DOCTOR A L T H E S I S

ISSN: 1402-1544
ISBN 978-91-7439-690-4 (print)
ISBN 978-91-7439-691-1 (pdf)
Lule University of Technology 2013

Cecilia Lund Mineralogical, Chemical and Textural Characterisation of the Malmberget Iron Ore Deposit for a Geometallurgical Model

Department of Civil, Environmental and Natural Resources Engineering

Division of Geosciences and Environmental Engineering

Mineralogical, Chemical and Textural Characterisation

of the Malmberget Iron Ore Deposit
for a Geometallurgical Model

Cecilia Lund

Mineralogical, Chemical and

Textural Characterisation
of the Malmberget Iron Ore
Deposit for a Geometallurgical Model

Cecilia Lund

Lule University of Technology

Department of Civil, Environmental and Natural Resources Engineering
Division of Geosciences and Environmental Engineering

Printed by Universitetstryckeriet, Lule 2013

ISSN: 1402-1544
ISBN 978-91-7439-690-4 (print)
ISBN 978-91-7439-691-1 (pdf)
Lule 2013
www.ltu.se

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

ABSTRACT
The northern Norrbotten ore province is an intensely mineralised area and has traditionally been a very important mining district. It contains Fe, Cu-Au, Au, and Ag deposits ranging from world-class ore deposits to
small and uneconomic ones (Martinsson 2004). The most important are the Kiruna and Malmberget iron ores
and the Aitik Cu-Au deposit.
The cross-discipline approach called geometallurgy connects two different but closely related areas in the
mining industry, namely geology and mineral processing. It involves understanding and measurements of the
ore properties signicant for its successful processing. Geometallurgy takes both the geological and mineral
processing information to create a spatially-based (3D) predictive model for product management in mining
operations (Lamberg, 2011).
This case study investigates how to establish a geometallurgical model using the Malmberget iron ore deposit
as a case study. A mineralogical approach (Lamberg 2011) was selected meaning that the focus is on mineralogy,
and therefore parameters like modal mineralogy, mineral textures, mineral associations, mineral grain sizes and
their relation to liberation characteristics are important. The main effort is to deliver a geological model which
gives quantitative rather than descriptive information to be used in a process submodel.
The ore characterisation (Papers I and II) gives new information on the chemical composition of minerals,
mineralogical composition of both ore and host rocks, as well as the variation within the individual ore bodies.
This sets a rm basis for the quantitative methods developed for routine analysis of modal mineralogy (Paper
III) and mineral textures (Paper IV). Also, this increases the understanding regarding the primary origin and
metamorphic evolution of the deposit, which is important since the origin of the apatite iron ore of the Kiruna
type is still controversial.
Based on the modal composition, preliminary geometallurgical (GEM) ore types were established for the
Malmberget ore body. Each of these GEM-types describes quantitatively: the minerals present, their chemical
composition, rules how to calculate the modal composition from routine chemical assays (element to mineral
conversion, EMC rules) and a textural archetype in a library of archetypes. Using these GEM-types it is possible
to calculate the modal mineralogy and the liberation distribution for every geological unit from the sample
level to GEM-types to be further used in building a GEM block model of the ore.
The applicability of the geological model was tested by developing a liberation based process model of simple
one stage dry magnetic separation for the GEM-types. The model returns the metallurgical response, in terms
of grade and recovery, of each of the developed GEM-types.The model was validated with another ore sample
representing the same archetype from a different ore body and with a different grade.The model forecasted the
recovery and concentrate grade within 2%-unit accuracy.
This is the rst published study where a full predictive geometallurgical model is entirely based on the mineralogy. The approach is a generic approach and valid not only for iron ores but also for other metallic mines.

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

TABLE OF CONTENTS
Abstract

List of Papers

Additional publications

Preface

Acknowledgements

11

A general abbreviation

13

The authors contribution to the appended papers

1. Introduction
2. Hypothesis and objectives of the study
3. Literature survey
3.1. The geology and metallogeny of the northern Norrbotten area with an emphasis
of the occurrence of apatite iron ore
3.2. Malmberget iron operations
3.3. Why Geometallurgy
3.4. The mineralogical approach
3.5. The geometallurgical approach at LTU
4. Sampling, experiment and methods
4.1. Preliminary study
4.2. Field work, drill core logging and sampling
4.3. Sala Mrtsell separation
4.4. Sample preparation
4.5. Optical microscopy
4.6. Geochemistry
4.7. Electron microprobe (EPMA)
4.8. QEMSCAN (Quantitative Evaluation of Minerals by Scanning Electron Microscopy)
4.9. HSC 7.1
5. Results
5.1. Ore characterisation
5.1.1. Classification of the Malmberget iron bodies
5.1.2. Massive and semi-massive ore
5.2. Lithology of the Malmberget host rocks
5.3. Developing a framework and tools for a geological model
5.3.1. Components of the geological model
5.3.2. Developing a routine method for modal analysis
5.3.3. Geometallurgical ore type classification
5.3.4. Developing a method to quantify mineral textures
5.4. Framework for a metallurgical model
5.4.1. Comminution - particle breakage model
5.4.2. Concentration model dry magnetic separation
5.5. Proof of concept
5.6. A guideline on how to gather geological data for geometallurgical modelling
6. Discussion
6.1. The advantage of a strategic method for geometallurgical modelling in the Malmberget deposit
6.2. The geometallurgical model based on a mineralogical approach
6.3. Contribution to the genetic interpretation of the Malmberget iron ore deposit
7. Conclusion
8. Recommendation for future work
9. References

15
17
18
20

20
24
24
28
31
33
33
34
34
35
35
35
35
36
36
39
39
41
47
48
50
51
52
57
58
71
71
72
76
78
81
81
82
84
87
89
91

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

LIST OF PAPERS
This doctoral thesis Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit
for a geometallurgical model consists of the following four manuscripts and is hereafter referred to by their
roman numeral.
I

Martinsson, O., Lund, C., Andersson, J., Debras, C. Character and origin of the host rock to the
Malmberget apatite iron ore, northern Sweden

II

Lund, C., Martinsson, O Andersen, J. C. . Origin and evolution of the metamorphosed

Malmberget apatite iron ore, northern Sweden

III

Lund, C., Lamberg, P., Lindberg, T. Practical way to quantify minerals from chemical assays at
Malmberget iron ore operations an important tool for the geometallurgical program

IV

Lund, C., Lamberg, P., Lindberg T. Incorporating ore and mineral textures in a geometallurgical
model of Malmberget iron ore

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

ADDITIONAL PUBLICATIONS
The following abstracts, papers and the technical report are submitted or have been published in different
conference proceedings but are not included in the doctoral thesis.
Lund, C., Martinsson, O. A characterising of the ore minerals due to the mineralogy, chemical and
textural

properties in Malmberget, Conference in Minerals Engineering, Lule, Sweden, Feb-

ruary 2008, pp 71-80 [Extended abstract]

Lund, C. and Martinsson, O. Trace-element chemistry of magnetite from the Malmberget
apatite-iron ore, Conference, The 33rd International Geological Congress Oslo, Norway 2008
[Abstract]
Lund, C., Mineralogical, chemical and textural properties of the Malmberget iron deposit
A process mineralogically characterisation. Licentiate thesis, Lule University of Technology, Lule,
Sweden, 2009, 100 pp [Thesis]
Oghazi, P., Lund, C., Plsson, B., Martinsson, O., 2010. Applying traceability in a Mine-to-Mill
context by using Particle Texture Analysis, Preprint 10-002 SME Annual Meeting, Phoenix, AZ
p.7-11. [Peer-review Paper]
Lund, C. and Martinsson, O., 2010. Magnetitkemiska analyser med elektronmikrosondteknik
(EPMA) frn Gruvbergets apatitjrnmalm. LKAB [internal report]
Lund, C., Lindberg T., Martinsson, O. Mineralogical-textural characterisation of different
apatite-iron ore bodies, Malmberget deposit, Sweden, treated in a sorting process in laboratory
scale. Conference, Process Mineralogy 10, Cape Town, South Africa, 10-12 November 2010 [Paper]
Lamberg, P. and Lund, C. Taking Liberation Information into a Geometallurgical Model Case
Study Malmberget, Northern Sweden Conference, Process Mineralogy 12, Cape Town, South
Africa, 7-9 November 2012 [Paper]
Lund, C. and Martinsson, O. Oxide mineralogy and magnetite chemistry of the Malmberget
apatite iron ore, Northern Norrbotten, Sweden, Proceeding of the 12th SGA Biennial Meeting,
Uppsala, Sweden, August, 2013 [Extended Abstract]
Lamberg, P., Rosenkranz, J., Wanhainen, C., Lund, C., Minz, F., Mwanga, A., Parian, M. Building a
Geometallurgical Model Using Mineralogical Approach with Liberation Data. The Second AusIMM International Geometallurgy Conference, 30 September 2 October 2013, in Brisbane,
Australia [Paper]
Lund, C., Rollinson, G. K., Martinsson, O. A quantitative mineralogical characterisation of the
Malmberget apatite-iron deposit, Sweden, using QEMSCAN (Preliminary study)

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

PREFACE
This part of my life has been a long living journey from the beginning of 2002 until today, summer 2013. It
started in February 2002 when I was employed by LKAB, Research and Development division in Malmberget. The rst acquaintance to the mining industry of iron ore was formed. I was educated about their mineral
processes and got the knowledge about the important inuence as the mineral properties (modal composition,
mineral associations and mineral textures) forms in the process. In 2003 I changed position and started as a
mine geologist at the mining department in Malmberget.This gave me the opportunity to get familiar with the
mining process and especially get a deeper understanding of the varying geological composition (mineralogy,
chemistry and textures) of the different ore bodies. During this time I participated at work shops, conferences
and excursions and got to know people from the Lule University of Technology who were working with ore
geology research in the Norrbotten and the Skellefte district.
I ended my employment in the second half of 2005 to work in another mining company, at the same time a
research project had been approved with nancial support from the Foundation of Hjalmar Lundbohm Research Centre (HLRC), collaboration between LKAB and Lule University of Technology. The aim of the
project was mainly the same questions and objectivities as I had worked within the LKAB. I thought that this
opportunity of being part of an interesting project combining the best of two disciplines would give me the
chance to understand the close connection between the mineral properties of an ore and its behaviour in the
process treatment.
In spring 2006 I started as a PhD student in Lule. The project has been challenging in many ways. Since this
research area did not follow a tested approach, the pleasure of exploring something new has been the true
motivation but the moments of difculties has also been present in many ways.
The close collaboration with Therese Lindberg, LKAB has been invaluable for the project. On short notice I
have had access to material, analyses, data and maybe the most important, always the opportunity for frequent
discussions. During this time, this project has also results in two master thesis, one in ore geology (Debras, 2010)
and one in geometallurgy (Koch, 2013).
Through the years, I have met people of a wonderful ability to share their time, knowledge and interesting in
the project. I am very pleased to all these people and grateful to get the occasion to nally share the results of
their contribution in this doctoral thesis.
This PhD project has occupied my thought days and periodically even nights. I am truly satised that it is now
nally ending but my journey continues with new adventures...
Cecilia Lund
June, 2013
Lule, Sweden

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

10

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

ACKNOWLEDGEMENTS
First of all I would like to thank my supervisors for their support. Im deeply grateful for your time and interest
in the project. Dr Olof Martinsson, your knowledge and skills in the geology of Norrbotten have been crucial
especially as the Malmberget ore geology is a tricky piece of work.
Professor Pertti Lamberg, your excellence in geometallurgy and guidance in this project made it possible for
me to reach the aim. Working with you is a true pleasure every day.
This is not something I could have done without all the support from professionals, friends and family. I am
especially thinking of all people of professions around the world that in one or another way contribute with
their skills and knowledge.
This project is nancial supported by the Hjalmar Lundbohm Research Centre (HLRC), which I am truly
grateful for.
A special thanks to all people at the LKAB, you have made it possible for me to get a true picture of the production, often only a phone call away. Agneta Nordmark and Joel Andersson, mining department, thank you for
all the data and support, it has been invaluable for me. Bengt Orrmalm, Eva Landstrm, Rune Aalatalo and Kjell
Isaksson, I really appreciated your support during the rst part of the project when all the drill core logging
and data gathering were performed, thank you. ke Sundvall, Eva berg, Magnus Stafstedt, Krister Taavoniku,
Kirsten Holme, you are all acknowledge for your support.
Therese Lindberg, thank you, you are fantastic.
Furthermore I would like to thank my colleagues at the department for a nice time working together. I will
give an extra thanks to Christina Wanhainen for your scientic guidance and support. Riia Chmielowski,
thank you for endless English grammatical correction and for proofreading the manuscript. Pejman Oghazi,
thank you for good collaboration. Milan Vnuk, thank you for digitizing my picture, the layout and technical
preparation of this thesis.
I would also thank all my good friends, who create a warm, sociable atmosphere with a lot of activities, ranging
from permissive company to outdoor expeditions, you are all warmly appreciated for just be there.
Finally, my dearest family who all are concerned of my happiness and well-being, which all of you have made
it possible for me to nish my PhD. My mother Henny, you have taken care of Torkel and the rest of your
grandchildren when we all are too occupied, my father Kjell-Arne for drawing all sketches and gures in the
thesis. Thank you for being there and believe in me. My brothers, Hkan, without the statistical discussion, no
geometallurgical model, Henrik for taking care of faulty cars to leaky faucets, thank you for all your support.
Wictoria and Torkel you are my everyday happiness. I will always love you from the bottom of my heart.

