Haven’t conducted a pXRF program in awhile? No worries, this post has you covered… to remind you to never ever forget to calibrate your equipment.

Like any geochemical program, you’re also going to need to setup a rigorous QA/QC program complete with standards (matrix-matched certified reference material), blanks (pure quartz) and duplicates. 

QA/QC Theory

Successful pXRF QA/QC will prevent contamination effects and allow for the accuracy and repeatability of results. 

• Contamination: a silicon blank tests the cleanliness of the analyzer (the front window). This should only measure silica and should be inserted at regular intervals during batch testing.
• Accuracy: ideally a matrix-matched certified reference material (MM-CRM) is analyzed approximately once per every 20 samples (~5% of samples). This tests the accuracy of the measurement by comparing it to a known value, as determined by multiple laboratories.
• Repeatability: this checks the stability of the analyzer by taking multiple measurements from the same sample (i.e., duplicates).

If there is a question on whether it is instrument error or sampling error:

• Take multiple readings from the sample in different areas to determine the heterogeneity of the sample, i.e., sample error.
• Take multiple readings from the sample in the same area to determine instrument error; improving sample preparation will likely help to improve this type of error.

It is also extremely important to:

• Calibrate for all ranges of elements you are interested in analyzing.
• Ensure that the sample type (pulps, coarse rejects, drillcore, etc.) are the same.
• Account for sample matrix effects by creating separate calibrations.
• Have consistent sample containers.
• Use consistent test times for the detector beams.

A good rule of thumb is to run QA/QC every 1 out of 20 samples in a pXRF program. The first sample of the day should be the blank to ensure that there is no contamination on the detector window. Inserting the blanks, MM-CRM and duplicates at a rate of ~5% is best practice.

Quick Tips For Sample Analysis:

• Ensure that the detector window is clean (e.g., dust, fingerprints, etc.).
• Place the detector window as flat as possible on the sample; smooth uniform surfaces (i.e., diamond saw) enable complete contact between the instrument and sample.
• Ensure that the samples are dry.
• Due to the small spot size of the instrument, only a small proportion of the sample will be analyzed. Therefore, it is important to take samples from areas with the smallest grain sizes to increase the homogeneity of the sample. Thus, the sample is more representative of the material analyzed.

Do not move the instrument when taking a sample.

pXRF Calibration

No analytical method is good over the entire range of concentrations potentially encountered with a single calibration. Properly calibrated, pXRF has a range of concentrations over which the linear calibration is assumed to hold for any particular element. It is important to note that pXRF typically underreports concentrations when a calibration range is exceeded.

The issue of calibration range is not necessarily a problem if one is simply “screening” rocks or soils for concentrations above or below a decision-making threshold. However, it is an issue if one is interested in estimating the average concentration of an element or when trying to establish comparability between pXRF results and a corresponding laboratory analysis.

Site Specific Calibration

It is important in pXRF analysis to develop site-specific calibrations, for example, high/low gold values or rock type 1/2/3. When properly calibrated, the pXRF will have a range of concentrations over which the linear calibration is assumed to hold for any particular element. This typically ranges from the instrument’s detection limit up to the percent range of concentrations. The pXRF will not accurately report concentrations above its calibration range. 

In order to determine if the pXRF needs a calibration factor for a specific element, plot the lab chemistry versus pXRF data of at least twenty samples (recommend to keep it to a similar matrix) looking for two features:

1. The data is correlated linearly with an R2 = ~1

2. There is no major offset of the data, which shows that the pXRF data is either under- or overestimating the lab chemistry data.

Olympus made a great video about this – LINK

Pro Tips from Around the Globe

• If your pXRF is not pressure (altitude) corrected that your calibration(s) should be done at site not back at at the sea level lab…this is especially true for the light elements that are more affected by the attenuation of air. -Alan Koenig

Any other questions about data you’ve already collected or are you looking to start a program? We’re happy to help!