Heavy metals by XRF

This project shows the performance of the hand held XRF (X Ray Fluorescence) analyser used for classifying heavy metal contaminated soil into hazardous and non hazardous waste. The instrument used was the Oxford Instruments XMET-3000TX.

The site was in a city centre with buildings on all sides. An earlier SI had identified very high (>50,000 ppm) concentrations of Pb, Zn and other heavy metals across the site. In order to create the foundations for the new building, deep pilings needed to be constructed. The original proposal was to remove the top layers of soil across the site and the piling arisings and dispose of this material as hazardous. The contractor, on the advice of their consultant, decided to assess on site analysis as a potential way of reducing the very expensive disposal costs anticipated.

In order to fully identify the extent of the metals contamination so that the areas that were hazardous and non hazardous could be accurately delineated, a high density sampling survey was carried out. Shallow trial pits were dug (<1m deep) at about 3m intervals and the XRF analyser used to scan 3 points in each trial pit, 1 at the base and 2 at the sides. The spread of the 3 results indicated the homogeneity of the contamination and the average of the 3 results was used to estimate the total metals concentration. The speed of the XRF analysis allowed 60+ trial pits to be constructed per day. As the project objective was to delineate the hazardous and non hazardous area, the XRF was set up for a 20 second scan on the FP setting. This low scan time would mean metal concentrations below 250 ppm would not be very accurately reported, but the high concentrations anticipated would not overload the detector and potentially give false negative results.

During this phase of the project, selected samples from the site were sent to the lab for MCERTS quality analysis to confirm the performance of the XRF. The results are shown below

The results highlighted in green show that all 3 results from the trial pit were within 50% of each other. Samples highlighted in yellow were within 100% of each other. Samples in orange and red were within 200% and >400% respectively. Samples showing as red or orange were usually near the edge of the dividing zone that divided the hazardous and non hazardous area.

The lab samples consisted of approximately 1.5kg of soil from each area. These had the very large lumps of brick and stone removed and the remainder homogenised by mixing. A 100g portion of this was removed and then dried and crushed to a <0.5mm particle size. Approximately 1g of this fraction was then extracted in aqua regia and analysed by ICP.

The remaining crushed portion was re-analysed by the XRF analyser using the 20 second scan setting using the FP calibration.

The lab results usually gave a slightly higher value than the on site XRF results, but this is probably due to the fact that the on site method did not try to remove the brick and stone present at the point of analysis as the aim was to assess the contamination at several relatively undisturbed points in the trial pit. The lab by removing the larger pieces of relatively uncontaminated brick and stone from the sample, effectively concentrates the contamination relative to the original sample.

The XRF results on the crushed sample that the lab analysed shows very good correlation for Lead (Pb) because the Pb concentration was sufficiently high to be within the accurate part of the FP calibration. The other metals were usually below the nominal 250 ppm quantitation limit and so the accuracy of the quantitation is lower. Using these scan parameters however, the absolute detection limit for the target metals was still around 50 ppm. This means that the analyser would be able to indicate if the target metals were present in the sample, even at concentrations well below the hazardous trigger threshold. The aim of this project was to establish the hazardous and non hazardous areas and samples with total metal concentrations below 2000 ppm were considered non hazardous. The data quality objective of being able to use the XRF on a 20 second scan to identify if the soil is above a nominal 2500 ppm was met in 13 out of 14 samples. Of the one sample where the on site XRF result did not agree with the lab result, the XRF result on the prepared lab sample did agree, showing the XRF method was working correctly. The difference in the lab and field result for this sample was probably due to the poor homogeneity of the soil in that area, as indicated by the red colour from the on site XRF result.

8 of the 14 samples tested by the lab and the XRF were borderline for classifying the soil as hazardous or non hazardous. These samples were selected to comprehensively demonstrate the technique was suitable for soil across the range of concentrations by showing the frequency of either classifying non hazardous material as hazardous (false positive) or classifying hazardous as non hazardous (false negative) was low. The results show that the frequency of false positive and false negative was very low.

The very good agreement with the on site result and the lab result is remarkable because the way the XRF analysis was done in the field. The homogeneity check was not expected to give such good correlation because the soil at the site was made ground with a lot of brick and stone present. The technique does indicate that by taking the average of several spot samples from a trial pit, a reasonable estimate of the metal concentration can be made. The usual sampling method employed for XRF analysis is to take a 500g sample, place it in a bag and mix the soil to make a representative sample.

The other metals shown in the above table are for interest only. The detection limits for Cr at this scan time is around 250 ppm, with a very poor standard deviation. Cr is one of the lightest of the “heavy metals” and requires a much longer scan time to accurately quantify its presence. For industrial applications a chromium concentration of >5000 ppm is needed to exceed the SGV guidelines, and this site was not expected to contain much chromium. Interestingly, chromium is naturally occurring in many soil types up to 800 ppm. The mineral form the chromium takes is however often resistant to extraction by aqua regia digest. This means that the lab may not achieve 100% extraction efficiency. XRF however is not constrained by extraction and can “see” all of the metal in the sample. This is often why the XRF result is higher than the lab result. Arsenic minerals can also be difficult to fully extract using aqua regia and this can also lead to the XRF result being higher than the lab result. The 20 second XRF scan makes a good arsenic result unlikely, but the data shows that none of the samples contained more than 200 ppm, with most being below 50 ppm.

The site may have been used to produce pigments for paint and pottery production. Nickel and Chrome salts were commonly used to produce intense colours in pottery. These pigments however are quite difficult to dissolve, even in acid.

The verification data produced above persuaded the contractor to use the XRF system to manage the excavation and disposal of the soil. The delineation exercise showed much less soil contained hazardous levels than was originally suspected and the depth of the contamination was also less than originally thought. A QROS chemist was put on site to manage the excavation and confirm only hazardous material was removed to the hazardous landfill.

The use of on site analysis saved the client approximately £500,000. The amount QROS charged for the on site work was just £7,000, giving a very significant cost reduction to the contractor.