGION	CONDITIONS	ANALYTES
MID Z	35 KV, 0.35 mA, O.13m Rh filter, 200 s livetime	Mn, Fe, Cu Zn, Pb, As
HIGH Z	50 KV, 0.35 mA, 0.63mm Cu filter, 200 s livetime	Cd

Table 1. Spectral acquisition conditions for the EDXRF analysis of soils.

Figure 1 is a mid Z spectrum of a soil sample that was found to contain 125 ppm As, 1100 ppm Pb, and 729 ppm Zn. A multiple linear least squares peak fitting routine was used for deconvolution of overlapped peaks.

Figure 1. Mid Z spectrum of a soil sample containing 1100 ppm Pb, 729 ppm Zn, and 
125 ppm As. Full scale on the y-axis is 2,000 counts.

The soil characterization method was standardized using four standard reference materials (SRM): NBS 1648 (urban particulate); NBS 2704 (river sediment); SO- 1 and SO-3, two soil standards available from the Canada Centre for Mineral and Energy Technology. Standards labeled NBS are available from the National Institute for Standards and Technology (NIST). These SRMs have certified concentrations of Fe, Mn, Cu, Zn, Pb, and Cd.

A fundamental parameters (FP) method (5) was employed as the data treatment scheme and used certified concentrations of Fe, Mn, Cu, Zn, Pb, and Cd in the four standard materials. To compute instrumental sensitivity (emission peak counts per second per ppm), the balance of the standard was assumed to be comprised of SiO2 to account for the contribution of the matrix on the measured analyte X-ray intensity. The balance component Si05 was selected to mimic the concentration of Si and 0 in typical soils, approximately 24% Si and 45% 0. Since none of the selected SRMs contain arsenic, As sensitivity (cps/ppm) was determined using a fundamental parameters theoretical calculation based on the computed Zn sensitivity. Table 2 lists the analyte sensitivities computed by the FP method.

ANALYTE	SENSITIVITY (cps/ppm)	LLD (ppm)
Mn	0.010	21
Fe	0.015	19
Cu	0.046	26
Zn	0.067	19
Pb	0.084	7
As	0.132	12
Cd	0.107	4

Table 2. Sensitivity and lower limits of detection for the analytes of interest.

There are some advantages to using an FP method for standardization compared to site specific soil standards. The FP method can use readily available, well-characterized SRMs to measure analyte sensitivities. Site specific soil standards, by contrast, are usually collected with a separate sampling mobilization. The FP method standardized with SRMs can provide accurate analyte concentrations to be determined in samples with fairly wide matrix variations without restandardization, unlike methods incorporating site specific standards.

RESULTS

Table 2 lists the lower limits of detection determined using the two sets of spectral acquisition conditions (Table 1). The pertinent equation is: LLD (ppm) = 3*sqrt(Ib)/m*sqrt(T), where Ib is the background intensity (cps), m is the analyte sensitivity (cps/ppm), and T is the acquisition livetime in seconds (6). Calculated LLD values are dependent upon spectrum acquisition times, sample matrix, and excitation conditions. The conditions in Table 1 were selected to optimize the Pb and Cd spectral regions. Improved LLDs are possible with EDXRF using longer spectrum acquisition livetimes and optimized excitation conditions for selected spectral regions.

Results for the determination of four analytes by EDXRF in 180 samples (43 cores at 4 levels, two SRMs, three samples in triplicate) were compared to independent analysis results in order to evaluate the level of agreement between the two methods. Table 3 lists the correlation plot data for the analytes in terms of actual slope, intercept, errors, and the correlation coefficient of the fit. Each analyte correlation plot included approximately 150 data points.

ANALYTE	SLOPE	INTERCEPT	CORRELATION COEFFICIENT
Pb	1.01 ± 0.03	10.0 ± 13.8	0.96
As	1.08 ± 0.05	0.98 ± 3.54	0.92
Cd	1.02 ± 0.03	3.09 ± 2.19	0.94
Zn	1.02 ± 0.02	63.0 ± 13.6	0.98

Table 3. Correlation plot data for the four analytes of environmental interest.

