(The synopsis of this post is that I still haven’t been able to put all of the pieces of the puzzle together.)
Previously I thought that assays for the Valley of Kings deposit would be accurate. I think I’m wrong on that count. The VOK deposit seems to have a lot of ‘coarse gold’ and exhibit a very high ‘nugget effect’. This nugget effect makes the gold very difficult to measure accurately. Problems in measuring the gold could explain why Strathcona’s sample tower results differ from the preliminary bulk sample results.
The September 18 2012 technical report by Snowden states the assay and sample preparation procedures for the project (page 12):
Samples at ALS were crushed to 70% passing 2 mm, (-10 mesh). Samples were riffle split and 500 g were pulverized to 85% passing 75 µm (-200 mesh). The remaining coarse reject material was returned to Pretivm for storage in their Smithers warehouse for possible future use.
Gold was determined using fire assay on a 30 g aliquot with an atomic absorption (AA) finish.
So first the drillcore is crushed to a smaller size. Then they take a portion of that material, using a riffle splitter to make sure that the material being removed isn’t biased. Then that 500g of material is ground down to an even finer size. 30g of it is used for the fire assay.
Fire assays will not work perfectly for a high-nugget gold deposit. The problem is that coarse/nugget gold is ductile. Instead of being crushed into smaller bits, it can deform and flatten out. Because the gold isn’t pulverized down into smaller bits, it is not evenly distributed throughout the material that is sent to the assay lab. This will result in poor accuracy as some 30g samples will contain an above-average number of nuggets and others will contain less.
There may also be biases in the sample preparation stage that result in poor precision. Apparently some gold may “plate out on the mill” during grinding (see this Act Labs webpage). I understand that to mean that some of the gold will stick to the milling/grinding equipment temporarily. This means that some of the gold is lost in the milling process. This would cause the assays to report lower grades on average.
For more information, see this PDF on Unrealistic Expectations of Assay Results. You can also look on page 132 of the June 26 2013 feasibility study (PDF). It lists the results of multiple assays on the same metallurgical sample. The feasibility study notes:
There is a significant deviation in the assay results between the two laboratories and also between the assay methods. This deviation indicates a substantial nugget effect on the gold and silver assay.
Fire assays are commonly used because they are cheap and work reasonably well.
If there is nugget gold involved, a metallics (screen) assay will do a slightly better job. As I understand it (I could be wrong here), a screen or mesh sieve is used to separate the sample into a coarse portion (the grains of the sample that didn’t pass through the screen) and a fine portion. Two fire assays are performed on the coarse portion and one fire assay on the fine portion. The reason why two assays are performed on the coarse portion is because it will have most of the coarse/nugget gold. This means that it will experience the ‘nugget effect’ more, making assays less accurate. Doing a second assay will reduce the impact of the nugget effect somewhat. All of the weights are calculated to figure out the gold grade of the original sample. This type of assay is more expensive so it often isn’t used at the exploration stage. Strathcona is apparently using “metallic screen assays” to obtain its sample tower data (according to Pretium’s news release).
Gold can also be measured by taking a large sample, grinding it, and subjecting it to cyanidation. This could be an improvement as it won’t have problems with the nugget effect. However, other factors can cause problems. The ore may contain elements that disrupt the cyanidation process by reacting with the reagents. The gold may precipitate out of the solution. The ore may not be ground finely enough and therefore not all of the gold in the grains is exposed to the cyanide solution.
There are also other methods to measure gold grade. This interesting blog post describes the trials and tribulations that Orex Exploration had in measuring the gold in their deposit, going through several different assay techniques. Orex Exploration has a webpage which says that the difference between metallurgical testing and historical fire assays was 2.78X. Orex also has technical reports on its website that contain more information about its metallurgical testing.
Relevance to Pretium
As I understand it, nugget gold can be very difficult to measure with accuracy and precision. This could potentially explain the large difference between Strathcona’s sample tower results (using metallic screen assays) and the preliminary bulk sample results. The sample tower data is based on metallic screen assays. It is not that accurate of an assay technique for the high-nugget Valley of Kings zones. There may be problems with the precision of the technique (which is a problem that Orex faced with various assay techniques including metallic screen assays).
I would note that Henrik Thalenhorst has been involved with many bulk sampling programs over his long career and has written some papers on the subject. I believe that he is very familiar with the difficulties in measuring the gold accurately. Here are some examples of his work that I can find online:
- A SEC filing of a letter/report to AMEC where Thalenhorst recommends sending the field sample for gravity concentration (in a Knelson concentrator) and then using intense cyanidation on the resulting material. I presume that he is recommending that method because he is well aware of the problems of measuring gold accurately.
- He has been working as a geologist since graduation in 1968 and for Strathcona since Jan 1986 (SEC filing).
- A paper by other geologists on a bulk sampling project cites Thalenhorst’s work: “Grade verification was approached by extensive bulk sampling. This is a recommended industry practice for gold deposits with a significant nugget effect (John & Thalenhorst 1991).” The conclusion of the paper thanks “Henrik Thalenhorst, Strathcona Mineral Services Limited, Canada, for his thorough introduction to bulk sampling practice.”
- Bulk sampling of mineral projects using a sample tower: lessons from the field – Paper published in the CIM journal. (I have not read it.) Thalenhorst has several papers that have been cited by other geologists if you look on Google Scholar.
Thalenhorst has probably thought about these issues and given himself a margin of error. If Strathcona is resigning, it may be because the difference between the preliminary results at the time and the resource model’s prediction is huge. Pretium has quoted Strathcona as saying that “…there are no valid gold mineral resources for the VOK Zone, and without mineral resources there can be no mineral reserves”. This is an extreme viewpoint as mineral resources have lower criteria than mineral reserves. This implies that the resource model overstates the resource by a very large margin.
Pretium’s latest data shows that the resource model is probably underestimating the resource (by a relatively smaller margin). Pretium’s MD&A states that the ore is being processed from lowest to highest grade. If there are 1815 tonnes remaining and assuming that it has the same grade as everything processed so far (an overly conservative estimate), then the number of final ounces will be 5,160 ounces, 29% higher than the 4,000 ounces projected by the Snowden resource model. There is a huge gap here. I don’t know the explanation for it.
*Disclosure: No position in the common; long much more puts than calls. I bought some call options intending to turn them into puts but then decided not to.