The results of the 11th Reynolds Cup are announced in this video.
The 2022 Reynolds Cup was organized by Rieko Adriaens and Gilles Mertens of Qmineral Analysis & Consulting. The Clay Minerals Society and the Deutsche Ton- und Tonmineralgruppe e.V. (DTTG) are acknowledged for direct financial support. The Royal Belgian Institute of Natural Sciences is acknowledged for in-kind contributions of minerals.
The three samples were prepared as mixtures of pure minerals that should represent realistic rock assemblages. The mimicked rock types for RC2022 were a shale deposit, a lacustrine deposit and a ball clay deposit.
In total 93 labs registered for the competition of which 68 labs submitted quantitative data for each sample. All participating labs have indicated to have used X-ray powder diffraction as a primary quantification technique whereas various complementary methods were used (see figure below).
The submitted data were judged by calculating the total bias which was defined as the difference between the reported mineral concentrations and the actual mineral concentrations. The concentration of unidentified and misidentified minerals were also part of the total bias calculation. The distributions of the results of the different participating labs is shown in the figure below:
Below, the top finishers are announced.
The 4th place was taken away by the team of Kristian Ufer, Stephan Kaufhold and Reiner Dohrmann (from left to right) from Bundesanstalt für Geowissenschaften und Rohstoffe, Hannover, Germany with a bias of 77.2
Three teams achieved a joint 3th place, as their bias is very similar.
The first team is that of Stanislav Jelavic, Nathaniel Findling and Bruno Lanson (from left to right) from ISTerre – OSUG C, Grenoble, France with a total bias of 73.4.
The second joint 3th place was achieved by Mariola Kowalik, Artur Kuligiewicz, Arkadiusz Derkowski, Zuzanna Ciesielska, Pawel Ziemianski and Marek Szczerba (from left to right) of the Institute of Geological Sciences Polish Academy of Sciences, Krakow, Poland with a total bias of 71.9.
The final joint 3th place was achieved by Marek Osacký, Ľubica Puškelová, Peter Uhlík (from left to right) of the Department of Mineralogy, Petrology and Economic Geology, Comenius University, Bratislava, Slovakia with a bias of 70.4.
Peter Uhlík decribed how they did it:
“
We began by analysing the as-received samples by binocular magnifier to distinguish some specific phases (e.g. anthracite, goethite, talc). Then 1 g of each sample was mixed with 0,25 g corundum (AL-OX-03-P with nominal grain size of 3.5 microns) and grind in a McCrone mill for 5 minutes with 4 ml ethanol. Dried ground samples were shaken for 10 minutes in a plastic vial with 3 plastic balls and then other 10 minutes with small quantity of hexane. Than the samples were sieved (250 μm) and sidepack the ground material into a holder to ensure random sample orientation, and then X-rayed. For bulk rock XRD interpretation and mineral quantification, RockJock 11 software was used. Another portion of the samples (about 1.3 g) were used to separate fractions < 2 µm using Stokes’ Law. These were mounted as oriented specimens and analysed with XRD in air-dry form and after saturation in ethylene glycol.
“
The two best labs in RC2022 really stand out as their bias, below 50, was significantly lower (see also results figure above) compared to 3th place bias.
The second place was achieved by Mark Raven, Rong Fan, Rodrigo Gomez-Camacho, Nathan Webster, Nick Owen, Shu Huang, Peter Self of CSIRO Mineral Resources, Waite Campus, South Australia. They submitted data with a fantastic bias of 46.0, which in every previous Reynolds Cup contest would have more than enough to win the contest.
Mark Raven and Peter Self won the Reynolds Cup in 2010 and ended up in 2nd place in 2018 and 2020.
When asked how they did it, Mark Raven answered:
“
The as received samples were initially lightly back pressed into sample holders for XRD analysis. Semi-quantitative elemental analysis using a Spectro energy dispersive XRF was also performed on the as received samples to help confirm minor phases that may otherwise be difficult to identify by XRD alone.
