1

Mineral Resource and Ore Reserve Estimation Second Edition

CHAPTER 3: GEOLOGICAL INTERPRETATION AND GEOLOGICAL MODELLING

Calculated Mineralogy and its Applications

S Halley

Contact author: Scott Halley Consulting Geochemist Mineral Mapping Pty Ltd 24 Webb Street, Rossmoyne, WA, 6148 0438 998583 minmap@westnet.com.au

2

Calculated Mineralogy and its Applications

ABSTRACT

One of the biggest problems in generating realistic geological and alteration models is the lack of

consistency in the logging data. Mine site drill hole data bases typically contain logging data collected over a

period of many years. In most cases the logging will have been completed by tens of different geologists.

Consistent logging requires a significant level of skill and experience, but this task is generally assigned to

the most junior geologists. As a result, there is a large degree of subjectivity in the recognition of basic rock

types, and an even greater lack of consistency in correctly identifying alteration mineralogy, particularly when

it involves logging RC chips. Recent advances in analytical techniques and the application of portable

analytical devices and spectrometers makes it possible to collect systematic, quantitative geochemical and

mineralogical data that is independent of operator bias. These applications include ICP-MS/AES

geochemistry, portable XRF devices, and SWIR spectrometers. From these types of data, it is possible to

use immobile trace element geochemistry to correctly identify rock types, major element geochemistry to

identify and quantify sulphide and silicate mineralogy, pathfinder chemistry to model haloes to mineralized

structures, and SWIR mineralogy to qualitatively recognise alteration mineralogy and solid solution mineral

haloes. Having a reliable geological and mineralogical framework can help improve the reliability of resource

models, assist in the domaining of ore types and spatially map areas with possible geotechnical problems. A

more accurate and quantitative model of mineral distribution throughout the resource model may also

increase the reliability of metallurgical predictions, an important modifying factor in the conversion of

resources to reserves.

INTRODUCTION

It is commonly difficult to use geological observations made from drill core and RC chips to create geological

models because in most deposits, logging lithology and alteration is not easy. It is a task commonly

assigned to the most inexperienced geologists, and also because of staff turnover, the relevant observations

may have been made by a large number of geologists. The result is a high degree of subjectivity and

inaccuracy in the data that is collected. In this paper, the use of geochemistry and spectral logging to create

3D models of geology and mineralogy is discussed. Three case studies are presented. The first case study

shows how a geological model of an orogenic gold deposit was created from systematic logging of drill

cuttings with a portable infrared spectrometer. This shows how a variogram could be created and applied to

grade interpolation using real geological constraints rather than just from spatial distribution of assay values.

The second case study shows the classification of the silicate alteration mineralogy in an epithermal gold

deposit by using the ratios of major elements from an ICP-AES geochemical assay suite. A classification of

mineralogy in this situation highlights the three dimensional distribution of clay species that may be

deleterious in the mineral processing side of the operation or perhaps cause geotechnical issues during

mining. The third case study shows the classification of sulphide mineralogy in a porphyry copper deposit by

using ratios of Cu, Fe, S, Au and Ag. A quantitative three dimensional model of the sulphide speciation

should assist in mine planning, as the nature of the mill feed will be predictable through the life of the mine.

3

CASE STUDY; SALT CREEK GOLD DEPOSIT

The Salt Creek Gold Deposit is located about 80km south east of Kalgoorlie (Integra Mining Limited, 2010). It

was discovered in 2007 by Integra Mining Limited. The pre-mining resource was 4.6MT @ 2.72g/t Au for

403,566 ounces of contained gold. The deposit is hosted in a mafic rock with pervasive, texturally destructive

chlorite alteration. There were no distinctive features within the mafic rock package that allowed recognition

of anything more than a very broad stratigraphic sequence. All of the early drilling was RC, which made rock

recognition even more difficult. The deposit has a very distinct sub-horizontal plunge, but in the early stages

of drilling, the geological controls on the plunge geometry were not obvious.

From very early on in the drill out, all drill holes were systematically logged with a portable infrared

spectrometer. The infrared logging protocol used at Salt Creek was to measure one spectrum per metre of

RC chips or diamond drill core using an Analytical Spectral Devices Terra Spec instrument. The spectra

were measured on the silicate groundmass (not on veins or sulphides) in order to consistently map the

alteration mineralogy and to measure solid solution variations in chlorite and sericite. The data were

interpreted using the CSIRO TSG (The Spectral Geologist) software (http://www.thespectralgeologist.com/).