11

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

12

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

A GENERAL ABBREVIATION
Type of deposits
BIF

Banded Iron Formation

AIO Apatite iron ore

IOCG Iron Oxide Copper Gold

Ore bodies/deposits
Fa

Fabian ore body

Vr
Pz

ViRi (Vitfors-Riddarstolpe) ore body

Printzskld ore body

Hn

Hens ore body

Vlkomma ore body

Grb
Ekb

Gruvberget ore deposit

Ekstrmsberg ore deposit

KUJ

Kiirunavaara ore body

PG

Per Geijer iron ores

Minerals
Mgt
Hem

Magnetite
Hematite

Ilm
Fsp

Ilmenite
Feldspar

Amph Amphibole
Px
Pyroxene
Qtz
Bt
Py
Ap

Quartz
Biotite
Pyrite
Apatite

EMC Element to mineral conversion

GEM-type
Geometallurgical ore type
AI
Association Index
LKAB Luossavaara-Kiirunavaara AB

13

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

14

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The authors contribution to the appended papers

The author participated in the geological eldwork and the sample preparation, as well as contributed

II

in the interpretations and evaluation of the data.

The geological mapping and logging, as well as the sample preparation were performed by the author.
The author made the analytical work (EPMA) together with an instrument operator and did the
interpretations and almost all writing. The other co-authors contributed in a supervisory capacity.

III

The author performed the geological mapping and logging, the experimental part, the sample
preparation, did the analytical work (QEMSCAN), developed the calculation routines (HSC
Chemistry), as well as did the interpretations and almost all writing. The other co-authors contributed
in a supervisory capacity.

IV

The author performed the geological mapping and logging, the experimental part, the sample
preparation, as well as did the analytical work (QEMSCAN), interpretations and major part of the
writing; Pertti Lamberg did the simulation work (HSC Chemistry) and some part of the writing.

15

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

16

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

1. INTRODUCTION
The northern Norrbotten ore province, located in the northern part of Sweden, is an important mining
area with several different economic deposits. These include i) epigenetic Cu-Au ores (e.g. Pahtohavare,
Aitik), ii) stratiform Cu-Fe deposits (e.g. Viscaria), and iii) iron deposits; BIF (e.g. Tornefors), skarn iron ore
(e.g. Sahavaara) and apatite iron ores of the Kiruna type (e.g. Kiirunavaara, Malmberget, Gruvberget) (Fig 1).
Some of them are in production, while others either have been mined out or mothballed, pending an economic viability to be mined, which also is the case for the sub economic mineralisations.
The two large underground mines in Kiruna and Malmberget and the open pit mine at Gruvberget are
operated by the LKAB (Luossavaara-Kiirunavaara AB), which is the most important producer of iron ore in
Europe. Over 2000 million tons (Mt) of ore has been mined in these deposits and they still have reserves of
1010 Mt @ ~ 46 % Fe (LKAB, 2011; Martinsson et al., 2012).
Large and high grade deposit like the Kiirunavaara with rather simple mineralogy, e.g. magnetite-apatite
(Nordstrand, 2012), is generally favourable for the production of high quality iron products and therefore
there is no acute need for increasing the understanding of geology and mineralogy for processing of the ore.
However, there are a number of iron deposits in the Norrbotten area which show lower grades, large geological variations within the ore and have much more challenging mineralogy for the production of saleable
iron concentrate. An example of such is Hannukainen containing, e.g. magnetite with associating chalcopyrite-pyritepyrrhotite (Hiltunen, 1982; Arvidson, 2013).
Geometallurgy takes both the geological and metallurgical information to create a spatially-based (3D)
predictive model for a mineral process (Lamberg, 2011). As the geometallurgical programs, which are the
industrial application, are needed for better resource control and to lower the risk in the operation related to
variation within the ore deposit, the geometallurgical concept spans the gap from ore characterisation to the
economic optimisation as mining management (Dunham et al., 2011).
Today, only few mines use geometallurgical programs in their production management, but this concept
will become more common in the future due to requirements of a more effective utilisation of the existing
ore resources such as mining larger volumes of lower grade ore, and development of ore characterisation
techniques. Similar conditions apply for new potential deposits as these also are most probably composed of
more complex ores.
This case study describes how to establish a geometallurgical model using the Malmberget iron ore deposit as a
case study. The mineralogical approach (Lamberg, 2011) was selected meaning that the focus is on mineralogy,
and therefore parameters like modal mineralogy, mineral textures, mineral associations, mineral grain sizes and
their relations to liberation characteristics are important.The main effort is to deliver a geological model which
gives quantitative rather than descriptive information to be used in a process sub-model.
The aim is to develop a geological sub-model of the geometallurgical system using a generic approach valid
not only for iron ores but also for other metallic mines.

17

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Karesuando

IN

68o

Soppero

Sweden

64o

Viscaria

5 Kiirunavaara
1

60o

Pahtohavare
Svappavaara

9LWWDQJL

3 Tornefors
Stora
Sahavaara

4
5
Aitik
Malmberget

30 km

Pajala

Trend

Paleoproterozoic intrusive rocks

Porphyrite Group

High grade gneisses

Hauki and Maattavaara quartzites

Pahakurkkio Group

Archaean basement

Kiruna Porphyries

Greenstone Group

Deposits

Fig. 1 Simplied geological map of the northern Norrbotten ore province with the economic deposits marked in red
modied from (Martinsson et al., 2004) 1) Epigenetic Cu-Au 2) Stratiform Cu deposit 3) BIF 4) Skarn iron ore 5)
Apatite iron ores.

2. HYPOTHESIS AND OBJECTIVES OF THE STUDY

The cross-discipline approach called geometallurgy connects two different but closely related areas in the mining industry, namely geology and mineral processing. It involves understanding and measurement of the ore
properties signicant for its successful processing.
Traditional characterisation tools used by geologists are either descriptive or do not measure the metallurgical
response. On the other hand most of the ore properties meaningful in minerals processing are directly related
to mineralogy. Therefore, the hypothesis of the thesis is:
The geological information collected traditionally by geologists is not enough to establish a geological model in a geometallurgical context; however, this can be fully done based on a proper mineralogical characterisation.
In order to answer the hypothesis, Malmberget was selected for the case study target and two main objectives
were set:

18

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Objective 1
Identify and characterise the mineralogical, chemical and textural properties of the Malmberget iron ore deposit important for a geometallurgical model.
The rst objective is ore deposit characterisation. The most known geological descriptions of this deposit are
from Geijer (1930); more recent published papers are rather rare. The necessity of correct and complete geological knowledge is always crucial since it has a strong impact on both the mining production and mineral
processes. In this thesis modern and updated geological descriptions are given on the ore and the host rock. It
expresses the varying characteristics of lithology, mineral composition, as well as mineral chemical and textural
properties.
This work will increase the understanding on the nature and important characteristics of the ore bodies. Furthermore, it will give new information to understand the origin and evolution, both primary and metamorphic, especially since the origin of the apatite iron ore of the Kiruna type is still controversial.

Objective 2
Establish a geological model consisting of modal mineralogy and mineral textures of the Malmberget iron ore deposit.
The second objective is to develop a geometallurgical model based on a mineralogical approach. Initially when
this project started, this was an undeveloped area, and the process mineralogical knowledge of the Malmberget
ore was rather limited. Since then, the understanding on what the varying geological characteristics within the
ores can inuence in the mineral processes has increased. As a geometallurgical model is based on a proper ore
characterisation, the results from objective 1 have set the directive to formulate objective 2. A preliminary study
was outlined to test what parameters were feasible to measure and which of those to be representative for the
ore deposit (see section 4.1). In this thesis the geological model is based on the mineral information such as
mineral grades and textures instead of the traditionally elemental grades.
Objectives 1 and 2 give the conditions to reach the main aim, to develop a framework for a mineralogy-based
geometallurgical program in the Malmberget iron ore deposit. The geometallurgical model must predict the
metallurgical response; otherwise, it cannot be used as a management tool in the production.

19

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Project

Geometallurgical model of Malmberget iron ore deposit

Geological model

Fa, Pz,Vr, Hn,V

Fa, Pz

Objectivity

Metallurgical sampling
Sample preparation
Comparison of Element
to mineral conversion
and QEMSCAN

Metallurgical sampling
Sample preparation
Mrtsell separation

Experimental

Field work
Drill core logging

Field work
Drill core logging

Mineralogical sampling

Mineralogical sampling

Optical microscopy
EPMA

Optical microscopy
Geochemistry
-ICP-MS
-ICP-AES

QEMSCAN

Geochemistry
-ICP-OES
-ICP-MS

Modal mineralogy

Optical microscopy
EPMA
QEMSCAN
XRF
Satmagan,
Element to mineral
conversion technique

Optical microscopy
Particle Tracking

Method

Mineral textures

Mineral chemistry
of the ore minerals

Geological description
of the host rock

Fa, Pz

Ore

Fa, Pz,Vr, Ba, Ma, Js, De,V

Subject

Ore deposit characterisation

Textural archetypes

Fig. 2 A ow chart showing the outline of this doctoral thesis as the objectivities, ores, experiment and methods.

3. LITERATURE SURVEY
3.1. The geology and metallogeny of the northern Norrbotten area with an emphasis
of the occurrence of apatite iron ore
The apatite iron ore type exists in a few different places around the world (Fig. 3). They were rst described in
Kiirunavaara by Geijer (1910), and later he declared that other deposits with similar geological features would
be named the Kiruna type of ore. They have the following characteristic features (Table 1):
APATITE IRON DEPOSITS

Kiruna district
Great Bear
magmatic Zone

Ber
Bergs
Berg
B
ergslagen
ergslag
erg
e
gslagen
g
sla
slage
slagen
a

SE Missouri

Avnik Area
Bafq provice

Cerro
erro de
e Mercado
Mercad
Me
erc
ercad
rcca
ado

El Laco

Fig. 3 Apatite-iron ore deposits around the world.

20

Gushan

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The age of the deposits of this kind ranges between Paleoproterozoic (e.g. Kiirunavaara) and Pliocene (e.g.
El Laco), and these are located in intracratonic settings or subduction zones (Frietsch and Perdahl, 1995). The
host rocks are calc-alkaline to alkaline volcanics with an andesitic to rhyolitic composition (Edfelt, 2007 and
references therein), and the ore mineralogy is dominated by magnetite-hematite with apatite and minor gangue
silicate minerals like actinolite and diopside.The alteration is comprised of albitisation, silicication, seritisation,
epidotisation, and minor alteration minerals are actinolite, scapolite, tourmaline, biotite and carbonates (Edfelt,
2007 and references therein).
Table 1 General descriptions of apatite-iron deposits (Edfelt, 2007 and references therein).
Main features

Apatite iron ore (Kiruna type)

Tectonic setting

Regional fault zones of intracratonic settings or subduction zones

Age

Paleoproterozoic to Pliocene-Pleistocene

Host rock

Calc-alkaline to alkaline volcanic rocks (intermediate to felsic)

Mineralogy

Magnetite-hematite and apatite-actinolite-diopside

Alteration

Albitisation, silicication, seritisation, epidotisation and minor carbonate

Ore genesis

Magmatic intrusive-extrusive and/or hydrothermal replacement

The apatite-iron ores have been the focus of genetic discussions for over 100 years. The magmatic model has
been suggested by Geijer (1910), Park (1961), Frietsch (1984), Nystrm and Henrques (1994) and Henrques
et al. (2003). The hydrothermal model is favoured by Park (1973), Hildebrand (1986), Hitzman et al. (1992)
and Sillitoe and Burrows (2002). As it is difcult to explain all of the features of the apatite iron ores with a
single genetic model, a combination of these two models, a magmatic-hydrothermal process, was suggested by
Martinsson (2004) regarding the formation of the Kiruna and Malmberget deposits at 1.89-1.88 Ga (Cliff et
al., 1990; Romer et al., 1994). However, there are major differences between these two deposits, due to later
overprinting by metamorphism, deformation and granitic intrusions (Bergman et al., 2001).
The geology of the ore province northern Norrbotten is characterised by bedrock sequences from Archean to
Proterozoic. The Archaean granitoid-gneiss basement is the lowest unit with an age of 2.8 Ga. Unconformable
overlaying sequences are supracrustal successions of Palaeoproterozoic age 2.5-2.0 Ga, followed by Svecofennian volcanic and sedimentary units dated c. 1.9 Ga. Forty apatite-iron deposits are known from the northern
Norrbotten ore province, and they are hosted by and probably also genetically related to the volcanic rocks in
the Svecofennian succession (Martinsson, 2004). The early orogenic Haparanda Suite predates the 1.86-1.88
Ga Perthite Monzonite Suite and was affected by deformation and metamorphism at 1.88 Ga. At 1.81-1.78
Ga a second deformation and metamorphic event occurred by the intrusion of the Lina granite-pegmatite
association close to the Malmberget apatite iron ore (Martinsson, 2004).
In the Malmberget ore eld, more than 20 different tabular to stock shaped ore bodies are known occupying
an area of 2.5 x 5 km2 (Fig 4). The Malmberget deposit was originally probably a more or less continuous ore
lens which experienced at least two phases of folding and metamorphism. By strong ductile deformation it was
torn into several lenses that today occupy a large-scale fold structure where the individual ore bodies stretch
parallel to the fold axis, which plunge 40-50 toward the SSW (Bergman et al., 2001).The apatite-iron ore was
affected by metamorphism under high temperature conditions during the emplacement of a large Lina granite
intrusion to the north at c. 1.8 Ga (Bergman et al., 2001) causing post ore alteration and recrystallisation at a
temperature of 650-700 C and a pressure of 2-4 kbar (Annersten, 1968). Stilbite-bearing mineral assemblages
indicate that the area was not heated above 150 C after 1.62-1.60 Ga (Romer, 1996).
21

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Tingvallskulle

N Alliansen
Vitfors

Josefina
Upland
Skne

Vlkomman

Printzskld 895 m

Hens
Baron

Dennewitz

Selet

Johannes

Koskullskulle

.Johannes
Viri
930 m

Kapten
Kapten
830 m

Fabian
830 m

Malmberget
0

1 km

Granite-pegmatite

Skarn and ore-breccia

Sillimanite gneiss
Metavolcanics rocks
(leptite)

Hematite ore
Magnetite ore

Fig. 4 Simplied geological map over the Malmberget iron ore deposit; the western eld is an almost continuous ore
horizon (Vlkomma to Tingvallskulle) and the eastern eld as individual ores (Kapten to Alliansen), modied from
(Bergman et al., 2001).