Sub-samples of ~ 1.5 g of each of the samples were then micronized under ethanol with a McCrone micronizing mill then oven dried at 60° C. After drying, the micronized samples were thoroughly mixed in an agate mortar and pestle to ensure homogeneity. The fine powders were lightly back pressed to minimise preferred orientation and XRD patterns collected on a PANalytical X’Pert Pro MPD using iron filtered cobalt K alpha radiation. Patterns were collected from 3 to 80° 2 theta at 0.017° steps. Total data collection time was ~30 minutes. The process of micronizing the samples under ethanol followed by oven drying partially dehydrates any swelling clay minerals resulting in broad asymmetric 00l peaks. The micronized samples were therefore calcium saturated to restore the 001 peaks of the swelling clay minerals to ~15 Å. This was achieved by washing the micronized samples twice with 1M CaCl2, washing with deionised water followed by ethanol (centrifuged at 6000rpm after each step) before oven drying at 60° C. The Ca saturated samples were again thoroughly mixed in an agate mortar and pestle to ensure homogeneity and lightly back pressed into sample holders for XRD measurement. XRD patterns were then re-collected. Comparison of the XRD patterns before and after Ca saturation helped identify water soluble phases (i.e. celestine and gypsum) in samples RC11-1 and RC11-3 respectively. Comparison of the micronized sample before and after Ca saturation of RC11-2 showed the emergence of a sharp 14.8 Å indicating the likely presence of vermiculite, however, vermiculite was not added to the sample by the organisers. This may indicate that phlogopite may have had some exchangeable potassium interlayers.
A further 2 g sub-samples of the as received materials were dispersed with 1M NaCl and centrifuged at various speeds to separate <0.2 µm, 0.2-2 µm and >2 µm fractions. The fractions were again Ca saturated and pressed powder and oriented, calcium and magnesium saturated and glycerolated specimens were prepared to help identify the clay mineral species. Oriented Mg and glycerolated XRD patterns of the <0.2 µm fractions of RC11-1 and RC11-3 indicated interstratified species which was modelled in NEWMOD2 with regularly interstratified illite-smectite clay. Kaolin polytypes dickite and nacrite were identified by minor mis-matched peaks at ~2.32Å and ~4.14Å respectively in RC11-1. Halloysite was primarily identified in the >2µm fraction of sample RC11-3 by a broad shoulder on the low angle edge of the kaolinite peak at ~7.3Å. Dioctahedral and trioctahedral smectite clays were identified by the positions of their respective 060 peaks at ~1.499Å (montmorillonite) and ~1.514Å (hectorite) from the <0.2µm random powder samples.
Quantification was performed on all RC sub-samples using TOPAS version 6. Disordered clay minerals were calibrated from purified clay mineral standards using a modified PONKCS (Partial Or No Known Crystal Structures) procedure. Amorphous content was determined using the internal standard method by difference after the addition of 50% by weight of a fine-grained corundum standard to each micronized Ca saturated sample. Values of amorphous content less than 5% were not reported in the final compositions.
“
Finally, the winners of RC2022 are Steve Hillier and Helen Pendlowski of The James Hutton Institute, Aberdeen, Scotland. This team got a staggering and impressive bias of 35.2 (!). Steve actually ended up in the top 3 of every Reynolds Cup contest that he participated in. Steve Hillier is also the first participant to win the Reynolds Cup 3 times, after 2008 and 2018 which proves this team is nothing but rock-solid.
Steve provided a detailed explanation of their analysis strategy and procedures:
“
The first thing we did was weigh each of the samples to establish how much material we had to work with – it turned out to be just over 3.9 g of each. Then, just out of curiosity, we ran a bulk powder XRD pattern on a portion of the sample, as received. The resulting XRD pattern is no use at all for anything quantitative, but its nigh on impossible to resist the temptation to do this, as it does give first sight, albeit in some unknown biased way, of some of the minerals that have been included in each mixture. And – sometimes – the lack of control over specimen properties such as particle size and preferred orientation does serve to emphasis the presence of a minor phase that may not be as obviously identified when a more rigorous sample preparation approach is applied. (Whatever you do, don’t attempt a quantitative analysis of a powder sample when you haven’t prepared it in any way and therefore don’t know anything about the particle properties of the specimen as ‘seen’ by the X-ray beam).