As shown in Table 3, slopes of the plots for Pb, Cd, Zn, and As are within 8% of 1.00 and all correlation coefficients are greater than 0.92. The calculated slope near 1.00 and correlation coefficients greater than 0.90 indicates agreement between the two analytical techniques. Figure 2 is a plot of 94 data points in the range of 0 to 300 ppm Pb. Figure 3 is a plot of 110 EDXRF and ICP analyzed samples in the range of 0 to 100 ppm Cd and also indicates agreement between the results of the two methods.

Figure 2. Pb correlation plot for 94 samples.

Figure 3. Cd correlation plot for 110 samples.

ACCURACY AND PRECISION

To evaluate the accuracy provided by the EDXRF method two SRMs were submitted as unknowns for EDXRF analysis as well as being submitted to the independent lab for analysis. Table 4 lists the results for SRM SO-2. EDXRF analysis of SO-2 provides results that are in good agreement with certified values. The independent ICP analysis of zinc in SO-2, however, is biased low by a factor of one-half.

Sample	ANALYTE	ICP	EDXRF	Certified
SO-2	Pb Zn	19 55	17 123	21 124

Table 4. Results of the analysis of SRM SO-2 by ICP and EDXRF
methods. All values in ppm.

Precision was evaluated by submitting three samples a total of three times for independent and EDXRF analysis. Table 5 shows the results for the two methods along with the calculated standard deviation (in ppm) of the three replicate analyses. Note that Cd in sample C was only reported by EDXRF to the nearest 1 ppm and three values of 9 ppm Cd were determined, hence the zero standard deviation for the three replicates. EDXRF precision is better than 10% relative standard deviation in all but one case (As in sample C) and compares well with that provided by the independent lab.

Sample	Element	Ind. Lab	EDXRF
A	As Cd Pb Zn	45 ± 4 20 ± 2 286 ± 28 185 ± 15	41 ± 3 31 ± 3 312 ± 12 134 ± 10
B	As Cd Pb Zn	17 ± 3 80 ± 6 141 ± 15 556 ± 39	14 ± 1 58 ± 4 158 ± 3 529 ± 46
C	As Cd Pb Zn	17 ± 1 10.0 ± 0.9 117 ± 8 173 ± 26	19 ± 4 9 ± 0 142 ± 14 128 ± 3

Table 6. EDXRF and independent lab results for three soil
samples each analyzed in triplicate. All values in ppm.

CONCLUSION

Field mobile EDXRF analysis of soils suspected of being contaminated provides information concerning the nature, extent, and magnitude of the contamination. Due to the minimal sample preparation necessary for EDXRF analysis, sampling to result turnaround time is relatively short so the most effective use of sampling resources is realized. EDXRF detection limits below 20 ppm were obtained for the elements of environmental concern. The effect of increasing lead concentration on arsenic detectability was quantified. Using the EDXRF method described here, reliable As results were found for those samples containing As/Pb concentration ratios above 0.083. Accuracy and precision for the analytes of interest using the EDXRF method was shown to be comparable to results obtained by independent analysis. Comparable results for Cd, As, Pb, and Zn between independent and EDXRF methods validates the use of EDXRF analysis for hazardous waste site investigation and remediation.

ACKNOWLEDGMENT

The author would like to acknowledge James P. Walsh and Associates for site sampling and providing the independent analysis data.

REFERENCES

(1) Vincent, H.; Field Screening Methods for Hazardous Waste Site Investigations Symposium Proceedings, (1988), 61.

(2) Perlis, R.; Chapin, M.; "Low Level XRF Screening Analysis of Hazardous Waste Sites", Field Screening Methods for Hazardous Waste Site Investigations Symposium Proceedings, (1988), 81.

(3) Madden, N.W.; Hanepen, G.H.; Clark, B.C.; "A Low Power High Resolution Thermoelectrically Cooled Si(Li) Spectrometer", IEEE Trans. Nuc. Sci., (1), (1986), 303.

(4) Data Oualitv Obiectives For Remedial Response Activities, U.S. EPA, EPA 540/G- 87,003, Appendix A, (1987).

(5) Criss, J.W.; Birks, L.S.; "Calculation Methods for Fluorescent X-ray Spectrometry- Empirical Coefficients vs. Fundamental Parameters", Analytical Chemistry, 40, (1968), 1080.

(6) Currie, L.A.; "Limits for Qualitative Detection and Quantitative Determination", Analytical Chemistry, 40, (1968), 586. 33,