Chapter 11

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

HOME

Chapter 11

Sampling, Analysis
and Quality Control

11.1. ERRORS IN SAMPLE important topic of sampling covered elsewhere in


PREPARATION AND ANALYSIS this monograph and should be mindful of the fact
that, of the numerous sources of error associated
Geologists should be aware of the many factors with the analysis of geological materials, the
that can adversely affect the quality of analytical initial field sampling is widely accepted as being
results reported by laboratories and should the greatest source of error (James and Radford,
strive to minimise, if not eliminate, these by 1980).
implementation of rigorous quality control
programs and consultation with laboratory
personnel. When implemented, a traceable quality Sample preparation errors
assurance/quality control (QA/QC) program The intent of sample preparation is the reduction
provides accountability, reduces uncertainty in particle size, through crushing and pulverising,
in ore reserve calculations and adds value to and in sample size, through splitting, while
projects. retaining representivity. Samples as large as 10 -
There are errors inherent in all three stages 30 kg are frequently presented to the laboratory
of the data acquisition process, from the initial for preparation and the final analysis is of a
in situ field sampling through to laboratory portion as little as 0.2 g. The expectation is that
sample preparation and analysis. Gy’s formula the result obtained on this 0.2 g aliquot is, within
has led to a greater understanding of the various acceptable limits, representative of the original
sources and magnitude of these errors and has field sample.
been discussed in detail by Gy (1992) and Pitard
(1993). Drying
All samples are dried as the first stage in sample
Sampling errors preparation at temperatures of 105 or 110°C
(note: quarantine regulations stipulate that
Sampling errors occur during the in situ (field)
some samples originating from overseas be dried
sampling of media (soil, sediment, outcrop, drill
at 160°C for the purpose of sterilisation; these
core, reverse circulation cuttings or vegetation). regulations can vary from one jurisdiction to
The intention of sampling is to obtain a sample another). Drying is generally for a minimum of
compositionally representative of a finite volume six hours or until constant mass is achieved. For
of the surrounding sample medium. The errors sulfide-rich samples, drying temperature may be
associated with sampling will vary depending on reduced to 65 - 75°C to minimise oxidation. This
the distribution of the metal of interest and are precaution is critical for mineralised samples
generally higher in precious metal ores (nuggety where pyrrhotite is the major sulfide as this
high-grade ores in particular) compared with mineral is susceptible to auto-combustion when
base metal ores. The reader is referred to the heated in a drying oven or even during grinding.

Field Geologists’ Manual 311

Chapter11.indd 311 27/07/2011 5:54:14 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Where hygroscopic moisture is not completely Pulverising


removed assay results will bias low by an amount
equivalent to the weight per cent residual The grind should be at least 85 per cent passing
moisture. 75 microns to maximise recovery by the chosen
method and to achieve a homogeneous pulp
In the case of heat sensitive materials, such as for subsampling. Appropriate grind size will
coal and sulfides of mercury, arsenic, antimony depend on the characteristics of the material
and bismuth, care should be exercised to avoid being ground, but is particularly important
overheating during drying as this may lead where the metal(s) of interest are present in low
to volatile losses and changes in chemical concentration and comprise a major component
composition and speciation. Gypsum can also of a trace mineral, eg zircon in a basalt. Sample
lose crystalline moisture at temperatures as low size will dictate bowl size and where various sized
as 90°C. samples are submitted they may be pulverised
in different bowls. This has implications for the
Crushing monitoring of contamination with coarse blanks.
Cross contamination from material adhering to
It is important that crushing and splitting the grinding bowl can be minimised by inclusion
equipment is cleaned properly between each of a quartz (or other barren material) wash or
sample to minimise cross contamination. Losses alcohol wipe between samples if air blowing is
of fines should be kept to a minimum as these insufficient. For iron ore samples a quartz wash
may be compositionally distinct from the coarse is not recommended as this will contaminate the
fraction and bias the sample. Samples and rejects sample more than an iron ore sample due to the
can be weighed at each stage of the size reduction chemical contrast. For gold analysis the sampling
process to monitor these losses. Crushing to at error depends on the number and size of gold
least 80 per cent (preferably 90 per cent) minus particles in the aliquot for analysis. Gold, being
2 mm is desirable to minimise sampling error malleable, may not undergo much comminution
during splitting. so fine grinding may not improve the sampling
error.
Splitting
Crusher and mill contamination
The particle size of the crushed sample should be Wear in the crushing and grinding media
fine enough to permit splitting of a representative (bowl, puck and rings) is another source of
sample for pulverising. Splitting is generally by contamination. Mill contamination depends on
riffle, cone or rotary divider. Where sample size is the type of mill being used. See Table 11.1.1.
less than 3 kg a ‘total sample prep’ is sometimes
employed. This involves bypassing the splitting Subsampling for analysis
stage and pulverisation of the entire sample in a A portion (generally 100 - 200 g) of the bulk
larger capacity bowl. Smee and Stanley (2005) pulp is transferred to a pulp packet from which
consider this inferior to splitting off a 1 kg or the assay portion is taken for analysis. The
smaller portion for grinding as the particle size amount analysed may be 25 - 50 g for fire assay
reduction (comminution) of a 3 kg pulverised of precious metal ores to as little as 0.2 - 1 g for
sample can be substandard. an acid digestion of base metal ores. Steps should

TABLE 11.1.1

Bowl type Sample capacity Main contaminant Minor contaminant


Standard mild steel bowl to 3.5 kg Fe, Cr Mo
Cr-free steel 500 g to 3 kg Fe Mn
Zirconia 100 g Zr, Hf Al
Tungsten carbide 150 g W, Co (Ta)
Agate 100 g Si

Source: SGS, 2011.

312 Field Geologists’ Manual

Chapter11.indd 312 27/07/2011 5:54:14 PM


CHAPTER 11 – SAMPLING, analysis and quality control

be taken to ensure that the pulp is homogeneous Inappropriate method selection


before taking an aliquot for analysis to keep the
subsampling error within acceptable limits. If The choice of analytical method should be made
a subset of pulp samples are submitted to an in close consultation with laboratory personnel.
external check laboratory for analysis the check For many sulfide-hosted base metal ores an aqua
laboratory should be instructed to re-homogenise regia digest may give total recoveries, but silicate-
the pulps prior to subsampling for analysis. This hosted nickel, for example, will require four acid
step overcomes possible segregation that can (where hydrofluoric acid is added to dissolve
occur during transport and can be achieved by silicates) or fusion treatment to report the total
mat rolling or re-pulverising. nickel. Fusion methods are required for metals
(such as Sn, W, rare earth elements (REE), Ti,
Mislabelling V) that may be hosted in refractory minerals
Many laboratories use bar coding to eliminate and for whole-rock major element oxides. As a
transcription errors in sample identification. general rule, the more aggressive the digestion
Labelling errors generally occur at the point (aqua regia four acid lithium borate/
of sample collection and tagging or at the sodium peroxide) the higher the lower limit of
laboratory, when the initial batch/job set up is detection (LLD) and the lower the sensitivity.
done. The submission of sample lists in electronic Ideally the LLD should be at least 20 times below
format can help reduce labelling errors in cases the expected concentration level. Where this is
where samples are not submitted with bar codes unknown a good rule of thumb for geochemical
already in place. Related sample preparation prospecting is to have the detection limit ten per
errors include unintentional mixing of fractions cent of crustal abundance (Clark value). For gold,
of different samples. The more complex the lead collection fire assay and neutron activation
analytical scheme (eg fire assay compared with analysis will give total concentrations, whereas
acid digest) the greater the chance of mislabelling recoveries by aqua regia and cyanide leach may
in the laboratory. be compromised by the presence of fine-grained
gold included in silicates and sulfides (cyanide
Analytical errors leach).