Next, we carefully split the three samples into portions, one (≈ 2.5 g) to be used for the preparation of a random powder specimen by McCrone milling in ethanol for 12 minutes (ensures an appropriate particle size distribution) followed by spray drying of the resultant slurry, directly from the mill (ensures a random powder specimen). The resultant powder specimens are shown in Figure 1.
FIGURE 1. Spray dried powders of the three RC11 samples.
Of the remaining material for each sample ≈ 0.3 g was used to disperse a clay size (< 2µm e.s.d) fraction, which was collected by sedimentation and prepared as an oriented (textured) specimen on a glass slide using a filter peel transfer procedure. Also, specimens of ≈ 0.5 g were sent to an independent commercial laboratory for chemical analysis by wavelength dispersive XRF using fused glass beads. So we had ≈ 0.6 g of each sample held in reserve, should we decide we needed to conduct any other tests.
The clay size fraction data were used primarily for the purpose of confirming the detailed identification of the clay minerals in the samples. The deliberate texture induced in a clay specimen collected on a filter then transferred to a glass slide substrate, serves to bias the diffraction pattern in favour of the basal reflections from clay minerals. Three XRD patterns were recorded for each sample, one as received ‘Air Dried’, a second (using the same specimen) following exposure to ethylene glycol ‘Glycolated’ overnight by vapour pressure, and a third (again using the same specimen) following heating on a hotplate at 300°C for one hour, ‘Heated’. This combination of treatments and the changes that occur between the three diffraction patterns in relation to basal reflections from the clay minerals when they are compared by overlaying the patterns, usually provides enough information to identify the various clay minerals in a sample to a reasonable level of detail. Provided that is that they are present in the < 2µm size fraction; and there are sometimes reasons why they might not such as failure to disperse, or larger particle size (clay minerals are not always of clay size!).
In sample RC11-1, the clay fraction data indicated the presence of kaolin, along with some illitic clay and two types of expandable clay minerals, which looked to be a mixed-layer illite/smectite and a smectite; in sample RC11-2 we could see a large amount of smectite, which based on the relative intensities of the 002 and 003 peaks looked like it was probably trioctahedral smectite (basal intensity ratios reflect chemical compositions which are correlated with octahedral occupancy), along with some talc, and traces of chlorite and serpentine; whilst in RC11-3 the clay fraction indicated the presence of kaolin (the peak shape of which suggested the probable presence of halloysite), some illitic/micaceous phase(s), and again, like sample RC11-1, two expandable phases, one a mixed-layer illite/smectite the other a smectite.
The information gained from the clay fraction analyses helped to inform the choice of standards that were included in our procedure for quantitative analysis, which relies entirely on the information in the diffraction patterns obtained from the random powder specimens. As we have done in previous Reynolds Cups, we used a method based on full pattern fitting of ‘prior determined’ standard patterns for every mineral identified in the samples. We use the term ‘prior determined’ as the standard reference patterns can be measured or (pre) calculated ones. Identification of minerals was made using standard search match procedures and reference to patterns in the ICDD PDF 4+ 2022 database, but also to tabulated data in the monograph by Brindley and Brown 1984 (pages 349-356). Additionally, identification is aided by the full pattern fitting method employed because it highlights collections of peaks and other regions where peak intensity is missing entirely or not adequately modelled in initial fitting attempts. Such peaks and regions can then be isolated in the powder pattern using modern search match software and used for more targeted search match procedures in order to identify minerals that may have escaped initial identification.
Our method of full pattern fitting has been incorporated into an opensource R package known as ‘powdR’ and all quantitative analyses were done using this package (Butler and Hillier 2021a, 2021b). One of the prime advantages of powdR is that it is much faster than our previous implementation of our methods in Excel. However, perhaps the biggest advantage of our prior determined full pattern fitting method compared to full pattern fitting with the Rietveld method is that the prior determined patterns include their associated background scattering. As such, if an amorphous phase is present, it becomes obvious that an appropriate pattern for it must be added to the fit and it is then quantified in exactly the same way as all other phases based on full pattern reference intensity ratios. There is no need to dilute the sample with an added internal standard to quantify amorphous phases, as must be done with the Rietveld method, but you do need to obtain appropriate reference patterns for the amorphous phases present as different amorphous phases have different patterns. Thus, we were quickly able to determine that RC11-1 and RC11-3 required the addition of amorphous phases, whereas RC11-2 did not.
