Sample Size and Meaningful Gold Analysis

It is over 30 years since the publication of the Clifton et al paper “Sample size and meaningful gold analysis”. However, confusion still exists among the protagonists of various types of sampling such as pan concentrate, -80# and BCL (bulk cyanide leach).

This communication is an informal overview of some sampling methods, with simple numerical examples applied from the Clifton et al nomograms, which are also reproduced in the Field Geologists’ Manual.

Perhaps that rare and diminishing breed of regional explorationists who still collect samples in the field will find this a useful, practical check for best practice. Those who assess open file data might also benefit through questioning some data validity and applicability, and perhaps carrying out field sampling checks with these common sense principles in mind.

12 September, 2002


Nick Marshall, consultant

Marshall Geoscience Services Pty Ltd

Sample Size and Meaningful Gold Analysis


This note gives some numerical scenarios to illustrate the severity of nugget effects. Please refer to the two graphs in the Field Geologists Manual, 2nd Edition 1982, p97-98. These are reproduced from the USGS Professional Paper 625-C by Clifton et al, 1969, “Sample size and meaningful gold analysis”.


1. Assume -80# samples (-177 microns).


Clifton et als’ sampling statistics considered a binomial distribution for particulate gold. To achieve ±50% precision 95% of the time (ie 20 particles of flake gold of -80 mesh required) on a 50g sample taken for analysis (eg by fire assay), the results will only be reliable at and above 3 ppm (3000 ppb). So much for this method of sampling and analysis! One can only get away with this, and have reproducible results, if the gold flakes within the -80# sediment are very much finer than 80#, as they mostly are. For example if the gold was only 30 micron sized (80# is 177 microns), then 50 g would be a reliable analytical sample at and above 16 ppb (graph page 98).


If visible gold can be panned, however, by definition nugget effects can be guaranteed some of the time, as such gold is relatively coarse.


Crustal abundance for gold is approximately 2 ppb. A 50g sample running 2 ppb (±50%) consistently, implies that the gold is as 8 micron spheres, or 15 micron flakes (graph, p98). This may be a barren, background result, indicating 20 grains of this size, or >20 grains of a smaller size none of which is pannable or visible. But suppose that within this spectrum of grain sizes we have one flake of coarser, 125 micron gold per kilogram which would report as perhaps 5 flakes of “VFG” in a typical pan concentrate. This might be sourced from a small, distal auriferous quartz vein.


Referring to the graph on p97, this flake would weigh approximately 3.5 micrograms. On a unit basis, this equals 0.0035 micrograms per kilogram, or 3.5 parts per billion. Added to the 2 ppb median, the result should be 5.5 ppb. However, a 50g sub-sample taken for aqua regia digest would include this 125 micron flake one time in twenty, reporting “72 ppb” as a strong spot anomaly.


Follow up of such spot anomalies consumes time and money.


NB. One should look at spatially clustered gold anomalous areas rather than following up isolated extreme highs, provided sampling density is adequate. The anomalous site should also make geological sense does it drain structures or formations/intrusions of potential interest? (check float lithologies in field log).


Reanalysis of 50g of the same sample would produce a value of only 2 ppb, which should alert one to the problem. Sieving to -200# or even -150# would have eliminated the random flake and given a reproducible 2 ppb result. In the 50g of -200# (75 microns) case, results above approximately 250 ppb would always be reliable (graph, p98). The same aqua regia digest could also be used to determine arsenic, bismuth and chalcophile elements.


BCL (or BLEG) sampling attempts to solve this problem by taking a large, more representative sample, say 2 kg, for analysis. Such a sample can be treated with an alkaline cyanide leach (the principle of CIP and CIL extraction) at room temperature, whereas fire assay or hot aqua regia digest on such a large sample would be difficult. In the above example, the one 3.5 microgram flake in a 2 kg sample would add 1.75 ppb to the median value.



Assume 2 kg of -40# BLEG.


In practice there are many variations, taking sample weights from several hundred grams to up to 10 kg, with sieve sizes ranging from fine to quite coarse, of the order of 10 or even 5 mesh. Some practitioners add Magnafloc to coagulate suspended clays, while analytical variations include using a static leach, where the sample is not agitated continuously. This would promote considerable scope for analytical error.


Methods of concentration of gold from the cyanide solution for final presentation to flameless AAS or ICP or ICP-MS include precipitation with zinc dust, coprecititation with tellurium on reduction, adsorbtion onto activated charcoal, or solvent extraction. Silver, palladium and copper can be determined from the same cyanide leach, although in my experience correlation of copper with conventional acid leach copper is not good.


Assuming 2 kg of -40# sample, a 4 ppb result would be reliable (± 50%) if all the gold within the -40# (-475 microns) fraction was present as 20 grains of 65 micron gold, or over 20 grains of finer gold. ( see graph in Field Geologists Manual, p 98). 200# is 75 microns = 0.075mm.