The ore bodies are recrystallised, coarse grained and elongated in the direction of the lineation of the rocks
(Martinsson and Virkkunen, 2004). The host rocks to the ore consist of felsic to mac, metavolcanic rocks that
locally contain sillimanite (Geijer, 1930). Especially in the footwall of the western part of the deposit, there exists gneiss with sillimanite, muscovite and quartz, while corundum and andalusite occasionally are found in the
ore (Geijer, 1930). Mac rocks are typically rich in biotite and can be found close to the ore as conformable to
discordant lenses. Some of them are probably metamorphosed dikes, while others seem to have formed as sills
or extrusions (Geijer, 1930).These metavolcanic rocks are traditionally called leptites and gneisses (Geijer, 1930;
Martinsson and Virkkunen, 2004) and are mainly distinguished from each other by the grade of metamorphism
and a coarser feldspar grain size of the latter (Geijer, 1930). These rocks may have an origin similar to those of
trachyandesitic to rhyodacitic composition hosting the Kiirunavaara deposit (Martinsson and Hansson, 2004).
The Malmberget deposit is divided into an eastern and a western part which include more than twenty known
ore bodies ranging in size and grade. Some of them are in production; others have been mined out or are in an
exploration stage. The individual ore bodies are described based on the location in the eastern or the western
part of the deposit (Fig. 5).
The Fabian (Fa) and ViRi (Vr) ore bodies are a part of the eastern eld of the Malmberget deposit and include
a number of more or less individual magnetite ore bodies characterised by a high magnetite grade and a low
apatite grade (Martinsson and Virkkunen, 2004). The Fabian ore body is a large irregularly shaped magnetite ore
lens, striking NE-SW and dipping from 75, SSE to almost vertical; an average content of the ore is 55.6% Fe
and 0.35% P.ViRi is a large tabular shaped magnetite ore body, striking E-W and dipping 60 to the SW with an
average content of 61.5% Fe and 0.15% P.The ore contacts shift from sharp to graded, generating in the latter case
a breccia style mineralisation where the wall rock also occurs as irregular inll in the massive ore.

22

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Fig. 5 A schematic 3D-picture illustrating the ore bodies in the Malmberget iron ore deposit (LKAB).

The Hens (Hn) and Vlkomma (V) ore bodies are located in the western part of the deposit and form part of
an almost continuous ore horizon, 5 km in length and consisting of both hematite and magnetite ore (Bergman
et al., 2001). The Hens ore has a tabular shape, striking ENE and dips 70 towards the SSE and is composed of
separate magnetite and hematite ore lenses with an average grade of 53.5 % Fe and 0.80 % P.The Vlkomma ore
body strikes SE and dips 55 towards the SW. A typical feature is coarse grained apatite, occurring as a regular
banding or disseminated. The Vlkomma ore body shows an almost similar mineralogical character as the Hens
ore body, but magnetite and hematite occur mostly together, often with porphyroblasts of one of the minerals
in a matrix of the other (Geijer, 1930).
The Printzskld (Pz) ore body is situated between the eastern and western parts of the Malmberget deposit.
This magnetite ore body has a lenticular shape, strikes ENE and dips 70 to the SSE and is partly rich in apatite and biotite giving the ore a banded structure. The grade is 50.5% Fe and 0.75% P. Smaller hematite bodies
occur close to the magnetite body, and minor amounts of hematite are also present in the magnetite ore.
In the Svappavaara ore eld, the Gruvberget (Grb) apatite iron ore is located; it strikes in the N-S direction,
dipping 50 to 75 to the E and follows the western limb of a syncline (Frietsch, 1966) (Fig. 1). The ore body
is disc shaped and is generally wider in the northern part. It is approximately 1300 m long in the N-S direction
and has an average width of 30 m in the E-W direction, being partly fractionated by faults. Four ore types
were recognised and distinguished by Frietsch (1966) including magnetite ore, hematite ore, hematite breccia
and weathered hematite. Hematite ore is the main part of the ore body, and the magnetite ore is located at the
northern part of the ore deposit.The proven ore reserve is estimated at being 7 Mt of magnetite ore at 53.1% Fe
(LKAB, 2013). The magnetite ore is enriched in calcium and phosphorous, with a mean value of 1.58% P2O5
and 5.75% CaO (Adolfsson and Fredriksson, 2011). The mineralogy reveals an irregular distribution of apatite
veins together with calcite, epidote and amphibole (Frietsch, 1966; Lindskog, 2001).

23

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

3.2. Malmberget iron operations

Ten deposits of apatite-iron ore have been mined in northern Sweden and the largest ones, Kiirunavaara and
Malmberget, are still mined by LKAB contributing about 90% of the iron ore production in Europe. The annual production in Malmberget is about 14 Mt of ore at 43.1% Fe and in Kiirunavaara 22.7 Mt at 42.8% Fe
(LKAB, 2011).
The Malmberget concentrator has two different lines, one for magnetite dominated ore and the other for
hematite dominated ore. The magnetite line in Malmberget consists of dry and wet concentration sections.
The dry process stage uses screens, cone crushers, autogeneous crushers (VSI) and low-intensity magnetic
drum separators producing two different pre-concentrates: a Sinter Feed (FAR) with a higher iron grade and
a slightly lower graded Pellet Feed (PAR) , both having a particle size of <10 mm . In the wet concentration
section, the Sinter Feed is processed further by screening to <2 mm followed by wet low intensity magnetic
separation (WLIMS). The Pellet Feed is further upgraded in three steps of ball mill grinding and WLIMS. In
the nal grinding step, the concentrate, i.e. Pellet Feed, is ground to approx. 68% <45 m to reach the optimum
size distribution for downstream processing (Kvarnstrm and Oghazi, 2008). The nal concentrates have high
requirements for purity, e.g. the Sinter Feed (MAF) has maximum limits of 0.8% for SiO2, 0.45% for TiO2 and
0.025% for P (LKAB, 2011).

3.3. Why Geometallurgy

The northern Norrbotten ore province is an intensely mineralised area and has traditionally been a very important mining district. It contains different types of mineral deposits (e.g. Fe, Cu, Au and Ag) ranging from
small and uneconomic up to world-class ore deposits (Martinsson, 2004). The increasing demand of raw material has resulted in iron ore being one of the most important commodities in the global market. The area is
actively exploited and explored by large and junior mine companies.
Different types of the Fe-oxide mineralisation occur in the Norrbotten area showing a quite large variation in
its composition. They can be classied by mineralisation type (massive, breccia, disseminates or veins), the main
mineralogy (e.g. magnetitehematite-apatite, magnetite-chalcopyrite-pyritepyrrhotite) or by genesis. Even
within a single ore type, (e.g. apatite iron ore) the character of the ore and its properties (mineralogy, grain
morphology, associating host rock) show a large variation (Table 2.)
Economically the most important deposits in the area have been the iron ores of Kiirunavaara and Malmberget. The global production of crude steel (1550 Mt) and the export of iron ores (1000 Mt) have almost been
doubled since the beginning of the 21st century (LKAB, 2013 and references therein) giving a competitive
market. LKAB plans to increase the production by the opening of new open pit mines in Gruvberget (already
in production), Mertainen, and Leveniemi, located in the Svappavaara ore eld (LKAB, 2013). By 2015 LKAB
aims for 37 Mt of nished iron ore products (LKAB, 2013). The strategy of LKAB is to provide high quality
products for the customers needs. As the ore quality varies both between and within the deposits, there is a
need to have good control over the plant feed and the renement.
Comparing the grades and tonnages between the different deposits in the Norrbotten area, it is apparent that
the future mines will be smaller and/or lower in the head grade (Fig. 6).

24

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 2. Summary of the characteristic properties of different types of iron ores in the Norrbotten area.
Ore eld

Ore
Type

Deposit

Pajala

BIF

Tornefors

Pajala
Kolari
(Finland)

Skarn
rich
IO
Skarn
rich
IO

Character
Lens,
banded

8 Mt 25 % Fe,
1-2 % S

Ore & gangue

minerals
Mag Amph+Px+
Grt

Stora Sahavaara

Lens,
banded

82Mt 41 % Fe,
2.5 % S 0.08%Cu

Hannukainen

Lenses

203Mt 33%
Fe, 0.16%Cu,
0.06%Au
>2000 Mt>60%
Fe, 0.02-2% P
20 Mt >60% Fe,
0.05-1.6% P
20 Mt 33% Fe,
3.5% P
~100 Mt 40-45%
Fe, 4-5% P
8 Mt 50% Fe,
2.5% P

Kiruna

AIO

Kiirunavaara

Kiruna

AIO

Luossavaara

AIO

Rektorn

Massive

AIO

Lappmalmen

Massive

AIO

Haukivaara

Massive

Kiruna

AIO

Ekstrmsberg

Massive

Svappavaara

AIO

Gruvberget

Massive
(Breccia)

Svappavaara

AIO

Mertainen

Breccia

Svappavaara

AIO

Leveniemi

Massive,
Breccia

Malmberget

AIO

Malmberget

Massive
(Breccia)

Kiruna
(Per Geijer)
Kiruna
(Per Geijer)
Kiruna
(Per Geijer)

Massive
(Breccia)
Massive
(Breccia)

Grade & Size

Host rock

Alteration

Mac tuff-tufte

Bt

Mag+Ccp+Py+Po

Black schist
mac tufte

Bt

Mag+Ccp+Py+PoAu
Di+Ab+Bt+Hbl

Diopside skarn

Amph, Px

Trachyandesite,
Rhyodacite
Trachyandesite,
Rhyodacite
Rhyodacite,
Ryolite
Rhyodacite,
Ryolite

Amph, Ab,
Bt
Amph, Ab,
Bt
Kfs, Qtz, Ser,
Chl, Bt
Kfs, Qtz, Ser,
Chl, Bt
Kfs, Qtz, Ser,
Chl, Bt

Mag Ap+Amph
Mag Ap+Amph
Hem+Mag
Ap+Qtz+Carb
Hem+Mag Ap
Hem Ap

Rhyodacite

37 Mt 55-60%
Fe, 1.3% P

Mag+Hem
Ap+Amph
Qtz+Ser

Rhyodacite

64 Mt 50-67%
Fe, 1% P

Mag+Hem Ap

Metaandesite

166 Mt 35% Fe,

0.05% P
204 Mt 64% Fe,
104 Mt 26% Fe
0.05-0.8% P

Mag Amph+Cal+Scp

Trachyandesite

Mag+Hem
Ap+Amph+Cal

Metaandesite

Ab, Bt, Ms
Scp, Amph

>900 Mt 51-61%
Fe <0.8% P

Mag+Hem
Ap+Amph+Bt

Andesite-Basalt,
Dacite-Andesite,
Rhyodacite

Ab, Kfs,
Amph, Bt,
Scp

Mar, Ser,
Amph
Ab, Kfs, Scp,
Ep, Amph,
Grt, Px
Ab, Amph,
Scp

Ore Reserve; Kiirunavaara 682 Mt 47.5 %Fe, Malmberget 271 Mt 41.8 %Fe, Gruvberget 7 Mt 53.1% Fe (LKAB 2013) References;
Grip and Frietsch (1973); Frietsch (1966, 1974); Lundborg and Smellie (1979); Lindskog (2001); Bergman et al. (2001)

10000
Economic status
In production
Feasibilty stage
Exploration stage
Closed mine/Feasibility stage
Closed mine
Subeconomic potential

Size (Mt)

1000

Kiirunavaara
Malmberget

Hannukainen
Mertainen

100

Leveniemi
Lappmalmen

Stora Sahavaara
Masugnbyn
Vathanvaara
Vittangi
Rektorn

10

Tornefors

Gruvberget
Ekstrmsberg
Tuolluvaara
Luossavaara

Haukivaara

1
20

30

40

50

60

70

Grade (Fe% )
Fig. 6 A grade and tonnage relationship of iron ore deposits in the Norrbotten area. Lines show equal metal content (Fe
vs. Mt).