Analytical errors, also referred to as measurement Geochemical versus ore grade methods
errors, are sources of error originating from the
point of weighing of the analytical subsample, For soil and stream sediment surveys and
through the digestion process and instrumental exploration drilling, geochemical or low level
reading, to the final reporting of results. methods are generally suitable. Lower limits of
detection are well below the lower concentration
Drying and weighing range of most geological materials and upper
limits for base metals are around one per cent.
For highly hygroscopic materials, such as some
An accuracy of ±10 per cent for geochemical
nickel laterite ores and iron ores, a further drying methods is usually quoted by laboratories and is
stage is required immediately prior to weighing considered ‘fit for purpose’ in these applications.
the portion to be analysed. The sample should be For resource work accuracy is critical and ore
dried at 105°C until constant weight is achieved grade methods are employed. These methods use
then transferred to a desiccator, containing fresh volumetric glassware and superior QC and an
desiccant, to allow to cool to ambient temperature accuracy of ±5 per cent is generally quoted. Many
prior to weighing. Alternatively, the sample can laboratories offer the convenience of analysing
be left to obtain hygroscopic equilibrium in the variably mineralised samples by geochemical
laboratory atmosphere and a separate sample methods and automatically reanalysing samples
taken for moisture determination. The results above the upper limit of detection by an ore
are then corrected to a dry basis based on the grade method. This protocol can be set up in
moisture content. It should be stressed that consultation with the laboratory manager. For
incomplete drying or inadequate correction for commodities such as copper, where cut off grades
moisture can be a major source of error in the are less than one per cent, the laboratory can be
analysis of hygroscopic materials such as goethite instructed to reduce the changeover to 0.5 per
iron ores and laterites. cent or less to ensure the quality of results.

Field Geologists’ Manual 313

Chapter11.indd 313 27/07/2011 5:54:15 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Precision errors are dependent on analytical ambient temperature or warming up of the


method and concentration and increase electronic components of the instrument during
dramatically as the lower limit of detection is operation. Progressive blocking of nebulisers
approached. For example, precision of 100 per or atomisation sources are a physical cause of
cent is acceptable at the detection limit for ore drift. If unrecognised, the results will be biased.
grade digests. Internal standards and normalisation standards
can be used to minimise the effects of drift
Voluming (Bloom, 2002).
Following acid digestion or fusion the sample is Noise is short-term instability of the
dissolved by leaching with HCl and then made instrument and hence signal. It can be the result
up to constant volume (eg 100 ml) by dilution of variation in atomisation and nebulisation
with distilled water. Additional dilutions may conditions or due to the electronic components.
be required before the solution is then drawn The noise of the instrument is a major factor
into the reading instrument for analysis. If the in governing detection limits, as the noisier the
glassware and pipettes are poorly calibrated, instrument, the higher the detection limit. This
non-volumetric or the operator careless, then parameter can be used to monitor degradation of
dilution errors will result. For example, voluming instruments or components.
to 101 ml instead of 100 ml can produce a one per
cent negative bias in the result. Memory effect
If analyte concentrations in successive samples
Instrument calibration fluctuate widely (over several orders of
Poor instrument calibration will generate bias in magnitude) the lines (tubing) of the reading
every sample of an analytical batch. This run bias instrument can become contaminated. The
is a large source of systematic error in analytical sample, or samples, immediately following the
chemistry. Ensuring that the uncertainty on the sample with elevated concentration will be
calibration is small will minimise this source of biased high due to this carry over effect unless
error. Instruments are calibrated using standard the lines are thoroughly flushed by the operator.
solutions prepared from spectrographically pure Pulp blanks can be used to monitor this potential
chemicals. The levels of analytes in the calibration problem. Gold is ‘sticky’ in peristaltic tubing
solutions should be in the appropriate range and thus prone to carryover. This is a very good
and similar to the samples to be measured. The reason not to submit ore grade gold CRMs in
solutions are made up at regular intervals and trace gold jobs.
at different grades appropriate to a particular
method. Once depleted, new solutions are Inter-element interferences
prepared. Provided these are made up with the
Due to the presence of constituents in the
proper pipettes and using excellent standard
sample at concentrations sufficiently elevated to
solutions the uncertainty on the final solution
interfere with the measurement of an analyte of
concentrations should be small in comparison
interest. The effect of the interference may lead
with the calibration uncertainty.
to either positive or negative bias in the analyte
All laboratories exhibit a degree of run-to-run and may be chemical or instrumental in nature.
bias. It is the responsibility of the laboratory
operator to ensure, through internal QC Crucible contamination
protocols, that this bias is within acceptable
limits. The geologist should also monitor this by There is an increasing tendency for commercial
insertion of appropriate reference materials. laboratories to offer analysis of base metal sulfide
ores by lithium borate fusion X-ray fluorescence
Instrument stability (XRF). The platinum crucibles used in the fusion
can become contaminated by alloying of metals
Can be adversely affected by drift or noise. Drift
over time and this degradation can be a source of
occurs when the instrumental response to a
given concentration of analyte or to background cross contamination.
conditions varies with time (Bloom, 2002). Nickel, zirconium and ceramic crucibles are
The drift may not be linear or progressive used for sodium peroxide fusions. These fusions
and is frequently the result of a change in cannot be used for the determination of:

314 Field Geologists’ Manual

Chapter11.indd 314 27/07/2011 5:54:15 PM


CHAPTER 11 – SAMPLING, analysis and quality control

•• sodium; dispersion), precision refers to tightness of results


•• the metal from which the crucible is (without regard to centring), and accuracy covers
both (on centre and tightly together) (Hauck et
manufactured and other metals, which may
be present in the crucible in significant al, 2008). Figure 11.1.1 illustrates the meaning of
amounts; or these terms.
•• elements subject to spectral interference by Figure 11.1.1
these elements. Schematic diagram illustrating the relationship
For example for zirconium crucibles zirconium, between trueness, precision and accuracy (from
hafnium and phosphorus cannot be determined, Flynn and Holdsworth, 2009).
(and rare earth elements are not advisable)
whereas for nickel crucibles nickel, cobalt and
molybdenum cannot be determined and the
analysis of chromium, copper, manganese and
zinc is not recommended (Genalysis, 2011).

Dilution errors
When analyte concentrations exceed the upper
limits for particular methods they can be brought
into range by dilution of the digestion liquor.
The degree of dilution needs to be performed
accurately and the appropriate dilution correction
calculations made. A preferred strategy is to
re-assay with a technique appropriate for the
concentration range.
Bias – the difference between a test result or the
Terminology mean of a series of test results and the accepted
reference value. It is the inverse of trueness
Accuracy – the closeness of agreement between and is either expressed in absolute terms in the
a test result and the ‘true’ value or accepted reporting unit of measure or as relative per cent
reference value. The term accuracy, when applied (positive or negative).
to a set of test results, involves a combination of
random components and a common systematic Geological blank – a field sample or
error or bias component. Accuracy is related to prepared pulp sample that contains very low
both trueness and precision (ISO 1993 and 1994). or undetectable concentration of the element
being sought. The field or coarse blank measures
Precision – the closeness of agreement
contamination in both the sample preparation and
between independent test results obtained under
analytical stages, while the pulp blank measures
stipulated conditions. Precision depends only on
the distribution of random errors and does not contamination in the analytical stage only. Care
relate to the true value or the specified value. should be taken when inserting coarse blanks to
The measure of precision is usually expressed in monitor possible carry over contamination from
terms of imprecision and computed as a standard specific high grade samples, as the crushing and
deviation of the test results (ISO 1993, 1994 and pulverising sequence varies between laboratories
2004). and even within laboratories, depending on
Trueness – the closeness of agreement between the mills and preparation schemes (robotic or
the average value obtained from a large series manual) used.
of test results and an accepted reference value. Reagent blank – the concentration of the
The measure of trueness is usually expressed in element of interest in the digest solution or
terms of bias (ISO 1993, 1994 and 2004). Note: fusion bead without the addition of the sample.
previously this definition applied to ‘Accuracy’ It is a measure of the contamination contributed
and this meaning is still in widespread use in by reagents.
current literature. Analytical blank – a combination of the reagent
In summary in ISO usage, trueness refers to blank and the signal component attributable to
being centred properly (but without regard to instrument noise.