One of the challenges of RC11-1 was the presence of three kaolin polytypes in the same sample, namely kaolinite, dickite and nacrite. The full pattern fitting quickly revealed the presence of all three polytypes, but unfortunately our reference library did not have a nacrite pattern! The quickest way to add one was to calculate a diffraction pattern for nacrite to which an appropriately scaled background, borrowed from a well crystallised kaolinite, was added. Figure 2 shows measured and fitted patterns for RC11-1, the fitted being the summation of the scaled standard patterns shown in the figure.
FIGURE 2. Comparison of measured and fitted XRD patterns for sample RC11-1. Note figure is zoomed into detail so peaks of high intensity are truncated in the figure. The reference patterns that sum to give the fitted pattern are also shown.
For sample RC11-2, the wide range of exclusively trioctahedral clays and phyllosilicates, including trioctahedral smectite, talc, serpentine, and chlorite will have probably posed quite a challenge for Rietveld based methods due to commonalties amongst their structures. However, the pattern fitting identified and quantified all four clay components of RC11-2 with a cumulative bias of just 2.9%.
FIGURE 3. Comparison of measured and fitted XRD patterns for sample RC11-2. Note figure is zoomed into detail so peaks of high intensity are truncated in the figure. The reference patterns that sum to give the fitted pattern are also shown.
For RC3-11, one of the most pleasing aspects of the fitting was the accurate determinations of both kaolinite and halloysite; pleasing, because this task still proves to be a difficult area for industry when both are present in the same sample.
FIGURE 4. Comparison of measured and fitted XRD patterns for sample RC11-3. Note figure is zoomed into detail so peaks of high intensity are truncated in the figure. The reference patterns that sum to give the fitted pattern are also shown.
Once the final fits were established for each of the three samples it was time to cross check the mineralogy against the measured geochemistry, recalling that any quantitative mineralogical analysis of an unknown sample must be compatible with its bulk chemical composition. This is about the only balance check you can make when you work with samples from real soils, sediments and rocks, and it is an extremely useful thing to do.
FIGURE 5. Comparison of measured and calculated sample chemical compositions. Elements represented as oxides measured by XRF are compared with oxide values calculated form the mineralogical analyses which gives the weight fraction of each mineral and its assumed chemical composition. Note axes are log scale
This comparison is shown in Figure 5 with the axes on a log scale due the wide range of concentrations of the various elements expressed as oxides. The fit with most points sitting close to a 1:1 line gave added confidence that the mineralogical composition determined from the full pattern fitting should be satisfactory. Fe and Mg can be tricky because their calculated values are affected by the Fe/Mg ratios assumed for any minerals that contain them. When dealing with real samples combining mineralogical and chemical data presents some interesting opportunities to optimise mineral chemical compositions and thereby learn more about the specific minerals present in your samples.
All that remained was to submit the mineralogical results to the organisers of the 11th Reynolds Cup. When the true mineralogical compositions of the samples were revealed the final total bias of the above three fits was 35.2. Several minerals that were not present were incorrectly identified in samples RC11-1 and RC11-3, but only at trace levels. In total these misidentified minerals contributed 3.5 and 2.1, respectively to the total bias. The inclusion of these minerals in our submitted results illustrates the difficulties of reliably identifying minerals present at trace levels in complex mixtures like those of the biennial Reynolds Cup competitions.
Butler, B. M., Hillier, S., 2021. powdR: An R package for quantitative mineralogy using full pattern summation of X-ray powder diffraction data. Comp. Geo. 147, 104662. https://doi.org/10.1016/j.cageo.2020.104662
Butler, B., Hillier, S. 2021. Automated full-pattern summation of x-ray powder diffraction data for high-throughput quantification of clay-bearing mixtures. Clays Clay Miner. 69, 38–51 (2021). https://doi.org/10.1007/s42860-020-00105-6
“