25

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

A brief review of the mining history shows the development in mining techniques and the iron ore products
during the years of production. The Malmberget deposit has been known since 1735, and there were many
mining pioneers during history that attempted to make the utilisation of the deposit protable. A problem that
was faced often was the transportation of ore across the Norrbotten wilderness. The invention of the Thomas process to produce good quality steel from high phosphorous ore in 1878 made the giant ore reserves in
Kiruna and Malmberget very attractive, and LKAB (Luossavaara-Kiirunavaara AB) was established at 1890 to
exploit these deposits. In order to obtain protability, a railway was discussed for a lengthy time, and in 1888,
after a period of four years, a railway was completed between Malmberget and Lule. Initially, the Malmberget
ore was pit-mined, but since the mid-1920s, all production has been underground using large scale sub-level
caving (LKAB, 2006). In the beginning the high phosphorus ore was sold as a lump ore, but to manage the
competition on the market LKAB was forced to start qualitative work on upgraded products and to specialise
on the pellets products. The rst pelletising plant began to operate in Malmberget in 1955 (LKAB, 2006). In
the 1970s a new direct reduction (DR) steelmaking process was developed (e.g. MIDREX and HYL), which
gave the opportunity for LKAB to differentiate the products of pellets (LKAB, 2013). Since then the focus has
been on the concentration properties of the ore and pelletisation of the concentrate. Today, this is classied as
a high technology area (Forsmo, 2007), and the blast furnace pellets with the DR pellets comprise 84% of the
total iron ore products (tot 26.3 Mt; LKAB, 2013).
Since a couple of decades back, a cross discipline approach called geometallurgy has been under development.
It combines geology and mineral processing to create spatially-based predictive models for mineral processing
plants (Lamberg, 2011). Some claim that it is a young analytical eld, while others think that it is a sub-discipline included in traditional mineral processing (Dunham et al., 2011). The term geometallurgy appeared in
literature approximately 40 years ago and has received a wide range of denitions and perspectives since then.
Much of the variation can be attributed to the background of the person and the purposes of the perspective
(Jackson et al., 2011).
Geologists have a long tradition of describing the ore to enable a production quality as high as possible. They
have been focusing on ore properties which help in dening the ore boundaries and locating the extensions
of the mineralisation. For the plant they have provided information on coming mill feed in terms of grades,
tonnages and main ore types or lithologies. Metallurgists, or mineral processing engineers, normally have good
understanding of what the feed grade will mean for the processing properties. In general, high grades mean
easier processing properties and higher recoveries, while low grades mean the opposite. Lithology or ore type
information provided by geologists is often linked with metallurgical response, but it is common to nd that
variation within a lithology or ore type is reasonably large.The idea of geometallurgy was to improve the characterisation of ore and improve methods to measure important parameters. This information was to be used to
design mineral processes, as well as manage and optimise the production (Batterham et al., 1992; Lorenzen and
van Deventer, 1994; Petruk, 2000; Sutherland et al., 2000).
The last decade has had a rather fast evolution in the eld of geometallurgy, and one of the large contributors
is the development and increased availability of advanced measurements and analytical techniques (Gottlieb et
al., 2000; Petruk, 2000; Gu, 2003; Moen, 2006; Liipo, 2012). Another main reason is the increased demand of
raw materials and that high-grade, surface-near mineralisations are depleting. Instead, there is an opportunity
for more multifaceted mining, as deep-mining, recycling of previously uneconomic waste load or use selective
mining of low grade ore (Dunham et al., 2011). Increased environmental restrictions, such as special tailing

26

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

areas for sulphur bearing tails, increase the processing costs and increase the needs to know more about the
sulphide gangue minerals.The economic pressure on the mining companies is always high and they will always
face key questions of keeping a reasonable mine valuation and signicant ore variability (Bye, 2011).
Three different levels of the modern geometallurgical denition can be outlined. A rather narrow vision that
solely has a focus on the ore instead of the mineral characterisation (Schouwstra et al., 2010) and where the
spatial characterisation is restricted to an ore type or a representative sampling (Helle et al., 2005). Under this
context also follows the denition of the process mineralogy that seldom has a spatial characterisation but instead focuses on solving process mineralogical problems (Jackson et al., 2011).
A more open and probably the most common view considers the use of geometallurgical modelling or mapping which is a spatial characterisation (Jackson et al., 2011). The perspective is limited to the impact of the
ore body and its primary extraction factors. It includes the process parameters, such as valuable minerals and
metals for process design, mine design and scheduling to be implanted within the block model (Dobby et al.,
2004; Jackson et al., 2011).
The broadest view uses the term geometallurgical sustainability performance (Jackson et al. 2011). In 2011, a
discussion started in order to enlarge the established geometallurgical eld and incorporate more externalfactors that inuence the context of geometallurgy in a global market perspective, such as the business dimension
(interpretation, analysis, evaluation and validation of all technical aspects), mine planning, risk management, sustainability (water, energy consumption and CO2 emission levels) and the geotechnical approach, e.g. identication of variable rock mass conditions (Vann et al., 2011; Dunham et al., 2011; Bye, 2011; Jackson et al., 2011).
Developing a valid industrial application, called a geometallurgical program, requires a number of important
and crucial steps which generally sums up the above mentioned criteria (modied after Dobby et al., 2004;
Bulled and McInnes, 2005; David, 2007; Lamberg, 2011).This means the following: 1) Collection of geological
data through drilling, drill core logging, measurements, chemical analyses and other analyses. 2) An ore sampling program for metallurgical testing where geological data is used in the identication of preferred locations
for the samples. 3) Laboratory testing of these samples in order to extract process model parameters (sometimes
called ore variability testing). 4) Checking the metallurgical validity of the geological ore-type denitions and,
where necessary, developing new ore-type denitions called geometallurgical domains. 5) Developing mathematical relationships for the estimation of important metallurgical parameters across the geological database. 6)
Developing a metallurgical model of the process. The model consists of unit operations which use the metallurgical parameters dened above. 7) Plant simulation using the metallurgical process model and the distributed
metallurgical parameters as the data set. 8) Calibration of the models via benchmarking for existing operations.
Basically two different approaches exist for linking the steps listed above. The rst one relies on geometallurgical testing, and the other approach is based on mineralogy (Lamberg, 2011). Geometallurgical tests are small
scale laboratory tests which aim to directly measure the metallurgical response of the samples. Examples of such
are the GeM Comminution Index test (JKTech, 2010), the JK Mineral Separability Indicator test (Bradshaw,
2010) and the Davis tube test (Niiranen and Bhm, 2012). The other approach relies on mineralogy (Lamberg,
2011). This means that both the ore and the process are quantitatively described on a mineralogical basis.
A common practice is to describe different geometallurgical domains.The domain is borrowed from geostatistics where it is dened as a closed volume where grades follow the same rule, i.e. show similar spatial continuity.
The establishing of a geostatistical domain is an iterative process where original geological information is used

27

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

to divide the ore body into several domains, and then their differences are compared through geostatistics. The
same is true for a geometallurgical domain; it is a closed area showing similar rules for the geometallurgical
properties, and domain denition is an iterative process (Lotter et al., 2011). To divide an ore body and use
geometallurgical units basically relies on a proper review of the geological data including the host rock,
alteration, grain size, texture, structural geology and grade (Lotter et al., 2011).
Estimation of the metal grade of the ore deposit and its variation goes through geostatistics and commonly this
is done by block modelling. This means that the ore body is divided into equal sized blocks, for example 40 m
x 40 m x 40 m and geostatistics is used to estimate the average grade and density of the block. Before the same
concept can be used for geometallurgical parameters, one needs to check that each parameter is additive, i.e. it is
possible to calculate an average for a group of samples. Elemental grades and mineral grades are additive, but for
example lithology is not. Elemental recoveries (e.g. copper recovery) sound like additives, but if two materials
with different recoveries are blended the resulting recovery is not necessarily a weighted average of these two.
The same is true for comminution properties. Comminution parameters are not necessarily additive or linear
in a correct statistical approach, e.g. textural variables (Dunham and Vann, 2007; Bye, 2011).
Geometallurgical programs of iron ores are very rare based on literature, but few examples can be found
(Niiranen and Bhm, 2012; Paine et al., 2012). A technique that is frequently used in evaluating the metallurgical variation in magnetite bearing iron ores is Davis tube testing (Farrell et al., 2011; Niiranen and Bhm, 2012).
It is a small scale physical separator that divides small samples into a magnetic concentrate and a non-magnetic
tail by using a strong magnetic eld. These products are chemically analysed, and the distribution of elements
is then calculated. The iron distribution (recovery) and the concentrate quality are used in predicting the
metallurgical response in a full scale operation.

3.4. The mineralogical approach

The mineral information forms a vital basis for designing, diagnosis and optimising the mineral process circuits
(Lamberg et al., 2013) but also when evaluating the ore, mapping the variation in the ore body and designing
the sampling for metallurgical testing. This requires a proper quantitative mineral characterisation of the ore.
Mineralogical information can be divided into different levels depending on the increasing complexity and
details (Lamberg et al., 2013). The lowest level is qualitative information. This is a traditional way of using
mineralogical information for identifying lithology and ore types. Quantitative information can be divided
into three dimensions.The one dimensional (1D) level gives the phases, their chemical composition and modal
composition. The 2D level gives the same information but on particle size class basis. The highest 3D level
describes mineralogical information on particles, i.e. liberation distribution basis. A necessary aspect to consider
when gathering the data (from samples of ore or process) is the selection and use of a technique that preferably
can handle a large amount of samples and data which should be fast, inexpensive and preferably at least partly
automated (Table 3).
Lithology or ore types have traditionally been determined by using optical microscopy and X-ray diffraction.
Often this involves optical microscopical study where proper lithological names are dened, and afterwards
identication in drill core logging goes through macroscopical identication with the help of chemical assays
and sometimes geophysical measurements. Examples of this are the ore types of Kiruna where identication is
based on chemical assays and geological logging.

28

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 3 A compilations of existing methods to determine the modal mineralogy in geological and process samples.
Technique

Advantage

Disadvantage

Methods

References

Optical mineralogy

Inexpensive, reliable

Tedious, small amount

samples, expertise

Optical microscopy

Henley 1992

Automated
mineralogy

Reliable, all types of ore,

complete mineralogical
information

Time consuming, sized

samples, expensive

QEMSCAN,
MLA, Inca Minerals, PTA

Sutherland and Gottlieb,

1991; Gu, 2003; Moen,
2006; Liipo et al., 2012

Quantitative
X-ray diffraction

Fast, inexpensive

High detection limits, not

for non-crystalline phases,
expertise

Rietveld

Rietveld, 1969, Paine

et al. 2011, Hestnes and
Srensen, 2012

Chemical mass
balancing

Fast, inexpensive, large

amount of samples, low
detection limits

Requires an extra analyses

method in complex ores

Element to mineral conversion

Braun, 1986; Paktunc,

1998; Whiten, 2008; Lamberg et al., 1997

Modal mineralogy, i.e. mineral grades, has traditionally been determined by optical microscopy with point
counting. It is time consuming (Petruk, 2000), and the quality is dependent on the skill of the mineralogist
(Henley, 1992). Image analysis with optical microscopy makes the method faster, but positive identication of
minerals is not always easy especially if a process sample is comprised of very ne grained particles. Recent
advances have improved the quality of quantitative X-ray diffraction (XRD), but the technique (Rietveld)
remains sensitive to preferred orientation, crystal size, crystallinity and variation in the chemical composition
of the several phases present in the mixture (Rietveld, 1969). Automated mineralogy, i.e. scanning electron
microscopy based image analysis, (QEMSCAN, MLA, PTA, IncaMineral) provides more rapid quantitative
analysis of mineral grades and textures (Jones and Gravilovic, 1970; Sutherland and Gottlieb, 1991; Gu, 2003;
Moen, 2006; Liipo et al., 2012).
Although automated mineralogy has not only largely replaced optical microscopy and quantitative XRD and
also greatly expanded the use of quantitative mineralogy, it still has some drawbacks which limit its usage. Firstly,
it is basically a standardless analysis method. Secondly, for accurate analysis from particulate materials, the sample
should have a narrow size distribution to avoid classication in sample preparation and to eliminate the need
for changing magnication during image collection (Kwitko-Ribeiro, 2011). This means that samples need
to be sized by sieving which is time consuming and also will multiply the number of samples to be analysed.
Thirdly, there are some limitations on the particle size. For sound statistics one needs to analyse thousands of
particles and, in coarse size fractions, multiple resin mounts will be required to contain the necessary numbers
of such relatively large particles. In very ne particle and mineral grain sizes (<5 microns), the resolution starts
to be a problem in mineral identication as the electron beam diameter approaches the size of individual grains.
Finally, minerals with very similar chemical compositions are difcult to identify in SEM based image analysis.
An example of such mineral pairs is magnetite and hematite (Fandrich et al., 2007; Gomes et al., 2012).
The use of multicomponent chemical mass balancing in solving mineral quantities from chemical analysis
(e.g. XRF) has been successfully used for decades by a number of researchers and laboratories (Braun, 1986;
Paktunc, 1998; Whiten, 2008; Lamberg et al., 1997; Lamberg, 2011). Normative calculations used by geologists
for classication purposes, for example CIPW (Cross, Iddings, Pirsson and Washington) (Cross et al., 1902),
differ from the mass balancing method by calculating hypothetical and not actual mineral composition for
samples, using stoichiometric compositions for minerals (e.g. CIPW uses volatile free minerals) and reporting
mineral grades as normative per cent, not as weight proportions. Therefore, for quantitative geometallurgical
purposes this method is not suitable.