Field Geologists’ Manual 315

Chapter11.indd 315 27/07/2011 5:54:15 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Certified reference material (CRM) – a say that 95 per cent of all confidence intervals
reference material for which one or more determined in a like manner (from different
parameters have been certified by a technically samples of the population) will include the
valid and recognised procedure and for which a recommended or certified value. It is important
certificate or other valid documentation has been to note that the 95 per cent confidence interval is
issued by a certifying body (ISO 2006). unrelated to control limits (ie it is not equivalent
Matrix-matched certified reference material to two standard deviations of the recommended
(MMCRM) – a CRM prepared from geological value). It is a measure of the reliability of the
material of identical source to the samples being recommended value; generally the narrower the
confidence interval the greater the certainty in
monitored. It has the advantage of having the
the recommended value.
same mineralogy and chemistry and hence
similar digestion and analytical characteristics. Control limits – a window of acceptability for
results obtained by a laboratory for a reference
Reference material (RM) – material for
material and generally calculated from multiples
which one or more parameters are sufficiently
of the standard deviation of the certification
well characterised to be used for the calibration
data. There are a number of methods in
of instruments, the assessment of analytical
use for the determination of the standard
methods, or for assigning values to unknown
deviation as applied to control limits. Ideally
materials.
standard deviations should include within-run
Standard reference material (SRM) – (repeatability) errors and between-run errors.
synonymous with reference material (note: in Determination of between-run errors require
the past this acronym has also been used for a temporal aspect to the certification program
‘secondary reference material’ – a material (ie submission of two or more batches to the
characterised to a lower level than and generally participating laboratories at timed intervals).
not conforming to all ISO criteria for a CRM). Where between-lab (bias) errors are included
Standard – literally this refers to a recognised the calculated standard deviations will be larger
and approved method or procedure (such as the than those calculated solely from within- and
standards published by ISO) but its use to describe between-run errors.
reference materials is also well entrenched in the Repeatability – the closeness of agreement
mineral and chemical industries. between results for the same parameter carried
Pigeon pair, paired offsets, bracketed out by the same method, by the same operator,
standards, variance standards – terms used with the same instrument, in the same laboratory,
to describe a pair of CRMs of almost identical over a short interval of time.
matrix that differ in concentration of one or more Reproducibility – the closeness of agreement
certified elements by an amount approximating between results for the same parameter carried
the typical measurement error of methods out by a different method, or different operator,
in common use at commercial analytical or different instrument, or in a different
laboratories. laboratory, or over an interval of time quite
Certified value – a statistically robust long in comparison to the duration of a single
estimate of the true value of a reference material measurement.
parameter (equivalent to the recommended Field duplicate – a second split of drill core,
value for a CRM). reverse circulation cuttings, outcrop, soil or
Recommended value – a statistically robust stream sediment from the same sampling interval
estimate of the true value of a reference material or area and collected using the same procedure.
parameter (equivalent to the certified value for They are bagged and numbered separately to the
a CRM). primary sample so as to be blind to the sample
Inferred or informational value – an preparation laboratory. Field duplicates contain
uncertified approximation of a reference material all sources of error: the original in situ sampling
parameter, generally derived from results of error, sample size reduction error, subsampling
fewer than six laboratories. error at the pulp level and analytical error. They
Ninety-five per cent confidence interval – can be used to determine the combined field,
an interval for which we can assign a 95 per preparation and analytical precision.
cent probability that the recommended or Preparation duplicate – a second split taken
certified value lies within; alternatively, we can after crushing but before pulverising. If the

316 Field Geologists’ Manual

Chapter11.indd 316 27/07/2011 5:54:15 PM


CHAPTER 11 – SAMPLING, analysis and quality control

crushing procedure is two-stage, preparation in the quality context is the relief of concern
duplicates should be taken at each stage. These about the quality of a product (Vaughn, 1990).
are normally taken routinely by the lab at a Quality control (QC) – is one component of
frequency of one in 40 samples. They contain the quality assurance. It is a system of measures
error of sample size reduction and analysis. intended to ensure that a product or service
Pulp duplicate – a second weighing and adheres to a defined set of quality criteria, or
analysis from the same pulp and usually done meets the requirements of the customer or client.
routinely by labs at a frequency of one in five to Random errors – errors inherent in all
one in 20 samples. The pulp duplicate has the methods of analysis due to fluctuations in
errors of analysis (including subsampling and measurement resulting from unstable power
weighing errors). supplies, instrument noise, non-reproducible
Re-reads – subsequent instrument readings atomisation, plasma fluctuations, X-ray intensity
from the same digestion solution as analysed variation and voluming solutions (human error
initially. These measurements have instrument component). Random errors impose a limit on
(and possibly final dilution) errors only. Re- analytical precision and while they can never be
reads are generally performed by the laboratory eliminated, they can be minimised and quantified
to check specific problems and the results are not by replicate measurement of the same sample.
generally reported to the client. Systematic errors – errors arising from
Lower level of detection (LLD) or detection matrix effects peculiar to a specific technique or
limit (DL) – the lowest concentration of an lithology, pronounced compositional differences
element that can be detected with a 95 per cent between sample and CRM, incorrect instrument
probability. It is calculated by taking a large calibration, human bias due to preconceived
number of instrumental readings at or near the notion of expected grades of CRMs run with
blank level, calculating the standard deviation sample batches. CRMs unfamiliar to the analyst
and determining the value corresponding to are the best way to monitor bias introduced
two or three standard deviations plus the blank through systematic errors.
value (Bloom, 2002; H Waldron, 2011, personal Gross errors – errors of substantial magnitude
communication). mostly originating from human error. They
Upper limit (UL) – the highest concentration are random in nature and include incorrect
of an element that can be quantified by a numbering of samples in the field, inadvertently
particular analytical method. It coincides with interchanging samples during analysis (eg
the high end of the calibration range of the sample swaps in fire assay), miscalculation of
reading instrument and may also be governed by the final result and other randomly introduced
digestion and recovery limitations. human errors. They can be detected and rectified
Sensitivity – refers to instrumental reading with appropriate quality control procedures such
resolution and is closely related to, but distinct as submission of blanks, duplicates and CRMs
from, the detection limit. Generally, the higher with each batch of samples.
the sensitivity the lower the detection limit and
vice versa. The sensitivity of a technique relates Quality control
to the slope of the calibration graph, it is the
slope of the relationship between instrument A sound QA/QC program is the cornerstone of all
signal and analytical concentration. For example, well-managed exploration projects and mining
an ore grade method may report to the nearest operations. It involves the monitoring of sample
0.01 per cent (100 ppm), compared with a low quality and measurement of analytical bias and
level method reporting to the nearest 0.0001 precision. Analytical bias is primarily quantified
per cent (1 ppm). It is customary to report base by the use of certified reference materials and
metals in weight per cent when determined by secondarily by the check assaying of a subset
ore grade methods and in ppm (or ppb) when of samples at a second external laboratory.
analysed by low level (geochemical) methods. Precision is quantified by the use of field,
Quality assurance (QA) – refers to all planned preparation and pulp duplicates. Quality control
or systematic actions necessary to provide entails monitoring laboratory performance by
adequate confidence that a product or service will acquisition of appropriate QC data followed by
satisfy given needs (Kirschling, 1991). Assurance critical evaluation.

Field Geologists’ Manual 317

Chapter11.indd 317 27/07/2011 5:54:15 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Data acquisition can be time consuming and therefore costly.