Chlorite has a very distinctive infrared spectrum (Figure 1). It has two characteristic absorption features at

around 2250nm and 2340nm (Pontual, 2007). These correspond to absorption of light by Fe-OH bonds and

Mg-OH bonds in the mineral structure. The exact position of the absorption minima varies depending on the

solid solution composition of the chlorite. Magnesium-rich chlorites generally have shorter wavelengths, and

iron-rich chlorites have longer wavelengths (Herrmann et al., 2001).

By selecting the subset of data points where the TSG software picked only chlorite as the phyllosilicate

phase, and then mapping the wavelength of the 2250nm feature in 3D, it was recognised that there was a

very distinct stratification in the chlorite chemistry (Figure 2). A batch of 200 half-core samples was assayed

by an ICP-MS/AES method to calibrate the spectral signatures against whole rock chemistry. The chemistry

demonstrated that the mafic host rock at Salt Creek was a fractionated dolerite sill. Given the fine-grained,

texturally destructive nature of the chlorite alteration, it was impossible to recognise the fractionation units

within the sill in drill core, let alone RC chips. The chlorite chemistry determined from the infrared spectra

provided a very sensitive proxy for mapping the fractionation within the dolerite. Surprisingly, the solid

solution variations in the chlorite chemistry are controlled by the pre-existing chemistry of the host rock, not

by the nature of the hydrothermal fluid. Orogenic gold systems are driven by low salinity, neutral pH fluids, in

which Fe and Mg have a relatively low mobility. In acidic, saline hydrothermal systems, the chlorite chemistry

would have reflected hydrothermal fluid conditions rather than host rock chemistry.

From the wavelength variations in the chlorites, upper and lower boundaries of the long wavelength domain

were modelled, corresponding to the fractionated quartz dolerite. An offset on the upper margin of the

fractionated quartz dolerite unit was mapped from the spectral signatures. This corresponded to a small

displacement shear which controlled the mineralisation within the sill. The vast majority of the mineralisation

is located in a brittle vein array adjacent to the shear zone, within the quartz dolerite (Figure 3). The

geological model derived from the spectral data demonstrates a very common Kalgoorlie theme. The plunge

of the orebody is controlled by the intersection of a small displacement shear with a brittle quartz dolerite.

4

In this model, 47,000 were spectra measured with a portable spectrometer. The average logging rate was

1,500m/day for RC chips, and 1,000m/day for diamond core. The total cost of the spectral logging, including

instrument hire, operator wages and interpretation was around 80c/meter. A model like this could not have

been created from visual core logging because the texturally destructive nature of the chlorite alteration

obscured the dolerite textures. The spectral logging eliminated subjective bias of the geologists.

The consequence of having a robust geological model is that the plunge geometry of the ore body is

predictable and should be along the intersection of the shear zone and the quartz dolerite. This provides an

independent validation of variograms generated entirely from grade distribution patterns.

CASE STUDY; ÇÖPLER GOLD DEPOSIT

The Çöpler Gold Mine is jointly owned by Alacer Gold (80%) and Lidya Mining (20%). It is located in Turkey,

120 km southwest of Erzincan and 550 km east of Ankara. It has Proven & Probable Reserves of 95.4Mt at

1.4g/t gold for 4.4 million ounces of contained gold, and Measured & Indicated Resources of 182.6Mt at

1.4g/t gold for 8.0 million ounces of contained gold (as at June 2012, 100% basis)

Çöpler is an epithermal gold deposit (Yigit, 2006). It is centred on a nested group of dioritic stocks that have

intruded into meta-siltstones and greywackes, and capped by a thick limestone unit. There are at least three

centres of early, low grade, porphyry Cu-Au mineralization, with phyllic (sericite-pyrite) alteration haloes,

predating the epithermal mineralisation. Intense, low temperature, intermediate-argillic (illite, illite-smectite,

smectite) alteration associated with the epithermal mineralization overprints the potassic and phyllic

alteration in the porphyry system.

All of the drill samples from Çöpler were assayed using a 4-acid digest and an ICP-AES method. This

analytical technique achieves close to a complete digest, particularly for the silicate minerals. The elements

assayed include gold, copper, a very broad suite of pathfinder elements, and all of the major elements other

than silica.

With a basic understanding of the alteration mineralogy, scatter plots of the major elements can be designed

that will allow a classification of the alteration signatures for each assay interval. In mineral systems like

Çöpler that involve feldspar-destructive alteration, the most useful plot is a molar ratio of K/Al versus Na/Al.