29

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Another mass balancing method, element-to-mineral conversion (EMC), is a mathematical mass balancing
where the bulk elemental composition of a sample is transformed to mineral quantities using information
on the chemical composition of the actual observed minerals, and the mineral quantities are reported as
weight percentages enabling study of even the trace elemental distribution between the minerals. Even if the
element-to-mineral conversion technique has been known for decades, it is an underused technique;
according to the literature it is systematically applied in mining operations very rarely (e.g. Kemi chromite
mine, Lamberg, 2011).
The chemical composition of minerals is needed not only for element to mineral conversion but also in
evaluating the distribution of elements between minerals and for evaluating the chemical composition of
concentrate. In nickel sulphide ore, for example, nickel is distributed between several minerals, and for accurate
estimations one needs to know the chemical composition of each nickel carrying mineral. In zinc ores, on the
other hand, the chemical composition of sphalerite, the zinc mineral, can vary widely, and in some deposits it
is easy to produce a zinc concentrate grading 55% Zn, whereas in other deposits this is not possible at all due
to low Zn content of sphalerite. The chemical composition of minerals is currently determined by electron
microprobe (Wave Length Dispersive, WDS, spectrometer). The detection limit for an electron microprobe
is commonly 0.1 wt%, but using special conditions it is possible to extend down to 10 ppm (Kojonen and
Johansson, 1998). For gold and palladium even lower grades in minerals can be signicant, and LA-ICPMS is
then commonly used (Johansson and Wanhainen, 2010). The chemical composition of the ore minerals within
the ore bodies does not often vary widely; therefore, it is generally enough to run a mineralogical study for
representative samples in the early stage of the evaluation of the ore body. The information on the chemical
compositions is then used throughout the utilisation of the ore body. Sometimes it is possible to identify the
variation in chemical composition of minerals from bulk chemical analysis.
The modal mineralogy is not sufcient to solely answer the ore behaviour when processing the ore. The
mineral textures play a signicant role and need to be considered when a geometallurgical model is developed.
The texture characterisations are usually very subjective (Bonnici et al., 2008) and traditionally more related to
ore characterisation than process mineralogy (Perez-Barnuevo et al., 2012).
In a mineral process, the mineral textures and the liberation are closely associated. The texture determines
the limitation in an upgrading process (Butcher, 2010). The purpose of the comminution stage is to liberate
ore minerals appropriately for the concentration process to enable reaching the required concentrate quality
with an adequate recovery. Therefore, it is important that textural attributes such as grain size, grain shape and
mineral associations could be quantied (Bonnici et al., 2008) for modelling and simulating the processes.
There is a need to include the degree of liberation and the mineral associations in the geometallurgical model.
However, the degree of liberation is not enough to forecast the behaviour of a mineral in the process as different types of mixed particles are present and behave in a unique way in the process (Lamberg and Vienna, 2007).
Therefore, it is also crucial to understand how particles are generated. By using liberation models, the aim is to
forecast what kinds of particles are produced from an ore in the comminution (Gay, 2004). The comminution
models are well-known and commonly used (Wills, 1988) and it is possible to predict product size distribution
due to changes in feed size distribution, particle hardness, feed rate and design of mill (Gay, 2004). When an
appropriate liberation model is selected, consideration needs to be taken to the purposes of the fragmentation
of the unbroken ore since each one of the liberation models has limitations.

30

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The random liberation model generally uses an image of rock and applies grids on the image to produce
particles of different sizes based on the grain structure and fracture pattern (Gaudin, 1939).The model fails since
the natural breaking is generally preferential (Fandrich et al., 1997; King 1979; Stamboliadis 2008; Gay 2004).
The Kernel model uses the kernel function which represents the frequency of the progeny particles (particular
in size and composition) which are produced when the particle (particular in composition and size) is broken
in comminution.This model has the limitation of the assumption of a binary system that can be too simplied,
but in a multiphase case the mathematical solutions become very complicated (Andrews and Mika, 1975; King
and Schneider, 1998; Gay, 2004).
The third type of liberation model uses empirical liberation measurements. Hunt et al. (2009, 2011) and
Bonnici et al. (2008, 2009) simulated the fragmentation of particles by applying chessboard segmentation on
the particle map of one size fraction (0.6-1.18 mm). By using the random breakage for naturally broken particles, it gives more realistic estimates than the pure random breakage models.
The particle breakage model consists of a general rule or collections of rules which quantitatively describe
which types of particles are generated from the ore in comminution.These particles are possible to measure and
also use for simulation.The model generates a liberation distribution based on geological information from the
actual particles and not from statistic predictions (Lamberg, 2011).
New innovations and the development of characterisation technology have created new tools to measure
textures quantitatively. Previously, this was both time-consuming and expensive; additionally, the handling of
large amounts of image data was not practical. Today, optical image analysis offers tools to identify minerals and
measure semi-quantitatively mineralogical information (Bonifazi et al., 1990; Neumann and Schneider, 2001;
Donskoi et al., 2007). It has become more common and is necessary to link the quantitative mineral identication to the metallurgical outcome (Lane et al., 2008; Schouwstra et al., 2010; Philander and Rozendaal, 2011;
Bushell, 2012). Others have developed new methods to produce the mineralogical and textural data, rapidly
and inexpensively. Hunt et al., (2009, 2011) and Bonnici et al., (2008, 2009) describe a method to extract the
required mineralogical information to generate mineral maps and then rank the samples with respect to the
metallurgical response. Another methodology characterises the different textural features within the mineral
particles by automated identication. A classication system is then developed based on the mineralurgical
indices on each textural type (Perez-Barnuevo et al., 2012).
The most common geometallurgical models rely on the metallurgical response from geometallurgical testing
without the mineralogical information (Alruiz et al., 2009; Suazo et al., 2010; Niiranen and Bhm, 2012). Hunt
et al. (2012) developed a method to estimate the potential copper recovery by systematic rules for representatively ore types. The modelling is based on archetypes. The gangue mineralogy associated with the sulphides
was expected to have specic liberation and recovery properties. A subset of samples that has unique mineralogical-textural information was then classied as an archetype. Each archetype representing a distinct signature
that expects to have a different metallurgical response.

3.5. The geometallurgical approach at LTU

The geometallurgy research group of Lule University of Technology was formed in 2011, under the leadership of Professor Pertti Lamberg. This group includes both geologists and metallurgists, and the research is in
close relation to the mining industry. The literature review reveals that there are rather few similar groups in

31

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

the world, and the long term effort is to dene and bridge each area in the geometallurgical approach to create
a denition of the geometallurgical concept, illustrated in a gurative pyramid (Fig. 7). The research is mostly
within the ore deposits in the Norrbotten and the Skellefte area, Sweden (Johansson and Wanhainen, 2010;
Minz et al., 2013; Lamberg et al., 2013; Lund et al., 2013), and the main focus is to develop mineralogical tools
and methods for geometallurgy. In the building of a geometallurgical model, we dene the two lower levels

The

geo

me

tall

urg

ica

l co

nce

pt

(Fig. 7) as internal factors, where it is the ore and process parameters from the actual deposits that inuence
the outcome.

(Geometallurgical domains)
Business dimension
Mining management
Metallurgical constrains
Geostatical modelling
Sustainability, resource efficiency
Geotechnical approach
Geometallurgical process
description

Ore characterisation

Fig. 7 A gurative schematic pyramid, which shows the model building using the geometallurgical approach.

The rst level is the fundamental ore characterisation in an ore deposit (e.g. mineralogy, chemistry and
textures). The responsibility and reliability rests on the geologists/geology department whose aim is to make
a correct and in-depth picture of the geology. Earlier identied uneconomic mineralisation could with
application of the geometallurgical principals shift to new ore bodies (Dunham et al., 2011). This assessment
also requires a fundamental understanding of the complete mineral distribution preferable as a 3D model of
the ore body (Minz et al., 2013), which cannot be neglected. Another limitation includes the great condence
in the metallurgical response which only relies on a small number of samples that are tested in the laboratory
(Lamberg, 2011) representing large volumes of ore. A challenge is to transfer the knowledge into the second
level, both regarding the actual data, using quantitative rather than descriptive information, and it is crucial for
geologists and metallurgists to discuss their denitions to the various terms used, since they are often different.
This needs to be taken into consideration as the geologists and the metallurgists use the same words but different meanings, causing misunderstanding.
The second level includes the mineral/metallurgical processes, often as laboratory-scale test work but also as
pilot-scale tests. These processes involve more or less different kinds of geometallurgical tests, such as rapid
tests, measuring the metallurgical response directly or small scale tests using a chosen sharply dened process
cell. In this step there is frequently a problem of the representativeness generated by a small number of samples
that should forecast the entire ore deposit (Lamberg, 2011). The goal here is to create a process model based
on particles, as similar particles behave in the same manner regardless of the spatial location, and then create a
simulations platform that will overcome the restrictions caused by including parameters.

32

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

A geometallurgical model should be capable of forecasting the behaviour of the ore in comminution. Industrial
comminution circuits are commonly operated to produce a given xed particle size distribution, measured for
example with 80% passing, and this is regulated mainly by changing the throughput. Therefore, the comminution characteristic of the ore, i.e. grindability, controls the production volume. The challenge in the particle
based models and simulation is to develop a working grinding model.
The three top levels in the illustrated pyramid are external factors which inuence the geometallurgical outcome. The lowest of these levels includes the geostatistic modelling, sustainability and resource efciency, and
the level above encompasses mining management and the metallurgical constrain in the process. All of these
aspects are in close relation to the prevailing nancial uctuations in the global market and the world economy. The uppermost level is the geometallurgical domaining, and there is a reason for the brackets. This model
building is based on geological and process samples that are given a quantitative number. As almost every
parameter in the model is additive, we do not know yet if this domaining really is important and necessary for
the model building.
This literature review demonstrates that there has been a remarkable method development in modern geometallurgy, from a rather narrow vision which includes the traditional geological, mineral process and metallurgical subjects to a much broader view that includes sustainability, production management and an expanded
scientic publication in the 21st century. To create a geometallurgical model of the Malmberget iron ore
deposit, consideration needs to be taken about all of the characteristics of the deposit regarding the geological
properties as well as the technical aspects and criteria. This was the reason why a preliminary study was
performed. One exiting question is how far this PhD study could take the mineralogical approach in the
geometallurgical modelling.

4. SAMPLING, EXPERIMENT AND METHODS

4.1. Preliminary study
The literature studies showed two geometallurgical approaches, one based on geometallurgical tests and the
other relying on mineralogy. A Ph.D. study using the rst approach was already in progress on the Kiruna
deposit (Niiranen and Bhm, 2012), and therefore it seemed to be appropriate to use the second one in the
Malmberget study since the mineralogy is characterised by coarse grained magnetite with considerable variation within the ore bodies. With this in consideration, the hypothesis and the objectives were framed.
When this project started in 2006, it was an open eld to establish and improve the knowledge about
process mineralogy in the Malmberget iron ore deposit. As has been described above, the new semi-automated
quantitative analysis methods had become available, and this enabled the running of a larger dataset and the
measurement of a large number of different mineralogical properties. The main questions were still remaining:
What can be done? How should it be done? What will be important to measure and exactly why would this be
representative for this deposit? Malmberget iron ore was selected for the case study because it is a large deposit
containing several ore bodies of different characters, and the knowledge level on their mineralogy was limited.
The preliminary study had the primary objective to characterise the ore bodies using traditional methods of
applied mineralogy. The important questions were if the QEMSCAN instrument could be applicable on the
iron ore and how. This study was planned and performed in a small scale using few samples, including only
a crushing stage and three particle size fractions. When following objective was framed and the aims were

33

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

formulated, it was particular the original ore textures preservation and their effect in terms of process response
that have pin point this thesis.
In this study different ore bodies and the associated host rock in the Malmberget ore deposit are sampled and
analysed, some in more detail (Fig. 4).The Fabian and Printzskld ore bodies are the main focus; the Fabian ore
body was selected as the calculated ore production in the ore body is expected to increase in the near future,
and the Printzskld ore body was considered to validate the results of Fabian but also to identify differences of
the ore bodies in the deposit. Together with Vlkomma, Hens and ViRi, this constitutes part of the rst objective, the ore characterisation, in order to understand the petrochemical and metamorphic inuence on a local
scale as the geometry and the alteration features are individual for each ore body.
In addition to these ore bodies, another apatite iron ore (AIO) Gruvberget is sampled and studied, located in
the Svappavaara ore eld (Fig. 1). This dataset is mainly represented as a reference object to the Malmberget
ore, both as an AIO and also as an ore deposit on a regional scale since Gruvberget represents a lower degree
of metamorphic degree compared to the Malmberget deposit. Therefore, we have an opportunity to verify the
data and results from the Malmberget deposit and evaluate these developed methods and their application to
other deposits elsewhere.