Misidentification of samples can also add to
The following procedures should be implemented sources of error.
prior to submission of sample batches for analysis.
Insertion of certified reference materials
Blanks should be inserted at regular intervals
(CRMs) or matrix-matched certified reference
in the sample stream to monitor contamination.
materials (MMCRMs) into the sample stream
Coarse blanks consisting of crushed barren
at a frequency of one in 20 or 30. CRMs are
material are subjected to the entire sample prep-
available commercially from recognised
aration process and therefore able to monitor
cross contamination in crushing, splitting and producers. Where possible, CRMs should match
grinding apparatus. Pulp blanks can be used to the lithology and mineralisation style of the
monitor contamination during sample digestion deposit being evaluated, as well as cover the
and measurement. As blanks are geological expected concentration range of target metals.
materials they contain trace amounts of most MMCRMs are prepared by CRM producers
elements, hence it is important that the material from ore samples taken from the same or
selected is appropriate and that control limits are similar deposit being evaluated or mined. They
set at sensible levels. are an ideal compositional and mineralogical
match and therefore better suited to resource
Coarse blanks are often inserted after
definition and grade control QA/QC. Key grades
mineralised samples where the risk of
are cut-off, run-of-mine (ROM) and the 90th
contamination is greatest. Waldron (2011,
percentile. Low-level to lightly mineralised
personal communication) recommends that
CRMs are available for greenfields exploration,
geologists be familiar with the preparation
soil and stream sediment geochemistry and
routine of their laboratory if specific positioning
lithogeochemical applications. They are prepared
of coarse blanks is critical. For example, manual
from oxidised or primary materials and generally
preparation staff frequently work with two
certified for a range of major, minor and trace
pulverisers simultaneously, so each alternate
elements including precious and base metals,
sample will be prepared in a separate mill. A blank
REE and pathfinder elements.
inserted to immediately follow a high sample
would therefore be milled in the other mill. Laboratory audits to confirm adherence to
Robotic preparation uses up to six mills with the contractual obligations and standard operating
first available one being used. In addition, mill protocols.
selection is based on capacity and sample size. The flow sheet depicted in Figure 11.1.2
Blanks are often submitted in smaller amounts illustrates a typical quality control protocol for
and consequently may not be pulverised in the drill core sampling, preparation and analysis.
same larger mills as the samples, thus rendering A QA/QC program should incorporate some or
the monitoring process ineffective. all of the following elements (Bloom, 2002):
Field, preparation and pulp duplicates should
be regularly taken and analysed to confirm the
•• submission of five to ten per cent field
duplicates,
veracity of field sampling and sample preparation
and subsampling protocols.
•• submission of three to five per cent sample
preparation duplicates (approximately 90 per
Routine lab versus check lab – five to ten cent passing 2 mm),
per cent of samples should be re-assayed by a
second unrelated laboratory for comparison
•• submission of three to five per cent pulp
duplicates,
with the routine laboratory’s results. If there
are discrepancies, analyses at a third or umpire •• insertion of three to five per cent of certified
reference materials (CRMs) or matrix-
laboratory may be required. External check matched certified reference materials
analyses can be done using the same pulp (MMCRMs),
packet as was originally analysed, or on a second
pulp split. It is often wise to prepare for inter- •• randomisation of sample numbers before
submission to a laboratory,
laboratory checks by requesting a second pulp
split be taken, at a set frequency, during initial •• comparison of multi-element trends for
sample preparation. This split can be set aside elements determined by different laboratory
until required for external check analyses. procedures,
Recovering excess pulps from storage and •• comparison of results for the same element
splitting off a second portion at a later date determined by different methods,

318 Field Geologists’ Manual

Chapter11.indd 318 27/07/2011 5:54:15 PM


CHAPTER 11 – SAMPLING, analysis and quality control

•• routine insertion of an unprepared barren •• analysis of five to ten per cent of sample pulps
sample (blank), at an independent check laboratory, and
•• routine insertion of a pulverised barren •• analysis of five to ten per cent of sample
sample (blank), rejects at an independent check laboratory.

Drill core sampling, sample preparation and analysis


Figure 11.1.2
Flow sheet for drill core sampling, sample preparation and analysis (modified from Bloom, 2002).

Field Geologists’ Manual 319

Chapter11.indd 319 27/07/2011 5:54:17 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Evaluation techniques – precision assumption of a normally distributed error, and


thus produces biased results when errors have
Scatter plots of field, preparation and pulp a skewed distribution. Abzalov instead concurs
duplicates offer a simple form of graphical with the proposal of Stanley and Lawie (2007)
representation and can be used to provide a to use the average coefficient of variation as the
visual comparison of precision at the various universal measure of relative precision error in
stages of sample size reduction.
mine geology applications.
Thompson-Howarth estimate of precision
Related precision estimates in common use
utilises the results of field, preparation and pulp
include plots of absolute relative difference
duplicates to quantify repeatability precision
(ARD, AMPD) or half absolute relative difference
(Thompson and Howarth, 1973, 1978; Stanley,
(HARD) versus percentile (rank) or duplicate
2003). Field duplicates contain all sources of
pair mean. A target of 90 per cent of samples
error: the original in situ sampling error, sample
having an ARD (per cent) precision better than
size reduction error and analytical error (note:
20 per cent for preparation duplicates, ten per
analytical error includes subsampling and
cent for pulp duplicates and 30 per cent for field
weighing errors at the pulp level). Preparation
duplicates is desirable but not always achievable,
duplicates contain the error of sample size
especially with core samples. Actual values
reduction and analysis, while pulp duplicates
for these parameters are grade and orebody
contain error of analysis only. By assembling and
dependent and should be established on a case
evaluating the field sampling, preparation, and
by case basis.
pulp duplicate results, the errors associated with
each step of sample treatment can be estimated Scatter plots of field, preparation and pulp
by subtracting ‘downstream error’ from each duplicates can be used to provide a visual
composite error estimate (Figure 11.1.3). The comparison of precision at the various stages of
source of the greatest uncertainty in the sampling sample size reduction.
and analytical program can thus be quantified
and targeted for reduction through procedural Evaluation techniques – bias
modifications, if necessary (Smee and Stanley, The Shewart Control Chart or Statistical Process
2005). Control (SPC) Chart is perhaps the most widely
According to Abzalov (2008) the Thompson- used tool to monitor laboratory performance
Howarth method produces consistently lower (Figure 11.1.4). Reference materials are inserted
results than other methods as it relies on the at a frequency of one in 20 to one in 30 into the

Figure 11.1.3
Modified Thompson-Howarth-type plot illustrating the magnitude of total, sampling (8 kg), preparation (1 kg)
and pulp (50 g) error, as determined by duplicate quality control monitoring from an intrusion related gold
deposit (from Smee and Stanley, 2005).

320 Field Geologists’ Manual

Chapter11.indd 320 27/07/2011 5:54:17 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Figure 11.1.4
Shewart control chart showing laboratory results for a gold CRM with Certified value of 2.55 g/t (green line),
two standard deviations lower and upper warning lines (orange) and three standard deviations lower and
upper control lines (red).

Results for Au in CRM (2.55g/t)


2.9
failure
2.8

2.7

2.6
Au (g/t)

2.5

2.4
failure
2.3

2.2
0 10 20 30 40 50
Sample Sequence

sample stream to monitor laboratory bias. The Incorrect use of standard deviations in this way
results obtained over time for each reference will lead to erroneous failures and unnecessary
material are plotted on a time chart together conflict with the laboratory.
with lines corresponding to the expected value, All results plotted on the chart should be
warning and control limits. The expected value random with no patterns or trends. They should
plus two standard deviations and expected value be distributed ‘Normally’, and fit a Gaussian
minus two standard deviations are generally ‘bell’ curve. This means that around 68 per cent
referred to as the upper warning limit (UWL) of the points lie within one standard deviation
and lower warning limit (LWL), respectively. The of the mean, 95 per cent of the points should
expected value plus three standard deviations and lie within two standard deviations of the mean,
expected value minus three standard deviations and 99.7 per cent of the points should lie within
are the upper control limit (UCL) and lower three standard deviations of the mean. It should
control limit (LCL), respectively. These limits be noted that some elemental abundances in
are based on how much variation is expected geological materials have an asymmetric rather
in the results and are multiples of the standard than normal distribution. Such distributions
deviation value provided by the producer of the usually have a preponderance of low values with
CRM. The probability of a ±3 standard deviations a long tail of high values, imparting a positive
failure is only 0.3 per cent and should be a rare
skewness to the frequency distribution curve.
event and as such, considered indicative of an
Data for nuggety gold ores generally typify this
anomalous occurrence worthy of investigation.
kind of distribution. These limitations, however,
For normally distributed data ~5 per cent of
should not apply to good quality CRMs where
results are expected to breach the ±2 standard
results for the certified elements are expected to
deviations warning limits.
be normally distributed.
In the application of control limits it is
If certain rules applying to randomly distributed
important that the method being monitored is
data when plotted on a control chart are not met
the same as that used in the original certification
of the reference material. For example, nickel then an investigation into the causes should be
laterite CRMs characterised by lithium borate initiated. These rules are somewhat arbitrary and
fusion XRF will produce standard deviations are applied at the discretion of the QA/QC analyst:
substantially smaller than those derived from •• A point lying outside the three standard
fusion inductively coupled plasma optical deviations limits – there is only a 0.3 per cent
emission spectrometry (ICP-OES) methods, chance that this is a result of a process under
owing to the superior precision of the former. control.