But, if we had two flakes of 0.22mm gold within this 2 kg sample, reference to graph p 97 shows this would add 16 ppb to the sample to make it 4 + 16 = 20 ppb instead of 4 ppb.


The answer is to eliminate random, coarser gold by fine sieving but still taking a reasonable bulk for BCL analysis. There is, however, a practical trade off in how long it takes to sieve a lot of fine sample, particularly from coarse soils or sediments, or high energy streams.


3. -150# (-105 micron) BCL.


Consider 250 g BCL on a -105 micron fraction.


This also is a compromise: 70 ppb is a reliable threshold if all gold is present as 20 flakes of this size.


A reproducible 4 ppb result implies that all gold is as 30 micron flakes or finer, which is reasonable.


One flake of approx. 90 micron gold (equivalent to 4 flakes in 1 kg; see graph, p 97) would add 4 ppb to the result. ie) result is 8 ppb instead of 4 ppb true.


While far from perfect, this is a better scenario than the previously discussed alternatives, and statistically there is less chance of one random flake in 250 g than in <8 random flakes in 2 kg of -40#.


Moreover, the sample is more constant (fine silt and clay only), than a varying mixture, from sandy site to silty site, of sand/silt/clay, due to variations in hydraulic load.


4. Stream Magnetic Concentrates.


Although only a few grams of relatively coarse sample are collected “as is” from the stream bed, (Marshall, 1995) results are usually fairly consistent and reproducible. There is some evidence, (subject to further case histories) that alluvial gold is not detected.


This is logical considering the geochemical (chemical rather than mechanical) dispersion mechanism involved. Here one is analyzing “molecular gold” which is coprecipitated and/or adsorbed from solution, along with Cu, Pb, Zn, Ba, Mo, As etc on to amorphous iron oxide, some of which becomes crystallized to maghemite, which is then readily extracted, or reconstituted from the bulk of diluent sand and clay grains. (Goethite is even more effective, but is not as readily concentrated in the field. In any case, goethite is probably included with maghemite as composites.) Hence the relative absence of “nugget effect” using this sampling medium. Alluvial gold is particulate, and in the absence of an oxidizing sulfide or solubilizing humic/fulvic acid or cyanogenic phase, there is no mechanism for its dissolution and subsequent coprecipitation and/or adsorbtion with FeOx on redox and pH change in the weathering environment.


5. Panned Concentrates


This traditional technique is very sensitive to alluvial gold and “noise” sources such as minor auriferous quartz sweat-veins of insignificant tonnage potential. Very fine grained gold from Carlin type and some epithermal type deposits may go undetected.


Results can be improved by wet sieving to a coarse fraction of constant volume prior to panning, and sampling from a consistent hydraulic load. High amounts of magnetite are removed if the concentrate is too bulky for fire assay. Unless the magnetic fraction being removed consists of pure magnetite, however, there is the danger of also removing occluded gold (Bheemalingeswara, 1995). Due to the relatively small analytical sample size, of the order of 30g or less for fire assay, and relatively coarse visible gold grain size, non-reproducible results are virtually guaranteed except where concentrations are high.


Day and Fletcher (1986) and Saxby and Fletcher (1986) discuss the effects of particle size and abundance of heavy minerals in relation to hydraulic load. Their field examples are a dramatic illustration of the problems of non reproducibility.


I regard this technique as qualitative only, for widely spaced regional sampling. Microscopic study of gold grain morphologies can be useful for suggesting different multiple sources of local versus distal origin.




Berkman, D. A., 1982. Field Geologists’ Manual. Aust. Inst. Min. and Metall.

Bheemalingeswara, K., 1995. Possible effects of iron oxide coating in the recovery of particulate gold from stream sediments. J. Geochem. Explor., 52, 373-380.

Clifton, H. E., Hunter, R. E., Swanson, F. G. and Phillips, R. L.., 1969. Sample size and meaningful gold analysis. U. S. Geological Survey Professional Paper, 625-C.

Day, S and Fletcher, K., 1986. Particle size and abundance of gold in selected stream sediments, southern British Columbia, Canada. J. Geochem. Explor., 26: 203-214.

Marshall, Nick, 1995. Magnetic concentrates a proven sampling medium for oxidized arid landscapes. 17th International Geochemical Exploration Symposium, Extended Abstracts, p95-98. EGRU Contribution 54. James Cook University, Townsville, Australia.

Saxby. D. and Fletcher, K., 1986. The geometric mean concentration ratio (GMCR) as an estimator of hydraulic effects in geochemical data for elements dispersed as heavy minerals. J. Geochem. Explor., 26: 223-230.