4.2. Field work, drill core logging and sampling

For the host rock characterisation, 193 samples were collected from 22 drill cores and 4 out crops of the ViRi,
Fabian, Dennewitz, Printzskld, Vlkomma, Baron, Marta and Josena ore bodies. The sampling was done in
three different campaigns. For the mineralogical study, 116 polished thin sections were prepared.
The ore mineral characterisation consists of two data sets, a mineralogical and a metallurgical. The mineralogical samples were taken from seven drill cores and hand specimens from the underground ore, representing
the ore bodies ViRi, Fabian Printzskld, Hens and Vlkomma. Over 100 mineralogical samples were selected
for polished thin sections to characterise the mineralogy and the textural properties of the ore, of which 46 of
them where selected for mineral chemistry analyses (EPMA). 95 samples were collected for major and trace
element analyses.
The metallurgical samples were sampled from ve drill cores of Fabian and Printzskld. The drill cores were
carefully logged, and three dominate ore types were identied, generating a total of >100 kg in ve sample
batches.
The Gruvberget dataset is represented by 30 samples from nine drill cores. 32 polished thin sections were
prepared for optical microscopy and electron microprobe analyses, which were carried out at the Finnish
Geological Survey (GTK).

4.3. Sala Mrtsell separation

At the mineral technical laboratory at LKAB in Malmberget, Sweden, the metallurgical samples were processed
in order to simulate a dry sorting process in a bench scale test. Two steps of crushing were used, a primary
crushing by a laboratory jaw crusher and a secondary crushing using a cone crusher in a closed circuit with
a 3 mm screen. A dry, low-intensity magnetic separator (Sala Mrtsell) was used to separate the ore feed into
concentrate and tailings. The feed rate was 125 ton/h with a drum speed of 2 m/s 270 mm.

34

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

4.4. Sample preparation

At the mineral technical laboratory at LKAB in Malmberget, Sweden, the concentrate and tail from the
Mrtsell separation test was split by a Jones splitter and sieved with a Ro-Tap shaker to fractions <38, 38, 75,
150, 300, 425, 600, 1190, 1680 and 3000 micron.
At the metallurgical laboratory, Malmberget, polished resin blocks of powder were prepared from ve of the
sized fractions between 38 and 600 m. First, the samples were split by microrifing to obtain representative subsamples, then subsamples were mixed with graphite to avoid differential settling (Pascoe et al., 2007),
aggregation and also to create a random mix of particles. Finally, the subsamples were mounted into 30 mm
round moulds and lled with epoxy resin to a height of approximately 6-7 mm. After hardening, the blocks
were labelled and backlled with resin to obtain a size/height of approximately 20 mm. All of the blocks were
polished to a high quality 1 m nish, carbon coated and analysed utilising QEMSCAN. A total of 35 polished
blocks were prepared.

4.5. Optical microscopy

Optical microscopy was used to identify minerals and determine properties like mineral assemblages, alterations,
grain sizes and the textural relations. Rough estimates on the modal composition were done. The petrographic
studies were carried out using polished thin sections on the mineralogical samples and the polished blocks on
the metallurgical samples using a standard petrographic microscopy (Nikon Eclipse E600 POL).

4.6. Geochemistry
For the chemical analyses the ore and host rock (mineralogical) samples were prepared by the laboratory of
ALS, Chemex in Pite, Sweden. The whole-rock chemical analysis was carried out at Activation Laboratories
Ltd, Canada for major (ICP-OES) and trace elements (ICP-MS).
The chemical analyses on the metallurgical samples were all performed at LKABs chemical laboratory at
Malmberget, using their standardised method for X-ray uorescence (XRF). Bulk samples and size fractions
were analysed with XRF for Fe, Al, Ti, Si, K, Ca, Mg, Mn, V, Na, S and P. The amount of divalent iron was
analysed with wet chemical titration (ISO 9035), and the amount of magnetic material was determined with a
SATMAGAN magnetic balance model 135 (Stradling, 1991).

4.7. Electron microprobe (EPMA)

The chemical composition of minerals was determined by electron microprobe at the Camborne School
of Mines, University of Exeter, Cornwall Campus, UK on a JEOL JXA-8200 superprobe with a technique
described by Potts et al. (1995). An accelerating voltage of 15 kV and a beam current of 30 nA were used for
the oxides and the silicates. The probe beam diameter was set to 1 m for the oxides and the phosphates and
5 micron for the silicates. The FeO and Fe2O3 ratio was calculated on the basis of stoichiometry and charge
balance (Droop, 1987). Based on the mineral chemistry and the whole rock analyses, an additional 16 samples
were selected for detailed mineralogical study by electron microprobe of Fe-Ti mineral pairs at the Finnish
Geological Survey (GTK) using a CAMECA SX 100. The accelerating voltage was 20 kV, and a beam current
was of 60 nA. The probe beam diameter was set to 10 m for the oxides. Standards for the different elements
and analyses in the two methods were used of natural silicates oxides and metals.

35

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

4.8. QEMSCAN (Quantitative Evaluation of Minerals by Scanning Electron Microscopy)

QEMSCAN was developed by CSIRO in Australia for the global mining industry to analyse ore and mill
products in respect to mineral processing to provide rapid, statistically reliable, repeatable mineralogical and
quantitative mineral analyses (Sutherland and Gottlieb, 1991; Petruk, 2001). The system can be operated in
different measurement modes: eld images scan, bulk mineral analysis (BMA), particle mineral analysis (PMA)
and trace mineral search (TMS) (Pirrie et al., 2004). The principle of this technology is to use a combination
of signals from the backscatter electron detector (BSE) and the energy dispersive X-ray detectors. The backscattered electrons will recognise particles based on density contrast, and the compositional differences from the
X-ray spectra will create digital mineral images of a sample based on chemical composition (Gottlieb et al.,
2000; Sutherland et al., 2000; Goodall et al., 2005). All of the analytical data are processed ofine by an integrated software application (iDiscover 4.2) to the QEMSCAN system. In order to be able to identify the minerals
and phases in the sample, every analysed pixel and obtained X-ray spectra is compared to a known database of
mineral composition (SIP - Species Identication Protocol), which is the basis of obtaining an accurate and
complete mineral identication (Gottlieb et al., 2000).
In this study the SIP-le had to be developed and specied for each ore body and ore type. The characterisation was made from the ore and the gangue minerals present in the samples. The important mineral in this
study is magnetite, but the gangue minerals also needed to be identied. Special attention was paid during the
data processing and SIP development to ensure the correct identication of the iron ore species. Specically,
magnetite, hematite and goethite are distinguished based on BSE thresholds due to their chemistry being too
similar for x-ray analysis to separate at the standard 1000 x-ray count rate that QEMSCAN uses (Andersen et
al., 2009).
Generally, more than 5000 particles were measured for each sample by QEMSCAN. The exception was coarse
grained fractions where the scanning of the whole sample resulted in a minimum of 390 particles.

4.9. HSC 7.1

In this study an element assay to mineral grades conversion and the Particle tracking was done with HSC
Chemistry software 7.1 developed by Outotec (Lamberg et al., 1997).
Element to mineral conversion is based on a set of linear algebraic equations where the bulk chemical composition of a sample is converted to mineral grades using a set of least-squares equations (Paktunc, 1998; Whiten,
2008; Lamberg et al., 1997). This can be expressed as an equation
A*x=b

(1)

where A is a table giving the weight fractions of elements in the minerals (e.g. electron microprobe analyses,
known), the x vector is the weight fractions of the minerals in a sample (unknown) and the b vector is the
weight fraction of elements in the sample (known). For a simple magnetite albite binary with Fe and Si analyses the equation would be (chemical composition of minerals from Table 8):
0.73

0.00

0.00

0.33

x(Mgt)
x

0.60
=

x(Ab)

0.06

36

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

There are two constraints for the solution. Mineral grades (x) cannot be negative and the sum of minerals must
to be smaller than or equal to 100%. The solution can be found by searching for the minimum value for the
residual vector (R) (equation 2).
R=|b-A*x|

(2)

If there are more elements than minerals, the case is over-determined, and there is more than one solution.
If the number of minerals is higher than elements, the case is under-determined and there is no unique solution
(Paktunc, 1998; Whiten, 2008). The multiple solutions can be solved, e.g. by Gaussian elimination and multiple
regression techniques to minimise the sum of the square of the deviations.
As minerals in nature are seldom stoichiometric, the actual elemental composition of minerals analysed by
electron microprobe should be used (A matrix). Quite often simple bulk chemical assays of a sample do not
provide enough information for estimating mineral grades. For example, in porphyry copper deposits there
may be several copper sulphide minerals present in the sample, and simple Cu, Fe and S assays do not enable
all of them to be quantied. For metallic mineral deposits a number of selective analysis techniques have been
developed. Most of them are traditional wet chemical methods. For gold they have been known for decades
(Lorenzen and Van Deventer, 1993). Also, for copper they are widely used (Lamberg et al., 1997), but they are
quite practical as well for nickel (Penttinen et al., 1977), molybdenum and zinc (Young, 1974). In this study
two different techniques were used to independently determine the magnetite and hematite ratio: titrimetric
divalent iron analysis and SATMAGAN magnetic balance. For each minerals paragenesis one should evaluate
whether diagnostic methods are available and could be used since they will improve the accuracy of the element to mineral conversion.
The HSC software includes a reference table in its mineral database for the dissolution of minerals in several
selective analysis techniques listed above. In solving equations (1) and (2), HSC includes several mathematical
options and here the Least Square and the Non-negative Least Squares solutions were used (Lawson and Hanson, 1974). For further information on mathematical methods, the reader is referred to Braun (1986), Paktunc
(1998) and Whiten (2008) and the HSC manual (Lamberg and Tommiska, 2009).
The purpose of the processing of the mineral liberation data was to quantitatively track how different kinds
of multiphase particles deriving from different textural types behave in processing. For that purpose a particle tracking technique by Lamberg and Vianna (2007) was used. The Particle tracking was done with HSC
Chemistry software developed by Outotec. The algorithms of the Particle Tracking were also used in textural
classication of the progeny particles, as described below.

37

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The full particle tracking algorithm solves liberation mass balance in a given mineral processing ow sheet
through the following steps:
1 Conversion of elemental assays to mineral quantities
2 Unsized mineral mass balance (1D)
3 Size-by-mineral mass balance (2D)
4 Reading liberation data and combining minerals
5 Adjusting particle grades to match the size-by-mineral mass balance
6 Particle classication and binning (basic and advanced)
7 Interpolating/extrapolating particle classes in missing size fractions
8 Smoothing liberation data
9 Mass balance of multiphase mineral particle classes (bins)
10 Error analysis of the Multiphase Mineral Particle Mass Balance
Particle tracking uses mass balanced mineral by size data. Instead of using mineral grades by QEMSCAN, the
modal analysis was done with the element to mineral conversion technique (step 1) and calculation procedure developed by Lund et al. (2013). Using HSC the raw mineral grades with solids ow rate data was rst
balanced on an unsized basis (1D) and then for size by a mineral basis (2D) with an additional constraint that
the 1D mass balance is conserved (Steps 2 and 3). The experimental data showed good quality and adjustments
in steps 2-3 were small.
The QEMSCAN analysis gave a total of 16 minerals, but to make this match with the mineralogy of element
to mineral conversion some of the trace minerals were combined to more abundant phases (see section 5.3.2)
(Step 4). This gave a total of six most important minerals to be used in textural classication and tracking. After
combining the minerals, the next step of the Particle Tracking is to adjust the mass proportions of the particles
to meet the mass balanced 2D modal composition (Step 5). This stage adjusts the mass proportion of measured
particles but does not change their composition. The adjustment for the liberation data is very little but is
important by making the particle data fully consistent with the 2D chemical mass balance. This means that if
the chemical composition of the size fractions and streams is back calculated from the adjusted particles, the
result is exactly the same as received from the 2D mass balance.
To mass balance particles they must be grouped somehow and in the particle tracking this is done by their
composition. The basic binning stage groups particles for different binary and ternary combinations (Step 6a).
In the system of six minerals and ve size fractions after this step, the total number of particles in each stream
was 2166. Some of the particle groups may have very few particles if any. Therefore, the advanced binning is a
stage which combines established groups to ensure that each group has enough particles for sound mass balancing of particle groups (Step 6b). The liberation measurement is challenging in very ne (<10 micron) particle
sizes, and in coarse size ranges the number of measured particles will be very small. Therefore, the Particle
Tracking includes a step to extrapolate the liberation information (i.e. particles) for size fractions which have
not been measured (Step 7). In this study liberation was measured in ve size fractions, and the remaining two
size fractions were extrapolated.
After the previously described stages, each process stream will be composed of similar particles classes, but
they are not in balance. The last step reconciles and mass balances the whole circuit for given particle classes

38

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

(Step 9). This is done by keeping the total solids ow rate of the stream xed and by minimising the weighed
sum of squares between measured and reconciled mass proportion of the particle classes. The particle tracing
has an option for smoothing the data (Step 8), but this step was not used in this study. The error analysis was
done with Monte Carlo simulation which means that based on given standard deviations HSC drew a new
dataset and solved the particle mass balance (Step 10). This was done 100 times, and nally the standard
deviation for the particle grades and recoveries was calculated.
In the textural classication there is only an ore sample and its liberation analysis; therefore, there is no process
to mass balance. However, it was found that classication of liberation data in a uniform way is very useful for
establishing the textural archetype library. Therefore, the treating of the liberation analysis of ore samples used
steps 1: element to mineral conversion; 4: reading the liberation data and combining the minerals; 5: adjusting particle grades to match the size-by-mineral mass balance; 6: particle classication and binning (basic and
advanced) and 7: interpolating/extrapolating particle classes in missing size fractions. This procedure gave, for
each ore sample, exactly the same number of particles, and a given particle type was compositionally similar in
each ore sample.The only difference was that the particle grade was different in each stream. For example, each
stream contained a binary particle composed of magnetite and albite in the ratio of 40-50:50-60 wt% of size
75-106 microns. The mass proportion of this particle was different in each ore sample, and it would be possible
that the grade is zero in some samples (e.g. in an ore sample poor in albite).