Field Geologists’ Manual 321

Chapter11.indd 321 27/07/2011 5:54:17 PM


CHAPTER 11 – SAMPLING, analysis and quality control

•• Two consecutive points lying outside two value, the probability that eight points in a
standard deviations – the probability that row fall outside of the one sigma line is less
any point will fall outside the two standard than one per cent. The above caveat applies.
deviations warning limit is only five per cent. These rules exist to highlight when something
The chances of two consecutive points falling improbable happens in a laboratory. The
outside the two standard deviations warning rationale is that if something happens that
limit is only about 0.2 per cent for a process is improbable due to normal causes, it has
under control. happened due to abnormal causes. The abnormal
•• Five, six, seven or eight consecutive points on cause is usually a change and since change to the
the same side of the expected value – there product is undesirable action needs to be taken.
is an equal chance that any given point will The standard Shewart Chart requires each
fall above or below the expected value. The CRM to be plotted separately with concentration
chances that a point falls on the same side on the y axis and sample sequence on the x axis.
of the mean as the one before it is one in It is often instructive to plot all CRMs on the
two. The odds that the next point will also same chart. This is achieved by converting the
fall on the same side of the mean is one in assay values to ‘z-scores’. Z-scores are calculated
four. The probability of getting eight points by subtracting the expected value from the
on the same side of the mean is only around analytical value and dividing it by the standard
one per cent. However, beware of hunting deviation. This converts the analytical value to
for this trend in a control chart. If you have standard deviation units above and below the
200 points then you have 192 sequences of mean, so all standards can be compared on the
eight points. The binomial probability of one same plot.
of these sequences all lying above or below Box and whisker plots provide a means
becomes very probable rather than remotely of graphically comparing the distribution
possible (R Holdsworth, 2011, personal characteristics of results from the primary and
communication). external check laboratories. The five parameters
•• Five, six, seven or eight consecutive points displayed on these plots are the 25th and
trending up or down – the same logic is used 75th percentiles forming the bottom and top
here as above and the above caveat applies. of the box, respectively, the 50th percentile
•• Four of five points falling outside one forming a band near the middle of the box,
standard deviation – in normal processing and the minimum and maximum of the range
68 per cent of points fall within one sigma of values forming the ends of the lower and
of the expected value and 32 per cent fall upper whiskers, respectively (Figure 11.1.5).
outside it. The probability that four out of five To eliminate the influence of outliers the 2nd
points fall outside of one standard deviation and 98th percentiles can be substituted for the
is only about three per cent. The above caveat minimum and maximum values.
applies. Scatter and relative difference plots
compare the original assays with those of the
•• Fourteen points alternating direction – this
check laboratory. In scatter plots the element
rule treats each pair of adjacent points as
concentration of the primary laboratory (x axis)
one unit. The chances that the second point
is plotted against that of the check laboratory
is always higher than (or always lower than)
(y axis). The paired results should lie along a
the preceding point, for all seven pairs is only 45° line corresponding to a 1:1 relationship,
about one per cent. The above caveat applies. with departures indicating bias at one or both
•• Fifteen points in a row within one standard laboratories. Linear-scaled plots highlight the
deviation – in normal operation 68 per cent of higher values and log-scaled plots provide detail
points will fall within one standard deviation at lower concentrations. In relative difference
of the expected value. The probability that plots the per cent relative difference between the
15 points in a row will do so is less than one primary and check laboratory assays (y axis) are
per cent. The above caveat applies. plotted against the mean concentration of each
corresponding pair (x axis).
•• Eight points in a row outside one standard
deviation – since 68 per cent of points lie The QQ plot or quantile quantile plot is a
within one standard deviation of the expected graphical technique for determining if two

322 Field Geologists’ Manual

Chapter11.indd 322 27/07/2011 5:54:17 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Figure 11.1.5 Digestion methods


Box and whisker plot comparing copper results for Perchloric acid digestion is an oxidative attack
primary and external check lab.
offering a cost-effective approach for the analysis
of trace elements in soils, sediments and rocks.
It is generally quantitative for base metal sulfides
and oxides but incompletely attacks silicates and
will not dissolve certain resistate minerals (eg
oxides such as chromite and cassiterite). It does
not totally digest the sample and is not suited to
resource quality work.
Aqua regia digestion uses a mixture of
concentrated hydrochloric and nitric acids able
to dissolve gold, sulfides, carbonates and most
oxides. The proportions of these acids vary
between laboratories, as does the temperature
of the digestion, resulting in variable degrees of
decomposition. Aqua regia digestion has limited
effect on barite, cassiterite, chromite, gahnite,
garnet, ilmenite, monazite, rutile, sphene and
zircon. Silicates are only slightly to moderately
attacked, depending on the mineral and the
degree to which it is altered. If gold is important,
data sets come from the same population. larger sample sizes should be used to reduce the
The quantiles of the first data set (primary sampling error.
laboratory) are plotted against the quantiles Aqua regia is a useful exploration tool and the
of the second data set (check laboratory). If method of choice for many grassroots exploration
the two sets come from a population with programs, as low levels of detection are possible
the same distribution, the points should fall for multi-element suites, including gold. The low
approximately along a 45 degree reference line. detection levels are achievable due to the low
Departures from this distribution are indicative dilution factors and low total dissolved salts. As
of measurement error. with the perchloric acid leach it is not generally
used for resource quality work, even though,
The advantages of the QQ plot are:
where base metals are incorporated exclusively
•• the samples from the check laboratory do in sulfides, oxides and carbonates, extraction is
not need to be paired with samples from almost quantitative.
the primary laboratory and the number of
Three acid digestion uses a mixture of
samples from each laboratory do not need to nitric, perchloric and hydrochloric acids. It is
be equal; and less common than aqua regia and four acid
•• many distributional aspects can be digestions and is not available at all commercial
simultaneously tested: shifts in location, laboratories. It is the method of choice for grade
shifts in scale, changes in symmetry and the control at many copper operations, particularly
presence of outliers. in South America.
Four acid digestion, often referred to as
mixed acid digestion (MAD), total acid digestion
11.2. BEST ANALYTICAL METHODS (TAD) or strong acid digestion, uses a mixture of
FOR COMMON METALS/ hydrofluoric, nitric, perchloric and hydrochloric
MINERALS/COMPOUNDS acids. It is the most potent acid attack in
geochemistry and approaches a total acid digest
Most analytical instrumentation require samples for most mineral species. It is particularly well
to be decomposed in one form or another suited to the analysis of base metal ores. The use
before samples can be read and concentrations of hydrofluoric acid achieves the breakdown of
determined. A variety of digestion methods are most silicates, however, some elements (Si, As,
commonly used. Sb, Cr, U and Au) may be erratically volatilised

