A Brief History of the Proficiency Testing Scheme
This quarter’s newsletter continues on the theme of history lessons – this time with some background to the Proficiency Testing Scheme.
Review of the 11 Years of Gold Proficiency Tests, by Mike Hinds
Mike Hinds (Consultant, LBMA)
The LBMA Proficiency Test Assay Program started in 2012 and has been an annual assay test for the past 11 years. This has been available to Good Delivery List (GDL) laboratories since the start and open to non-GDL laboratories since 2016.
This program has provided assay laboratories with the ability to determine gold and silver in the range of 995.0 – 999.9‰ and feedback on their results anonymously (each laboratory identified by a code). The collating of results and statistical analysis is done by Fapas (Food Analysis Performance Assessment Scheme) under contract to LBMA.
The gold and silver test samples are provided by the referee laboratories to LBMA. These samples are rigorously analysed by both fire assay and Inductively Coupled Plasma – Optical Emission Spectrometry (ICP-OES) methods and are checked to ensure the metal samples are homogeneous.
An overview of 11 years of gold assay data from GDL laboratories is presented in Table 1. The measure of a good proficiency test result is given by a Z-Score which is within ± 2. The goal of a laboratory is to have a Z-Score as close to 0 as possible, which indicates that the value obtained by the laboratory matches the assigned value for the proficiency test sample. The Z-Score is defined as:
Where x = the laboratory’s reported result Xa = the assigned gold concentration value αp = the standard deviation from all the test proficiency test results.
The data in Table 1 shows the year, number of GDL laboratories participating, the number of laboratories with a Z-Score between -2 and 2, % laboratories with a Z-Score within ± 2, and the assigned Au concentration in fineness (‰). For the majority of the proficiency tests conducted, more than 85% of the GDL laboratories or better had a Z-Score within ± 2 and frequently this percentage is above 90%. The only exception to this was in 2017, when it’s likely there were unforeseen homogeneity issues with the test samples.
A comparison of standard deviations from the two methods is revealing, as shown in Figure 1. At lower concentrations of gold (995 - 996‰), the standard deviations from both methods tend to be similar and somewhat large. This is most likely due to slight variations of impurity element concentrations (Ag, Cu, Pd, Pt, etc) in the gold. The higher standard deviations observed are a combination of the variability of the material and the variability of the analytical method. At higher concentrations of gold, the trend line indicates that the standard deviation of the ICP-OES determinations tend to be lower than the standard deviation of fire assay determinations. The impurity element concentrations are lower and therefore the variations of these elements would be lower as well. The standard deviation from fire assay determinations at higher concentrations of gold tend to become constant. This suggests that the standard deviations observed is more due to the intrinsic reproducibility of the fire assay method rather the variability of the material.
Results from non-GDL laboratories identified since 2016 are summarised in Table 3. There are fewer numbers of non-GDL laboratories participating in this proficiency test compared to GDL laboratories, and few of the same laboratories participate annually. The low number of laboratories reporting results that have Z-scores within ± 2 suggest that these laboratories may need to improve their methods of determining gold at these concentration levels.
This voluntary Proficiency Test Program has provided participating laboratories with valuable feedback on their ability to assay high purity Au (995 – 999.9‰) for many years. It is hoped that the high number of laboratories participating in this program will continue.
Table 1: Summary of Results of GDL Laboratories Determining Au in Proficiency Tests from 2012 – 2022.
| Year | Total # Labs | # Labs Within Z-Score ± 2 | % Labs Within Z-Score ± 2 | Au, ‰ |
|---|---|---|---|---|
| 2012 | 40 | 34 | 85 | 995.34 |
| 2013 | 40 | 38 | 95 | 998.34 |
| 2014 | 41 | 36 | 88 | 996.77 |
| 2015 | 49 | 46 | 94 | 999.67 |
| 2016 | 48 | 41 | 85 | 996.28 |
| 2017* | 48 | 30 | 63 | 994.67 |
| 2018** | 49 | 47 | 96 | 999.72 |
| 2019 | 42 | 40 | 95 | 999.53 |
| 2020 | 42 | 41 | 98 | 995.09 |
| 2021 | 37 | 35 | 95 | 998.43 |
| 2022 | 42 | 37 | 88 | 995.07 |
*2017 Au results suspect.
**2018 GDL and non-GDL laboratory results not separated.
Table 2: Comparison of Methods Bias and Reproducibility (standard deviation) from fire assay (FA) and ICP-OES for the determination of gold in proficiency test samples.
| Year | Au, ‰ | #FA | FA Bias | FA Std Dev | #ICP-OES | ICP Bias | ICP Std Dev |
|---|---|---|---|---|---|---|---|
| 2012 | 995.34 | 33 | 7 | ||||
| 2013 | 998.34 | 30 | 10 | ||||
| 2014 | 996.77 | 24 | 12 | ||||
| 2015 | 999.67 | 26 | 0.00 | 0.047 | 21 | -0.01 | 0.012 |
| 2016 | 996.28 | 34 | 0.00 | 0.049 | 13 | 0.01 | 0.040 |
| 2017* | 994.67 | 35 | 11 | ||||
| 2018 | 999.72 | 25 | 0.01 | 0.038 | 22 | -0.01 | 0.014 |
| 2019 | 999.53 | 24 | 0.00 | 0.037 | 17 | 0.00 | 0.015 |
| 2020 | 995.09 | 32 | 0.00 | 0.053 | 9 | -0.02 | 0.044 |
| 2021 | 998.43 | 22 | 0.00 | 0.041 | 12 | 0.00 | 0.030 |
| 2022 | 995.07 | 28 | 0.01 | 0.058 | 11 | 0.01 | 0.074 |
*2017 Au results were suspect and so excluded.
Table 3: Summary of Results of non-GDL Laboratories Determining Au in Proficiency Tests from 2016 – 2022.
| Year | Total # Labs | % Labs Within Z-Score ± 2 | % Labs Within Z-Score ± 2 | Au, ‰ |
|---|---|---|---|---|
| 2016 | 14 | 9 | 64 | 996.28 |
| 2017* | 4 | 2 | 50 | 994.67 |
| 2018** | 999.72 | |||
| 2019 | 7 | 5 | 71 | 999.53 |
| 2020 | 9 | 5 | 56 | 995.09 |
| 2021 | 8 | 3 | 38 | 998.43 |
| 2022 | 10 | 6 | 60 | 995.07 |
*2017 Au results suspect.
**2018 GDL and non-GDL laboratory results not separated.
Figure 1. The standard deviations from fire assay (blue) and ICP-OES (red) are correlated with the Au concentration of each proficiency test (2017 excluded).
Looking Ahead to the 2023 Gold and Silver Proficiency Testing Scheme
The GDL team would like to take the opportunity to thank all those who participated in the survey, the feedback has been reviewed and the following changes have been made to the schedule:
- January: Dispatch samples, avoiding the December holidays.
- Additional time has been factored in to prepare the Fapas report for timely publication in April.
Registrations for the 2023 Gold and Silver Proficiency Testing Scheme will open towards the end of September.