Consider a rock that is totally sericitised. The mineralogy of the rock might be muscovite-quartz-carbonate-

pyrite. All of the K and Al in that rock is contained within muscovite which has a composition of

KAl3Si3O10(OH)2. Therefore the ratio of K:Al in the sericitised rock is 1:3. Similarly, a totally KSpar (KAlSi3O8)

altered rock has a K:Al ratio of 1:1. In the same way, albitisation can also be tracked. Albite is NaAlSi3O8 with

Na:Al =1:1.

Figure 4 shows a K/Al versus Na/Al molar ratio plot constructed from 160,000 assay points from the Çöpler

data base. For clarity, a point density contour overlay has been applied. This figure shows an extraordinary

extent of Na-depletion. The sodium-depleted rocks have complete spectrum of K/Al ratios from 0 to 0.33.

This shows a continuum of alteration from intense phyllic alteration through to intense argillic alteration. The

5

mineralogy ranges from muscovite to illite to interlayered illite-smectite to smectite and/or kaolinite. The clay

mineralogy cannot be resolved from the assays. Infrared spectra were needed to determine the clay

mineralogy. The hypogene clays are predominantly smectite, but feldspar-bearing rocks are weathered to

kaolinite within the supergene zone. It is impossible for the geologists to visually distinguish intense illite from

intense smectite alteration, or any gradation in between the two end-members, particularly when most of the

drilling is RC.

The down hole assay intervals were classified according to the mineralogy defined from the K/Al versus

Na/Al molar ratio plot in Figure 4. From this plot, the intensely Na-depleted samples were classified as strong

illite, illite-smectite, or smectite. Other points were classified as weakly or moderately sericitised depending

on how far they had shifted from the projected position of least-altered diorites or sediments towards the

projected composition of muscovite. Phengitic white mica may have K/Al ratios up to 0.45. Samples with a

K/Al ratio of greater than 0.45 must contain another potassic mineral in addition to phengitic sericite. These

points were classified as KSpar-bearing. The limestone and dolomite-bearing samples were classified from a

Ca vs Mg plot.

The mineralogy classifications are plotted on a long section through Çöpler in Figure 5. This figure highlights

the role of the limestone in acting as a cap on the hydrothermal system. A low grade porphyry copper centre

is apparent from the copper assays in Figure 6. There is also a narrow band of copper enrichment at the

base of the limestone. The alkali element contents do not change significantly in the potassic alteration zone,

because feldspars are stable here. There is a minor potassium addition where hornblende is replaced by

biotite. Gold mineralization is primarily related to the intense illite alteration (Figure 7).

A mineralogical model like this can serve a number of purposes in relation to the estimation of mineral

resource and other applications. Firstly it provides some important clues in the resource modelling to help

decision making about grade continuity and grade correlations from hole to hole. It is very useful from a

geotechnical perspective, because rock hardness will be closely correlated with the bulk mineralogy of the

rock. It provides the basis for recognising potential problems with pit wall stability, because the distribution of

clay-rich domains within planned pit walls is clearly highlighted. It is also very useful from a metallurgical

perspective. Swelling clays may cause problems on the leach pads, or in mill circuits. Knowing the 3D

distribution of the clays allows a targeted campaign of test work so that potential problems can be evaluated

before they impact on the operation.

CASE STUDY; HAQUIRA PORPHYRY COPPER DEPOSIT

The Haquira porphyry copper deposit is owned by First Quantum Mining Limited. It is located in the

Andahuaylas – Yauri porphyry copper belt in Southern Peru. It is a copper-molybdenum porphyry with minor

gold credits. Haquira currently has reported measured and indicated resources of 3.7 million tonnes of

contained copper equivalent and inferred resources of 2.4 million tonnes of contained copper equivalent.

All drill samples from Haquira were assayed with an aqua regia digest and an ICP-AES assay method. Aqua

regia dissolves sulphides, carbonates, some secondary oxides, clay minerals and some chlorite, but it will

6

not dissolve most of the silicate phases. Therefore, from an aqua regia digest, it is not possible to chemically

classify rock types and alteration mineral assemblages as was done in the Çöpler case study. What can be

done with this data is a classification of the sulphide mineralogy from Cu:Fe:S:Au:Ag ratios.