Field Geologists’ Manual 323

Chapter11.indd 323 27/07/2011 5:54:18 PM


CHAPTER 11 – SAMPLING, analysis and quality control

depending on digestion conditions, and certain the expensive platinum ware or the addition of
mineral species may not be totally dissolved an oxidant to the flux.
or may not go into solution (eg zircon, sphene, Other fusions such as oxidative alkaline
magnetite, barite, gahnite, chromite and fusions are used for specific applications such as
cassiterite). Whilst major elements (excluding in the determination of barium, strontium and
Si) can be determined by this method, fusion fluoride.
techniques are preferable.
Fire assay – see section headed ‘Gold’.
Four acid digestion is useful for many
applications, including exploration using multi-
element suites. The digestion can be optimised
Reading methods
for higher grade material, where it is particularly A variety of reading methods are available
well suited to the analysis of base metal ores. for quantifying the concentration of elements
Other acid digestions exist for specific in geological materials. Each method has
applications, such as the determination of its strengths and weaknesses and the most
mercury and soluble copper. appropriate instrumentation will depend on the
target elements, concentration ranges, possible
Sodium peroxide fusion provides total
interferences and chemical and mineralogical
decomposition of the sample and is therefore
composition of the samples. A periodic table
well suited to samples containing minerals of
of recommended reading methods for the
interest that are resistant to acid attack. These
commonly measured elements is presented in
include minerals containing rare earth elements
Figure 11.2.1. The selection of which method
(REE), high field strength elements (HFSE), Sn,
or methods to use should be made in close
W, Ti and V. Following fusion at 700°C in nickel,
consultation with laboratory personnel.
zirconium or alumina crucibles the fusion cake
is leached with hydrochloric acid, diluted with Atomic absorption spectrometry (AAS) is
distilled water and elements measured by ICP- a spectroanalytical procedure for the deter-
OES or ICP-MS (inductively coupled plasma mination of the concentration of elements
mass spectrometry). Disadvantages of the based on absorption of light by free atoms in
method are: the gaseous state. The modern form of AAS was
developed in the 1950s and is still regarded today
•• the potential loss of volatiles on fusion;
as a valuable tool in modern geochemistry. Its
•• elevation of detection limits compared with main applications in today’s mineral industry are
acid digest methods – this is due to the for the determination of gold and base metals.
greater levels of dilution required to offset It is cost effective where the number of elements
the high salt levels in solution that may affect of interest are limited to one to four and is the
instruments such as ICP-MS; and finish of choice at many commercial labs for over
•• contaminants from the flux and crucibles range (>1 per cent) base metals. Low detection
(eg Na and Ni, Cr, Cu, Mn, Zn, Zr, Hf and Al, limits for gold can be obtained using graphite
depending on the crucible type). furnace AAS.
Lithium borate fusion is used to prepare a wide Inductively coupled plasma optical emission
range of silicate rocks and ores for analysis by spectrometry (ICP-OES) also referred to as
X-ray fluorescence, inductively-coupled plasma inductively coupled plasma atomic emission
optical emission, inductively-coupled plasma spectrometry (ICP-AES) is a form of emission
mass spectrometry and atomic absorption.
spectroscopy using an inductively coupled
Samples are combined with a mixture of lithium
plasma to produce excited atoms and ions that
metaborate and lithium tetraborate fluxes and
emit electromagnetic radiation at wavelengths
fused in platinum crucibles at temperatures
between 900°C to 1250°C. The resultant glass characteristic of a particular element. The
can be either formed into a disc and presented intensity of this emission is proportional to the
to an XRF instrument for direct analysis or concentration of the element in the sample.
dissolved in dilute acid for measurement by ICP- Compared with AAS, ICP-OES and ICP-MS are
OES, ICP-MS or AAS. As with sodium peroxide modern techniques that have revolutionised
fusion refractory minerals and major oxides, the geochemical study of geological samples.
including SiO2, are readily dissolved. High The main advantage of ICP-OES is the ability
sulfide bearing rocks may require pre-roasting of to rapidly and simultaneously determine multi-
the sulfides prior to fusion to prevent attack on element suites that can be tailored to the various

324 Field Geologists’ Manual

Chapter11.indd 324 27/07/2011 5:54:18 PM


Chapter11.indd 325

Figure 11.2.1
Best analytical methods for common minerals/metals/compounds1.

CHAPTER 11 – SAMPLING, analysis and quality control


Field Geologists’ Manual

1. Modified from Intertek Genalysis Laboratories, with permission.


27/07/2011 5:54:20 PM

325
CHAPTER 11 – SAMPLING, analysis and quality control

commodities. Major and minor elements and They are extremely accurate methods and
some trace elements can be determined by ICP- therefore used extensively in umpire analysis of
OES. It does not have the sensitivity of ICP-MS shipping concentrates.
for many elements. A classical method is frequently used for
Inductively coupled plasma mass spectro- the determination of gold in high grade ores.
metry (ICP-MS) is based on coupling together This entails conventional fire assay with gold
an inductively coupled plasma as a method recovered from the silver prill by parting
of producing ions (ionisation) with a mass (dissolution of the silver in nitric acid) and
spectrometer as a method of separating and weighing. At many commercial laboratories it
detecting the ions. It has superior sensitivity to is the default method at concentrations above
10 ppm while some offer it as an option at grades
ICP-OES and can be used to complement the
as low as 3 - 4 ppm Au.
latter by extending the lower limit of detection
to well below the element concentration ranges
commonly encountered in unmineralised Non-destructive methods
rocks and soils. With appropriate dilutions and These are methods where elements are
calibration controls, high levels can be analysed determined directly on an intact sample, that
too. The two finishes, ICP-OES and ICP-MS, is without prior decomposition of the sample.
are frequently combined using the same digest Generally samples are presented for analysis
solution, thus providing an extremely powerful in powder form to achieve a homogeneous and
analytical tool for the determination of major, representative sample.
minor and trace elements. Pressed powder X-ray fluorescence is a low-
X-ray fluorescence spectrometry (XRF) is cost, rapid turnaround technique for the analysis
based on the emission of secondary (fluorescent) of trace to minor quantities of single elements.
X-rays from elements in a glass disc that have The pulverised sample is mixed with a binder and
been excited by bombardment with high energy pressed into a briquette under a press. It is well
X-rays or gamma rays. The glass disc is prepared suited to elements held in refractory minerals
by fusion of a rock sample in a lithium borate that pose difficulties for acid digest methods in
flux, pouring into a mould and quenching into a particular. Consistent particle size distribution is
disc-shaped glass. The method has been in use essential to minimise errors due to mineralogical
for many decades and it remains the preferred and particle size effects. Matrix matched controls
method for the analysis of major element are important when rapid matrix correction
routines are used. It is not considered suitable
oxides in silicate rocks. For a time following the
for resource evaluation or lighter major elements
advent of ICP technology, XRF was somewhat
although it is in limited use in robotic mine
overshadowed, but with recent improvements in
laboratories for grade control.
instrumentation and automation of disc making
it has undergone a resurgence in popularity. Instrumental neutron activation analysis
Today XRF is the method of choice in whole rock (INAA) is not widely used due to the scarcity
major element analysis, in particular iron and of nuclear reactor facilities; however, it can be
nickel laterite ores and increasingly, base metal a powerful and effective form of analysis for
sulfide ores. It is capable of a very high degree of modest batch sizes. The technique involves
accuracy and precision. bombarding the sample with neutrons in a
reactor and measuring the induced gamma-
ray activity at various wavelengths unique to
Classical methods the particular elements being measured. It is a
With the exception of gold, classical methods, highly precise method particularly suited to Au,
sometimes referred to as ‘wet chem’ methods, REE and U. Elements that can be determined by
are generally reserved for the analysis of very this method include Ag, As, Au, Ba, Br, Ca, Cd,
high grade ores and concentrates. They are Ce, Co, Cr, Cs, Eu, Fe, Hg, Ho, Hf, Ir, K, La, Lu,
expensive methods and have elevated detection Mo, Na, Nd, Ni, Rb, Sb, Sc, Se, Sm, Sn, Sr, Ta, Tb,
levels appropriate to materials with high metal Te, Th, U, W, Yb, Zn and Zr.
concentrations. Broadly they can be divided into
gravimetric methods (eg electro-gravimetric Gold
determination of copper, determination of Zn by
dimethylglyoxime precipitation) and titrimetric A variety of methods are available for the
methods (eg short iodide titration for copper). determination of gold owing to its wide