5. RESULTS
This thesis describes the development of a geological submodel within the geometallurgical model for the
Malmberget iron ore deposit. Applied methodology is a so-called particle-based approach consisting of three
steps, developed and modied from a concept by Lamberg (2011). The rst step is the establishment of a
geological model relying on a proper ore characterisation to which Paper I and Paper II have contributed. The
conclusion is that a geological model requires at least two types of quantitative information: modal composition
and mineral texture. Paper III shows how to construct the modal mineralogy by using an element to mineral
conversion technique, and Paper IV develops a technique to measure and describe mineral textures through
textural archetypes.
In the second step of the building of a geometallurgical model, the reliability of the established geological
model is tested by a simulated mineral process to forecast the metallurgical response of each of the developed
textural archetypes. The particle brakeage model gives the liberation distribution of each textural archetype.
The model enables numerical prediction of the liberation spectrum with varying modal mineralogy. The third
step includes the process model generating quantitatively data of how the particle behaves in each unit process stage and therefore returns the metallurgical response (grade and recovery) for any given geological unit
(sample, block, domain). The developed geometallurgical model is the rst one for the Malmberget iron ore;
see section 6.2, for the summarised outcome and the state of originality.

5.1. Ore characterisation

The ore characterisation includes traditional geological description of the ores and the host rock of the
Malmberget iron ore deposit. Information on different characteristics of lithology, mineral composition,
mineral chemistry and textural properties is important and forms the basis in a geometallurgical modelling.

39

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The iron ores of Norrbotten are characterised by high Fe grade 30-70 % and a varying P grade up to 5 %,
often of large tonnage (Bergman et al., 2001). The main minerals are magnetite and hematite, and the gangue
minerals are apatite and amphibole.The accessory mineral includes pyrite, chalcopyrite and titanite.The apatite
is often present in signicant amounts and is generally strongly enriched in LREE (Martinsson, 2004).
The Malmberget deposit consists of several ore bodies of massive magnetite and hematite. Compared to the
Kiirunavaara ore body, Malmberget is lower in tonnages, slightly lower in Fe grade and has a moderate but
varying P content (Fig. 6). Mineralogically, the iron ores are rather simple. Magnetite and hematite are the
main ore minerals, and typical gangue minerals are apatite and amphibole-pyroxene. In the massive ore a broad
variation of mineral-texture relations can be identied. The magnetite ore has both a ne and a coarse grain
texture where the latter is more common (Fig. 8). In the apatite rich ore, the apatite exists either as single grain
interstitial to magnetite or aligned in the direction of the host rock lineation. The amphibole (-pyroxene) rich
ore shows two typical textures, a coarse grained and a ne grained amphibole (-pyroxene) where the proportion of the amphibole (-pyroxene) increases gradually to the skarn (amphibole-pyroxene) alterations.

Fig. 8 Different types of massive ore; A) ne grain magnetite, B) coarse grain magnetite, C) apatite-rich ore, D) coarse
grained amphibole rich ore and E) ne grained amphibole rich ore.

In the Malmberget deposit the extensive regional metamorphism and deformation have left very few primary
features. The chemical composition of the ore bodies has not changed much, but the magnetite and hematite
ores are otherwise fully metamorphically overprinted.These are seen as metamorphic textures, different oxidation textures and a redistribution of elements between the magnetite and hematite, which need to be considered even in a mineral processing perspective.
The massive magnetite has textural variations mainly in the magnetite grain size, grain shape and association.
Based on grain size magnetite can be classied into three types: ne grained (100-400 m), medium grained
(400-800 m) and coarse grained (800-1400 m) (Fig. 9). The magnetite grains are generally euhedral to
subhedral varying 0.4-1.4mm in size. The texture is mostly granoblastic with distinct triple junctions at the
grain boundaries. Coarse grained magnetite occurs as porphyroblasts or as veins in a ner grained matrix of
magnetite. Larger magnetite grains can have an elongated shape following the mineral lineation in the host
rock. In the western part of the deposit, hematite is the main mineral together with magnetite, occurring as
euhedral to anhedral grains 0.2-2mm in size within the hematite ore or as larger porphyroblasts (5-20mm) in
the magnetite ore.

40

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The Fe-Ti oxide minerals (ilmenite, magnetite and hematite) show certain oxidation textures which are only
present in the massive ore. Texturally there is variation between the individual ore bodies. Homogeneous magnetite, magnetite with intergrowths of ilmenite, rutile and spinel and also oxidation assemblages of hematite
and rutile occur. Magnetite may contain oriented ilmenite lamellas (trellis and sandwich) with a thickness of
about 2 m or exsolved spinel patches 50-200 m in size (Fig. 9). Rutile can be associated with magnetite that
has largely been altered to hematite. Single ilmenite grains occurring interstitially to magnetite contain, in their
central part, exsolved hematite lenses, ranging between 4.5-270 m in size. Euhedral grains of hematite have
inclusions of 20-100 m thick ilmenite lamellas, or they can contain 10-50 m large rutile needles exsolved at
the rim of the grain.

Fig. 9 Photomicrographs showing the grain morphology (size and shape) of magnetite and the oxidation textures
of magnetite, ilmenite and hematite at the Malmberget iron ore deposit (reected light). A) A coarse grain magnetite
porphyroblastic texture, B) A ne grain magnetite granoblastic texture, C) Exsolution lamellae of hematite in magnetite
grain, associated to ilmenite grains (Vr, Fa and Pz) and D) Magnetite with intergrowths of spinel and rutile (Pz and Hn).

Even if the iron grade generally is high regarding the production grade and recovery relation, the presence of
penalty elements and the locking relation of associated minerals requires a classication based on the mineral
processing aspects.
5.1.1. Classication of the Malmberget iron bodies

From the production point of view, the most important mineralogical and chemical differences between the ore
bodies are related to hematite/magnetite ratio, presence of apatite and the chemical composition of magnetite.
When the ore bodies are classied according to these features, they show a certain distribution geographically:

41

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

High V & Ti low P magnetite bodies:ViRi

Low V & Ti low P magnetite bodies: Fabian
Low V & Ti high P magnetite bodies: Printzskld
Low V & Ti high P magnetite bodies with hematite: Hens and Vlkomma
Low V/High Ti low P hematite bodies:Vlkomma
A varying V and Ti content in magnetite and hematite, and the presences of P in different ore bodies are prominent features that can be distinguished by the mineralogy, mineral chemistry and the textures.V is carried by
magnetite and ilmenite, but in the ore the ilmenite is subordinate to the magnetite (e.g. approximate content
is 1% ilmenite) (Table 4). The Ti is carried by magnetite, ilmenite, hematite and titanite. Simple mass balance
calculations indicate that about 80% of titanium is carried by ilmenite. However, direct correlation can be seen
between the TiO2 content of the ore and the TiO2 content in magnetite. The whole rock chemistry of the
magnetite ore shows the primary signature of V and Ti which is also reected in the chemical composition of
magnetite (Fig. 10).
Table 4 The average chemical composition (wt. %) of the Malmberget oxide minerals is analysed by electron microprobe in 16 samples, giving the elemental variation in the ore bodies.
Magnetite (Fe3O4)

Ilmenite (FeTiO3)

Hematite (Fe2O3)

Ore body

Vr

Fa

Pz

Hn

Vr

Fa

Pz

Pz

Hn

# analyses

30

36

97

18

23

30

40

30

16

31

22

FeO

94.0

93.4

93.8

93.6

94.2

47.0

45.0

47.6

89.6

89.6

90.3

TiO2

0.17

0.24

0.03

0.01

0.00

50.4

49.8

49.3

0.04

1.10

0.38

V2O3

0.28

0.24

0.18

0.10

0.10

0.31

0.28

0.28

0.13

0.16

0.18

Al2O3

0.17

0.17

0.08

0.12

0.09

0.00

0.00

0.00

0.04

0.03

0.19

MgO

0.06

0.12

0.04

0.16

0.05

0.47

2.04

0.92

0.00

0.06

0.00

MnO

0.00

0.02

0.02

0.05

0.08

1.71

2.92

2.05

0.00

0.01

0.00

Total

94.7

94.5

94.3

94.1

94.5

100.3

99.5

99.9

91.7

90.9

91.1

Fe2O3 (calc)

69.3

68.9

69.4

69.5

69.8

4.8

7.5

7.7

99.6

98.7

100.0

FeO (calc)

31.6

31.4

31.3

31.1

31.3

42.7

38.2

40.6

0.02

0.87

0.34

Recalc.sum

101.6

101.1

101.1

101.0

101.5

100.3

100.8

100.9

99.8

100.9

101.1

Phosphorous in the ore bodies is carried totally by the apatite. The iron and the phosphorous in general do
not have any good correlation; therefore the Fe vs. P gure (Fig. 10) shows the average level for phosphorous
in the ores.
The rst group (1) is magnetite ore that is rich in V - Ti and low in P. This type is represented by the ViRi ore
body. The second group (2), a magnetite ore low in both V - Ti and low in P, is typical for the Fabian ore body.
These two groups are located in the eastern part of the deposit. Both the vanadium and the titanium are in the
ore exclusively carried by magnetite; the ilmenite is rare and scarce. The chemical composition of magnetite
analysed by an electron microprobe reproduces the Ti-V graph of the different ores, and this conrms that the
observed trend in the rapidly lowering V with a steady decrease in Ti derives from the chemical composition
of the magnetite (Fig. 11).
The whole-rock analyses of the ore were used to identify the variations on the macro scale of the different
ores. Most of the ore bodies show quite monotonous T,V and P levels, but the Fabian ore body is an exception.
Based on the elemental distribution and the mineralogical composition, three subgroups could be differenti-

42

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Whole rock chemistry

2.0

0.4
Deposit
Vr
Fa
Pz
Hn
V (mix)
Grb
Ekb
KUJ
PG Mag
PG Hem

1.0
1
4,5

0.5

0.3

V2O3 (wt% )

TiO2 (wt%)

1.5

0.2

0.1

3
4

Mineral chemistry

0.30

Vr
Fa
Pz
Hn
V

0.20
1

0.15

P (wt%)

TiO2 (wt%)

Hn (44)

Whole rock chemistry

1.5

Deposit

0.25

Pz (1948)

V2O3 (wt% )

Fa (3487)

0.0

Vr (1274)

0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30

0.10

Deposit
Vr
Fa
Pz
Hn
Grb
Ekb
KUJ
PG Mag
PG Hem

1.0
4
3

0.5

0.05

2
1

3
4
0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30

0.0
40

50

V2O3 (wt% )

60
Fe (wt% )

70

80

Fig. 10 Chemical plots show the average (median) elemental distribution of different apatite iron deposits in the Norrbotten area. Top left: the whole rock analyses of V2O3 vs. TiO2 (wt. %), Top right: Box and Whisker diagram illustrating
the V2O3 (wt. %) variation of the whole rock content in different ore bodies with a Fe grade >50%. Bottom left: the
mineral chemistry of magnetite V2O3 vs. TiO2 (wt. %) from Malmberget ores and each classication number. Bottom
right: the whole rock analyses Fe vs. P (wt. %) of different ore bodies.Viri (Vr), Fabian (Fa), Printzskld (Pz), Hens (Hn),
Vlkomma (V) and Ekstrmsberg (Ekb), Gruvberget (Grb) and Kiirunavaara (KUJ B-ore), LKAB unpublished data and
Per Geijer (PG) ore (Park, 1973)
Classificationof ore bodies
3.0
2.5

4.

P2O5 (wt%)

2.0

3.

1.5
1.0

2.

0.5

1.
0

0.05

0.10

0.15

0.20

0.25

0.30

V2O3 (wt%)

Fig. 11 A discrimination diagram for the classication of different ore bodies where the average compositions of the
analyses should be used. The extension of each group is based on the chemical average from the 25th to the 75th percentile of the samples with an Fe grade >50%. 1 Viri, 2 Fabian, 3 Printzskld, 4 Hens

ated in the Fabian ore body. One subgroup contains higher amounts of Ti,V and Ga, and another is rich in P.
The third subgroup is more heterogeneous in character but is generally low in Ti,V, P and Ga.These chemically
different ore sections can be correlated between the drill holes at the Fabian ore body (Fig. 12) and are interpreted to reect primary variation within the ore. The mineralogy such as the associated pyrite, chalcopyrite
and anhydrite occur locally in small amounts disseminated and as veinlets in the ore, and are mainly restricted
to the ore sections rich in Ti-V-Ga. In the ViRi ore body, the average chemical composition of the ore is rather

43

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

uniform corresponding to the high Ti-V ore section found in the Fabian ore body with a similar mineralogy.
The whole rock analyses of the ore mimics the mineral chemistry, and in the Fabian ore the differentiated
mineral chemistry of both the V and Ti in magnetite are explained by these chemically different ore sections.
Fabian is therefore interpreted to be a key ore body that reects the large chemical variations within the ore
body.
Fa
6500 Fe (%)

50 50 65 75

300

Ti (%)

P (%)
0 0.4

1.1

0.4

0.9

Pz
6852

V (%) Ga (ppm)
0

0.13 0.26 0

55

Fe (%) P (%)

Ti (%)

V (%) Ga (ppm)

30

65 75

0 0.4

0.4 0.9

0.15

40

Magnetite
Skarn
Intermediate metavolcanic rock

400

Intermediate to felsic metavolcanic rock

100

Felsic metavolcanic rock

Ore Breccia
Apatite
200

Amphibole / Pyroxene
Sulphides

Meter

500

600

Feldspar
Biotite

300

Magnetite

700

400

465
800

877

Fig. 12 Schematic drill core logs from the Fabian and Printzskld ore bodies show the ore sections enriched in Ti,V, P
and Ga as colour shaded areas in the geochemical logs. Based on these variations, geochemical subtypes of massive ore in
Fabian ore occur as several rather regular bodies, up to 40 m wide that can be traced between the drill cores.