Porphyry copper deposits are formed from very oxidized hydrothermal fluids (Seedorf, 2005). Initially, all of

the sulphur that is exsolved from the underlying magma chambers is in the form of SO2. As the fluid cools

between 450°C and 350°C, SO2 is converted to H2S via a disproportionation reaction;

4SO2 + 4H2O = H2S + 3H2SO4

The ore fluid initially contains copper in excess of reduced sulphur, so bornite, with a high ratio of Cu to S is

generally the first copper mineral precipitated in a porphyry deposit. As the supply of reduced sulphur

increases, chalcopyrite is precipitated, and when there is an oversupply of reduced sulphur, pyrite is formed.

Therefore porphyry copper deposits generally show a zonation pattern from an inner bornite zone, to

chalcopyrite, to an outer and higher pyrite-rich zone (Dilles and Einaudi, 1992). Modelling the distribution of

the sulphide minerals is important because the metallurgical characteristics of the ore will change

significantly from zone to zone. This has traditionally been done from visual logging, with recognition and

visual estimates of the proportions of bornite, chalcopyrite, pyrite etc. This requires a significant level of skill

and experience on the part of the loggers, and a degree of subjectivity and inconsistency is unavoidable.

Figure 8 shows a Cu:Fe:S ternary plot with the projected positions of bornite, chalcopyrite, pyrite, chalcocite

and malachite. From the ternary plot, samples with a Cu:S ratio greater than that of bornite (Cu5FeS4) were

selected, coloured green, and labelled as oxide. Samples with a Cu:S ratio greater than that of chalcopyrite

(CuFeS2) and up to the ratio of bornite were selected, coloured purple, and labelled as bornite. Samples

within this bornite group are likely to have a mixture of bornite and chalcopyrite. Similarly, three other groups

were created for pyrite-dominant, mixed pyrite-chalcopyrite, and chalcopyrite-dominant samples.

All samples with less than 0.1% Cu were assigned a grey colour and labelled as “barren”. A small group of

enargite-bearing samples were identified from a plot of Cu versus As. The enargite bearing samples have a

linear trend with a Cu:As ratio corresponding to the stoichiometry of enargite.

Figure 9 shows scatterplots of Cu vs Au and Cu vs Ag for the two colour groups defined as “bornite” and

“chalcopyrite”. Note that both of these plots show bimodal populations. Virtually all of the points classified as

bornite-bearing occur in the high-Au, high-Ag populations. The reason for this bimodal distribution is that

bornite contains significantly higher levels of Au and Ag in solid solution than chalcopyrite. Ten percent of the

points classified as “chalcopyrite” overlap with the bornite points on the Cu vs Au and Cu vs Ag plots. When

the nature of the veining at Haquira is examined, the reason for this overlap becomes apparent. There is a

later generation of quartz-molybdenite-pyrite veins that overprints earlier bornite and chalcopyrite veining.

The points with high Au/Cu and Ag/Cu classified as “chalcopyrite” bearing, are actually samples with bornite

veins, overprinted with quartz-molybdenite-pyrite, thus shifting them to lower copper:sulphur ratios.

Thus, by considering the Cu:Fe:S:Au:Ag ratios within the assay samples, every assay interval can be

classified in terms of the likely sulphide mineralogy within the samples. The metal values within each assay

7

interval are a simply a reflection of the mineralogy of those samples. Recalculating the assays as weight

percent of minerals rather than weight percent of elements provides the basis for making a quantitative

mineralogical model. In the example illustrated here, this should greatly improve the reliability of

metallurgical predictions for the ore body as a whole. Bornite-rich ore makes a higher grade concentrate than

chalcopyrite-pyrite ore.

CONCLUSIONS

Over the last 40 years the standard commercially available assay techniques have changed from AAS to

INAA to ICP-AES to ICP-MS. The real cost per assay has barely changed, but we can now have 40 or more

elements routinely assayed with detection limits way lower than average crustal abundance levels. In

addition, we have an array of portable instruments that measure chemistry or directly measure mineralogy at

a rate fast enough to use as routine logging instruments. This rapid, cheap, off-the-shelf technology

quantifies core logging in a way that is vastly superior to what can be logged visually. The data is

independent of observer bias. The results can be presented in a categorized format (for example to describe

rock type signatures or alteration mineralogy characteristics), or in numeric formats to describe abundances

or solid solution compositions of minerals, or metal abundances. The data formats are ideal for 3D

modelling.

The potential applications of good quality ICP-MS/AES data and spectral logging are limited only by the

imagination of the geoscientists. Immobile trace elements can be used to geochemically fingerprint

lithological units. The major elements can be used to quantify the alteration mineralogy. Mapping clay

distributions can be used to predict potential metallurgical problems or geotechnical problems. Rock

hardness is primarily a function of mineralogy, so correlating a small number of test results against

mineralogy estimated from whole rock analyses may allow a work index to be fully predictable in a 3D model

where appropriate assay methods have been used.