326 Field Geologists’ Manual

Chapter11.indd 326 27/07/2011 5:54:20 PM


CHAPTER 11 – SAMPLING, analysis and quality control

concentration range from less than one part per generally at 100 or 150 microns using a cloth
billion (ppb) in barren soils, sediments and rocks screen. The entire oversize fraction along with
to in excess of 20 000 ppb (20 ppm or g/t) in the cloth screen is analysed by fire assay to
high grade ores. In addition, the distribution of determine the coarse gold component while the
gold in geological samples can be sporadic and undersize fraction is analysed in duplicate by
cause severe sampling problems. Conventional routine fire assay to determine the non-coarse
lead collection fire assay remains the generally gold component. A weighted average is then
preferred method, however, aqua regia digestion calculated to determine the overall gold content
and cyanide leach methods are useful options for of the original 1 kg sample. The sample should
particular applications. not be pulverised too finely as some gold may
Fire assay involves the firing of a 25 - 50 g plate out on the mill and be lost.
sample in the presence of a borax, soda ash, Aqua regia gold determinations are mainly
litharge (Pb oxide), baking flour, trace silver used for grassroots exploration and are in most
and other additives (dependent on sample instances inappropriate for resource definition.
composition) in a crucible or fire assay pot. Although this digest is incomplete for silicates
For difficult matrices (including concentrates) and refractory minerals, many elements, in
special fluxes and/or a reduced charge weight as particular base metals and many pathfinder
low as 5 g may be used. During the firing process elements, can be determined simultaneously
the lead oxide is reduced to molten lead which with gold. The choice of reading method is
quantitatively scavenges the gold (+ Pt and Pd) dependent on the expected grade range with
and silver from the silicate melt. As the molten the same criteria used for fire assay applicable.
lead is considerably heavier than the silicate Where gold is occluded in silicates recoveries
melt, it collects at the bottom of the crucible.
may be compromised.
The molten lead (and slag) is poured from the
crucible into a mould where the lead forms a Cyanide leaches are used to extract cyanide
button which can be readily separated from the soluble gold from large samples (up to 5 kg) for
adhering slag. The lead button is then placed in both low level and high grade gold analysis.
a cupel made from magnesia. In the cupellation For very low level work, such as grassroots
process the lead and other base metal impurities exploration, bulk leach extractable gold (BLEG)
are oxidised and absorbed into the cupel leaving analysis is typically performed on screened
behind a silver prill into which the gold has stream sediment samples. Detection limits of
quantitatively segregated. The prill is treated with 0.01 ppb are achievable by this method and
nitric acid to remove the silver. The gold residue allow detection of anomalies in the 5 - 20 ppb
is either weighed directly (gravimetric finish) or
range or lower in regional surveys. This method
dissolved in aqua regia and diluted for reading by
can accurately reproduce ppb gold anomalies in
AAS, ICP-OES or ICP-MS (instrumental finish).
calcrete, for example.
For low level gold, where a 1 - 2 ppb detection
level is required, methods such as graphite The large sample size permitted by cyanide
furnace AAS, solvent extraction flame AAS, leaching is ideally suited to the evaluation of
ICP-OES or ICP-MS are employed. For ore nuggety ores where the presence of coarse gold
grade samples relevant to resource definition produces highly variable results on conventional
and grade control, a 10 ppb detection level is fire assay charge weights (25 - 50 g). Extended
generally acceptable and flame AAS or ICP-OES grinding times for nuggety ores will generally not
is used. For sulfide-rich ores a pre-roast step overcome the nugget effect but result in plating
may be included prior to firing to oxidise the out of gold on the mill. It should be noted that
sulfides. Platinum and palladium can also be extraction efficiencies by cyanide leach are less
quantitatively determined by lead collection fire than total (typically around 90 per cent) and
assay with ICP-OES (5 ppb detection) or ICP-MS mimic those achieved in carbon in leach (CIL)
(1 ppb detection) instrumental finish. and carbon in pulp (CIP) plants. Cyanide in
Screen fire assay is typically used for the excess of 0.02 per cent dissolves gold at a rate of
analysis of samples containing coarse gold and 3.36 μg/h, so a typical 24 hour cyanide leach will
where a typical 50 g fire assay would be subject dissolve a gold particle up to 80 µm in diameter.
to unacceptably large sampling errors. A 1 kg Recoveries can be determined by fire assay
representative pulverised split is screened analysis of the leach residue.

Field Geologists’ Manual 327

Chapter11.indd 327 27/07/2011 5:54:20 PM


CHAPTER 11 – SAMPLING, analysis and quality control

The LeachWELL™ accelerated cyanide leach prior to weighing for analysis. The alternative
process is frequently used for high-grade ores. treatment is to allow samples to equilibrate to
This can dissolve gold at a rate of 50 times that the laboratory atmosphere, weigh for analysis,
of normal BLEG. determine moisture on a separate aliquot and
then correct results to a dry basis.
Base metals Ferrous iron determination is important for
magnetite ores, where a significant component of
The chief commodity base metals copper, nickel, iron is present in the divalent state. The sample
zinc and lead occur in measurable concentrations is dissolved in acid and ferrous iron determined
in all geological materials. For low level volumetrically by titration.
reconnaissance evaluation of stream sediment
and soil samples or lightly mineralised rock Davis Tube recovery (DTR) simulates plant
samples acid digest ICP-OES or AAS analysis recoveries of ferromagnetic concentrates using
are the preferred methods. The accuracy of these a Davis Tube. Recoveries of ferromagnetic
‘geochem’ methods is generally ±10 per cent and constituents can be assessed through control of
considered fit for purpose. If gold, platinum or magnetic field strength and sample particle size.
palladium are required then an aqua regia digest, The chemical compositions of both the magnetic
preferably on a larger sample, and with an ICP- and non-magnetic fractions are analysed by XRF.
MS finish is recommended.
Base metal sulfide or oxide ores, where Nickel laterite ores
expected metal concentrations exceed 0.5 to The major element analytical methods used for
one per cent, often require methods with greater nickel laterite ores are the same as those for iron
accuracy (±5 per cent) and precision. Acceptable
ores. Nickel laterites can be highly hygroscopic
methods for resource definition or grade control
(more so than some desiccants) so extreme
work are ore grade four acid digest ICP-OES
care should be taken during the drying stage.
or AAS, sodium peroxide fusion ICP-OES and
lithium borate fusion XRF. Incomplete drying of samples or re-absorption of
moisture following drying can be major sources
of error in the analysis of these materials. For
Iron ores this reason the alternative treatment described
Whole rock major element analysis of iron above for iron ores should be considered where
ores requires a fusion process to obtain total ores are known to be hygroscopic. As with iron
element abundances. Samples are dried at 105°C ores, lithium borate fusion XRF is the preferred
and analysed using lithium borate fusion XRF, method, but lithium borate or sodium peroxide
or lithium borate or sodium peroxide fusion fusion ICP-OES are also acceptable. Where the
ICP-OES. Because of its superior accuracy and full major element suite is not required four acid
precision XRF is widely regarded as the preferred digest ICP-OES will give total concentrations for
method. Loss on ignition (LOI) is determined nickel and cobalt comparable to fusion methods.
on a separate aliquot of the dried sample and
measures weight loss between 105°C and 1000°C Uranium
due to volatilisation of structural OH- molecules
from hydrous minerals (goethite, limonite, clays, For exploration applications involving analysis
micas, amphiboles, etc), CO2 from carbonates of soils and sediments free of resistate minerals,
and SO2 from sulfides, which, in turn, oxidise. an aqua regia digest followed by ICP-MS is
For magnetite ores where ferrous iron (Fe2+) is suitable, particularly if gold is also required.
substantial, a gain in weight may be recorded Pressed powder pellet XRF also provides a rapid,
due to the uptake of oxygen accompanying the cost-effective method of analysis for grassroots
exploration. Higher grade uranium can be
conversion of Fe2+ to Fe3+. In such instances the
determined by ore grade four acid digestion
weight increase associated with oxygen uptake
and ICP-MS analysis or by fusion techniques.
is greater than the combined effects of volatile
Although uranium is remarkably soluble by this
losses. digestion in resistate minerals such as davidite,
Some ores are highly hygroscopic and apatite, zircon, monazite and sphene, other
rehydration following drying can be problematic. constituents are incompletely digested. For this
Dried samples should be cooled in a desiccator reason fusion techniques are recommended

328 Field Geologists’ Manual

Chapter11.indd 328 27/07/2011 5:54:20 PM


CHAPTER 11 – SAMPLING, analysis and quality control

for refractory minerals likely to host uranium. open and the NiS button recovered from the slag.
Lithium borate fusion with XRF or ICP-MS There is no cupellation stage. The NiS button is
finish or sodium peroxide fusion ICP-MS are the simply pulverised and a portion of it digested in
preferred methods. hydrochloric acid. The nickel sulfide is dissolved
In summary, depending on compositional leaving a residue of insoluble PGE sulfides which
matrices, the following methods are suitable for are dissolved in aqua regia and then analysed
the grade ranges indicated: by ICP-MS or, if grades are sufficiently high
(>10  ppb), by ICP-OES. The insoluble PGE
•• geochem aqua regia or four acid digest ICP-
sulfides can be analysed directly by INAA, but
MS: 0.01 ppm to one per cent U,
with poorer sensitivity than ICP-MS.
•• pressed pellet XRF: 5 ppm to one per cent,
It should be noted that nickel sulfide collection
•• ore grade four acid digest ICP-MS: 0.5 ppm fire assay is a highly specialised method and not
to >30 per cent, widely available.
•• delayed neutron counting: 0.1 ppm to one per
cent, Partial selective leach
•• lithium borate fusion ICP-MS: 0.5 ppm to
Mobile metal ions and related pathfinder
>30 per cent, elements associated with orebodies can be
•• sodium peroxide fusion ICP-MS: 0.5 ppm to mobilised in groundwaters and migrate upwards
>60 per cent, to be fixed in surficial soils and sediments. Partial,
•• lithium borate fusion XRF: 10 ppm to >60 per selective or sequential leach geochemistry
cent, and extracts these loosely bound ions from the
soil and sediment media and provides an
•• instrumental neutron activation analysis:
enhanced geochemical response to underlying
0.5 to 50 per cent.
mineralisation. Conventional strong acid
digestions are less selective than partial leaches
Rare earth elements (REE) and high and in mature terrains the bulk chemical
field strength elements (HFSE) signature of deep overburden may mask the
Lithium borate fusion ICP-MS is the best presence of these subtle anomalies.
method for the analysis of REE and HFSE (Y, There are a range of partial leaches available
Zr, Hf, Nb, Ta, Th, U) as the host minerals are that vary in their ability to selectively and
commonly refractory and the heavy REE not sequentially dissolve specific components of
completely attacked by acid digestion. Sodium the sample. They are analogous to leaches
peroxide fusion ICP-MS will also provide total used for agricultural and environmental
concentrations but detection levels are elevated purposes, although many of the methods used
in geochemical exploration are proprietary or
compared with borate fusion. It is important
licensed to commercial laboratories. Examples
that samples are finely ground (95 per cent
of the more widely available leaches include
minus 75 micron) to ensure complete fusion of water, ammonium acetate, ionic, cold and hot
refractory phases. hydroxylamine hydrochloride and sodium
pyrophosphate leaches. Consultation with the
Platinum group elements (PGE) laboratories is highly recommended to optimise
sampling protocols and digest requirements. An
The conventional lead collection fire assay orientation study is also valuable. Leaches are
method is capable of quantitatively recovering generally done on screened samples that have
platinum and palladium in addition to gold. For not been pulverised.
all six PGE (Pt, Pd, Ru, Rh, Ir and Os) nickel
Water – a leach using demineralised water for
sulfide (NiS) is used as the collector in place of
extracting the water-soluble component.
lead. The PGEs have a very high partitioning
coefficient into a NiS melt relative to a silicate Ammonium acetate – a buffered leach to
melt and are consequently quantitatively extract exchangeable cations adsorbed by clay
scavenged into the NiS phase. Gold may not and elements co-precipitated with carbonates.
be completely partitioned into the NiS. Unlike It is a very weak leach targeting the most weakly
lead collection fire assay, the molten material is bound ions and often used in sequential leaches
allowed to cool in the pot before the pot is broken prior to more aggressive leaches.

Field Geologists’ Manual 329

Chapter11.indd 329 27/07/2011 5:54:20 PM


CHAPTER 11 – SAMPLING, analysis and quality control

Ionic – a buffered sodium cyanide leach MS instruments incorporating electro thermal


targeting ions weakly bound to the surface of vaporisation technology, gold can now be directly
particles in regolith or soil. determined in groundwaters to a detection limit
Cold hydroxylamine hydrochloride – this of 0.5 parts per trillion. Brines may have to be
widely used leach is selective for manganese diluted before analysis, with some degradation
oxides which are powerful scavengers of specific of sensitivity as a consequence. Being a more
heavy metals. Manganese occurs as oxides and homogeneous medium, a water sample will be
hydroxides in several oxidation states and in more representative of the water body than a
various amorphous and crystalline forms having rock sample will be of the surrounding bedrock.
an extremely high cation exchange capacity. Large areas can be explored with relatively few
samples, potentially providing a cost-effective
These compounds are therefore very efficient
exploration tool.
scavengers of many trace elements. The leach
will not attack crystalline iron oxides, but will
dissolve very small amounts of amorphous Biogeochemistry
hydrous iron oxide. Biogeochemistry, or phytogeochemistry, uses
Hot hydroxylamine hydrochloride – a leach living plant tissue as the medium for exploration.
combining higher hydroxylamine concentration Plants scavenge elements and translocate them
and elevated temperature and capable of to foliage, twigs, bark, flowers and seeds making
dissolving amorphous hydrous iron hydroxide them efficient geochemical samplers, able to
and crystalline manganese hydroxide while integrate material from a much wider area than
leaving crystalline iron oxide substantially intact. the equivalent single soil sample. Foliage and
Amorphous iron hydroxides are more powerful bark generally give the strongest geochemical
scavengers of mobile metal ions than crystalline signature. Samples are easy and cheap to collect
forms of iron oxide. and, compared with conventional forms of
Sodium pyrophosphate – this leach specifically geochemical exploration, environmental impact
targets metals forming chelation complexes with is minimal. Analysis of plant tissue provides
the humus component of soils and sediments. It information on the underlying geology and can
will not attack sulfides nor dissolve significant be particularly effective in areas of transported
amounts of amorphous iron oxides. Variations in cover or regolith-dominated terrain.
the concentrations of organically-bound metals
Trace element levels in plants are generally
such as copper, zinc, cobalt, nickel, iron and
much lower than in soils and consultation with
aluminium have been shown to provide good
the laboratory regarding sampling protocols
anomaly contrast.
to minimise contamination is recommended.
An orientation study is advisable to determine
Hydrogeochemistry the most appropriate species too collect and
The chemical analysis of ground waters can be which tissue type works best. Samples are dried,
a valuable adjunct to exploration programs. milled in specialised equipment and digested
Trace elements in ground waters are generally in a modified aqua regia solution. Analysis
present at extremely low levels. Special by combined ICP-OES and ICP-MS provides
handling and storage procedures are needed multi-element suites that are the most useful for
for the preservation of samples. Waters can interpreting biogeochemical data.
be analysed directly with ICP-MS and with the Samples may be ashed prior to digestion to
dramatic improvements in sensitivity of ICP- enhance sensitivity for some applications.

330 Field Geologists’ Manual

Chapter11.indd 330 27/07/2011 5:54:20 PM


CHAPTER 11 – SAMPLING, analysis and quality control

11.3. GENERAL PREFERRED SAMPLE MASS NOMOGRAM1

Maximum ore mineral particle diameter in mm

1. From Gy, P, 1956. Nomogramme d’Echantillonnage (Societe de Minerais et Metaux: Paris).

Field Geologists’ Manual 331

Chapter11.indd 331 27/07/2011 5:54:22 PM


CHAPTER 11 – SAMPLING, analysis and quality control

11.4. GRAPHS OF PARTICLE SIZE AND PREFERRED SAMPLE MASS FOR


GOLD ASSAYS1
Figure 11.4.1
Relationship between number of particles per 1 kg sample, particle mass (assuming all particles to be of
uniform mass), and grade or tenor of the sample in parts per million. Scales to right relate grain size of
gold spheres and flakes to particle mass.

1. From Clifton, H E, Hunter, R E, Swanson, F J and Phillips, R L, 1969. Sample size and meaningful gold analysis,
USGS prof paper 625-C.

332 Field Geologists’ Manual

Chapter11.indd 332 27/07/2011 5:54:23 PM


MINING AND ECONOMIC GEOLOGY

CHAPTER 11 – SAMPLING, analysis and quality control

Figure 11.4.2
Size of sample required to contain an expected 20 particles of gold as a function of the combination of gold
particle size and grade, assuming all gold particles to be of uniform size and randomly distributed in the deposit.

Size of sample required to contain an expected 20 particles of gold as a function of the combination of gold particle size and
grade, assuming all gold particles to be of uniform size and randomly distributed in the deposit.

124 Field Geologists’ Manual

Field Geologists’ Manual 333

Chapter11.indd 333 27/07/2011 5:54:23 PM

You might also like