Systematic geochemical and spectral logging does not replace the need for routine core logging. Rather than

wasting time on aspects of logging that can be performed far better and more consistently with other

methods, the geologists time will be freed up to concentrate on logging textures, structure and geometric

aspects that cannot be measured chemically. Over the next few years, routine ICP-MS/AES assaying and

spectral logging should become industry best practice.

ACKNOWLEDGEMENTS

Integra Mining (now part of Silver Lake Resources Limited), Alacer Gold Corporation, and First Quantum

Minerals Limited are thanked for permission to use and publish their data.

REFERENCES

8

Dilles, J.H. and Einaudi, M.T., 1992. Wall-rock Alteration and Hydrothermal Flow Paths about the Ann-Mason Poprhyry Copper Deposit, Nevada; a 6km Vertical Reconstruction. Econ Geol., vol. 87, 1963-2001

Herrmann, W., Michael Blake, M., Doyle, M., Huston, D., Kamprad, J., Merry, N., Pontual, S., 2001. Short Wavelength Infrared (SWIR) Spectral Analysis of Hydrothermal Alteration Zones Associated with Base Metal Sulfide Deposits at Rosebery and Western Tharsis, Tasmania, and Highway-Reward, Queensland. Econ. Geol., vol. 96, 939-955

Integra Mining Limited, 2010. Annual Report

Pontual, S., Merry, N and Gamson, P., 2007. GMEX, Volume 1, Spectral Interpretation Field Manual. Published by

AusSpec International.

Seedorff,E., Dilles, J.H., Proffett, J.M., Einaudi, M.T., Zurcher, L., Stavast, W.J.A., Johnson, D.A., and Barton, M.D., 2005. Porphyry Deposits: Characteristics and Origin of Hypogene Features. Economic Geology, 100

th Anniversary

Volume, 251 – 298

Yigit, O., 2006. Gold in Turkey—a missing link in Tethyan metallogeny. Ore Geology Reviews, Volume 28, 147–179

9

FIGURES

FIG 1. Stacked spectra of chlorite, showing the characteristic absorption features at 2250 and 2340

nanometers. Note how the wavelength at the minimum point for the 2250 feature shifts from 2250 for the

blue spectrum, up to 2260nm for the red spectrum.

10

FIG 2. 3D view showing the distribution of chlorite-bearing spectra from the Salt Creek Gold deposit. This

view is looking directly down the plunge of the Salt Creek dolerite sill. The sample points are coloured by the

wavelength of the 2250nm Fe-OH absorption features from the SWIR spectra, as shown in Figure 1.

FIG 3. Cross section of Salt Creek showing the upper and lower boundaries of the fractionated quartz

dolerite unit and the modelled shear zone. The drill strings are coloured by gold assays; blue < 0.2ppm Au,

red > 2.0 ppm Au.

11

FIG 4. Feldspar-Sericite K/Al versus Na/Al molar ratio plot for the Çöpler assay data, also showing the

projected compositions of some common alteration minerals. This plot forms the basis for graphically

defining the alteration mineralogy signatures for all the assay intervals.

FIG 5. Alteration mineral classifications from molar ratio plots on a long section of the Çöpler deposit.

12

FIG 6. Çöpler long section with copper assays. blue = Cu < 100ppm, red = Cu > 1500ppm.

FIG 7. Çöpler long section with gold assays. blue = Au < 0.1 ppm, red = Au > 1.5ppm. The red shape is a

volume that encloses assay intervals where the alteration has been geochemically classified as “strong illite”.

13

FIG 8. Ternary plot of Cu:Fe:S from the Haquira East porphyry copper deposit. The grey points are samples

with Cu assays less than 0.1%. The colour scheme shows the sulphide mineralogy predicted from this

ternary plot.

FIG 9. Copper versus gold and copper versus silver scatterplots from Haquira East, showing red points

(defined as the chalcopyrite group) and purple points (defined as the bornite group) from the ternary plot in

Figure 8. Note the bimodal distribution on both plots.

14

FIG 10. The Haquira East sulfide mineralogy predicted from the ternary plot in Figure 8 and the scatterplots

in Figure 9 was plotted in 3D, and a volume was created to enclose the bornite-bearing assay intervals.

FIG 11. The modelled chalcopyrite shell from Haquira East. Assay intervals within this shell contain bornite

(Figure 10) or chalcopyrite, with only minor pyrite.