The third group (3), the magnetite ore low in V-Ti but high in P, is typical for the Printzskld ore body but also
for the magnetite ores bodies in the western part of the deposit. The magnetite is associated with apatite. The
ore bodies are lenticular in shape and varying in the apatite grade giving a banded appearance of the ore. Apatite
occurs interstitially to magnetite or as elongated grains often aligned in the direction of host rock lineation and
are similar morphologically and correlated to the coexisting magnetite.
The occurrence of hematite is more complicated. Firstly, the ore bodies composed exclusively of hematite
are classied as a separate group (5). The magnetite ore bodies with a minor amount of hematite are classied
as group (4). Hematite in these two existences shows completely different chemical composition. Group (4),
a magnetite ore low in V-Ti and high in P, shows similar characteristics as group (3), but it contains minor
amounts of hematite.The ore bodies of group 4 are spatially closely associated with massive hematite ore (Hens,
Vlkomma) of type (5). The low grades of V and Ti reect the composition of the original ore with magnetite.
The metamorphic recrystallisation has only redistribution elements between magnetite and hematite.
The effect of metamorphic overprinting and deformation within the Malmberget deposit is seen as differences
in the morphologic features (grain size, shape and textures), modal composition, as well as the chemical composition of the magnetite and hematite. These are results of an increased progressed oxidation which is slightly
higher in the western part compared to the eastern part of the deposit as also interpreted by Romer (1996)
and Annersten (1968). Some large scale deformation can be found in the western part of the deposit giving a
banded and stretched appearance for the ore (Geijer, 1930; Bergman et al., 2001).
In ore bodies where the magnetite to hematite ratio varies, these two minerals show variation in mineral textures and mineral chemistry. These features all point toward an alteration of magnetite to hematite. Different

44

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

textural relations with an increasing intensity of replacement indicate that the local oxidation conditions of
the ore bodies also have controlled the redistribution of trace elements between magnetite and hematite. The
oxide mineral assemblage changes from magnetite-ilmenite to hematite-magnetite, and several generations of
hematite occur in the ores. Hematite is found along the edges of magnetite grains and along fractures. A developed martitisation is seen as oxidation surfaces on magnetite but also as single grains or larger porphyroblasts
(>20mm) interstitial to magnetite.
The chemical composition of hematite shows much larger variation than magnetite. The Ti content can be up
to 12% TiO2 in the porphyroblasts where magnetite is the dominant Fe oxide and the hematite grade is low.
The co-existing magnetite is clearly poorer in Ti, and the Ti content in the whole rock is still comparable to
the corresponding original magnetite ores.This indicates that the system has been closed for Ti and V, and these
elements have been redistributed between hematite and magnetite favouring hematite (e.g. Printzskld and
Hens). When the hematite content increases in the magnetite ore, the TiO2 content lowers approaching to the
values of the original magnetite (e.g.Vlkomma, Fig. 13).

Magnetite

0.6

12

Vr
Fa
Pz 1
Pz 2
Hn
V

0.4

10

TiO2 (wt %)

0.5

MgO (wt %)

Hematite

Ore body

0.3
0.2

Ore body
Hn alteration
Hn porpyroblast
Pz alteration
Pz porpyroblast
V porpyroblast

0.1
0.0
0.0

0.1

0.2
V2O3 (wt % )

0.3

0
0.0

0.4

0.1

0.2
V2O3 (wt % )

0.3

0.4

Fig. 13 The chemical composition of magnetite and hematite from ViRi (Vr), Fabian (Fa), Printzskld (Pz), Hens (Hn)
and Vlkomma (V) ore bodies analysed by electron microprobe. A) The V2O3/MgO variation in magnetite from Vr, Fa,
Pz, Hn and V B)The V2O3/TiO2 variation in hematite shows two subgroups from Pz, Hn and V.

The predominating Fe-Ti oxide mineral pair in the ViRi and Fabian ore bodies is magnetite-ilmenite, and the
Ti and V content in magnetite is rather high (Fig. 13 and 14). Magnetite and ilmenite can coexist in a wide
T-fO2 range, but their chemical composition reects the stabilisation conditions in terms of temperature and
oxygen fugacity (Lindsley, 1991). As temperature decreases magnetite begins to exsolve TiO2 due to intensive
diffusion (Ramdohr, 1980), which ultimately results in trellis and sandwich textures (Mcke, 2003). A similar
process is seen and also explained for the coexistent magnetite and hematite in the Printzskld and Hens ore
bodies where the magnetite grains that have a higher content of Mg show exsolved spinel texture. A progress
in the oxidation state westwards is revealed by the transition to an almost pure end member magnetite and the
presence of hematite. This is typically seen in the Printzskld ore body where the original coexisting ilmenite-magnetite is gradually changing to a magnetite-hematite mineral pair, and the Ti is redistributed between
magnetite and hematite with a favour to the latter. The hematite in Printzskld and Hens ore occurs as Ti-rich
porphyroblasts and Ti-poor oxidation surfaces of the magnetite. According to Lindsley (1962), high Ti-content
in hematite gives an increased stability for the hematite at lower oxygen pressures because of the solubility of
ilmenite in hematite. In the Vlkomma ore body, hematite is a dominant Fe mineral, and the Ti content in
hematite corresponds to the whole rock Ti of the ore which is equally high as in the corresponding magnetite
ores.

45

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

300
400
10
Hem+Rt

600

700

100

tite
ne
Hem+Ilm
ag
-M
ite
at
Mgn+Rt
m
e
H

ViRi & Fabian

Ilm

10

log TiO2 (wt%)

-log 10 f (O2)

15

T (C)
500

20

1
Mag

0.1
Mgn+Ilm

25

Vr
Fa
Pz

0.01
0.01

100

100

Printzskld

0.1

log V2O3 (wt% )

30

Printzskld

Ilm

Ilm
Hem porphyroblast

10

log TiO2 (wt%)

log TiO2 (wt%)

10

1
Mag

0.1

0.1
Mag

Hem Martite

0.01
0.01

0.1

log V2O3 (wt% )

100

Hens
Hem porphyroblast

log TiO2 (wt%)

10

Mag-hem ite

0.01
0.01

Mag

Vlkom m a

Hem m assive

0.1
Mag

Hem m artite

0.1

10

0.1

0.1

log V2O3 (wt% )

log TiO2 (wt%)

100

0.01
0.01

0.01
0.01

log V2O3 (wt% )

0.1

log V2O3 (wt% )

Fig. 14 The elemental redistribution of V and Ti in magnetite and hematite during increased metamorphic conditions
in the Malmberget iron deposit.

In Malmberget a similar pattern is seen for the V distribution between magnetite and hematite, and this is
explained by the high prevailing oxygen fugacity and temperature (Fig. 14a). Magnetite in ViRi and Fabian
coexists with ilmenite, and both minerals are rich in V2O3 (Fig 14b). Schuiling and Feenstra (1980) have shown
that ilmenite in magnetite-ilmenite pairs close to the magnetite-hematite buffer will contain more vanadium
than magnetite because V in this state is more commonly V4+ and will substitute for Ti4+. However, the amount
of ilmenite is small, and the magnetite has a rather high V content. In the magnetite-hematite pairs the hematite
always concentrates more V since the number of available sites for Fe3+ to be substituted by V4+ is only 2/3 of
what exists in the coexisting magnetite. This is clearly seen in high V and Ti contents in the hematite porphyroblasts in Printzskld and Hens. The low content of V2O3 in magnetite in Hens and Vlkomma ore compared
to ViRi and Fabian ore is therefore a result of the redistribution of vanadium during extensive formation of
hematite porphyroblasts (Fig. 14c&f). Finally, when the hematite content increases, the titanium and vanadium
content in hematite sets to a level of magnetite, and this is true in massive hematite ores, too.
Norrbotten is a low to intermediate pressure province, metamorphosed under variable conditions; the few
temperature-pressure determinations which have been made indicate ranges of 500-800 C, and 2-6kbar

46

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

(Bergman et al., 2001). Previous work at the Malmberget deposit has indicated the metamorphic temperature
of 500-550 C using the coexisting iron oxide and Mg-Fe silicates but also co-existing magnetite and hematite (Annersten, 1968). Measurements given in Paper II based on the magnetite-ilmenite pair give very similar
results for both the temperature and the oxygen fugacity.
The hematite ores of group ve (5) have low V but high Ti grades compared to magnetite ores with similar Fe
grades (Fig. 10). The hematite ores are out of the scope of this study, and readers are referred to Geijer (1930)
for more information regarding these ores. However, the chemical and mineralogical characteristics are differentiated from the magnetite and magnetite-hematite ores studied here (types 1-4) indicating that the hematite
ores of type (5) have another origin.
5.1.2. Massive and semi-massive ore

Besides the iron mineralogy and the presence of apatite, the ore bodies have a signicant variation in their iron
grades and the abundance of silicate minerals.
The massive ore bodies, high in Fe and low in SiO2, are in the eastern part of the Malmberget deposit surrounded by a semi-massive mineralisation that can be several tens of meters thick with a decreasing iron grade.
This low grade Fe ore type has traditionally been called ore breccia. In the Kiirunavaara deposit Geijer (1910)
has described that the massive ore has intruded into felsic volcanic rocks and brecciated it. A similar description has also been given concerning the Malmberget (Geijer, 1930). However, for the Malmberget deposit the
breccia ore and cross-cutting relationships are more complicated as both deformations and the intrusion by
several types of dykes have generated different kinds of semi-massive ores. Besides existing at the rims of the
massive ore body, the semi-massive parts can be found as inclusions in the massive ore. Because of the complicated conditions and the absence of detailed petrological studies, these ore types here are called semi-massive
ore when the geometallurgical approach is considered (non-genetic term) and ore breccia in the ore geological
approach (Paper II).
The denition for the semi-massive ore is as follows: iron (magnetite and/or hematite) ore with iron grade less
than 55% Fe, and the main non-iron minerals are the silicates, i.e. amphibole, feldspars (albite and K-feldspar),
quartz and biotite, not the apatite (Fig. 15).
Na2O+K2O (wt%)

70

Massive ore
Semi-massive ore

60

Fe (wt%)

Na2O (wt%)

Ore type

50

40

30

20
0

10

20

30

40

50

P2O5 (wt%)

MgO (wt%)

SiO2 (wt% )

MgO (wt%)

Fig. 15 The whole rock chemistry reects the mass proportion of magnetite content in the massive ore and in
semi-massive ore.

47

K2O (wt%)

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The silicate mineralogy varies both within and between the different ore bodies. In a geometallurgical context
the semi-massive ores need to be studied in a more detailed way to nd out whether there exists a large variation in the processing properties within the ore type and if there is a need for subdivision.
In many places the massive ore gradually changes into a semi-massive ore. Massive ore changes to veins of
magnetite, apatite and amphibole-pyroxene developing into networks in the paler host rock rich in leucocratic
silicate minerals. That generates a semi-massive ore with a mineral composition and structures which are more
complex and comprise different mineral assemblages (Table 5). Ilmenite, pyrite, chalcopyrite, titanite and anhydrite are mainly found as accessory constituents in the semi-massive ore.
Table 5 The mineral assemblages that vary in the massive ore.
No.

Mineral assemblages in the ore

Magnetite+apatite+amphibole+pyroxenebiotite

Magnetite+amphibole+pyroxene+biotite+chalcopyrite+pyrite+feldsparapatite

3a

Magnetite+plagioclase+quartz+amphibole+chalcopyrite+pyritebiotit

3b

Magnetite+plagioclase+quartz+amphibole+biotitechalcopyritepyritetitanite

Magnetite+amphibole+anhydrite+chalcopyrite+pyritecalciteplagioclase

Magnetite+biotite+chlorite+feldsparamphibolepyroxene

Magnetite+plagioclase+quartz+amphibolechalcopyriteilmenitetitanitepyrite.

The main element chemistry reects the mass proportion of magnetite against the leucocratic-felsic material
resulting in a high content of Fe,V, Ti and P in the magnetite rich end and a high content of Si, Ca, Mg, Na, K
and S in the leucocratic-felsic material rich end (Fig. 15).The magnetite in the semi-massive ore is, in addition,
very low in titanium and vanadium compared to the massive ore.
The distinct chemical discrepancies of massive ore and the semi massive ore are important to consider. The ore
resource estimation uses traditional 3D block models which solely are based on the chemical analyses of the
ore samples. The amounts of semi-massive ore that are present in the process are consequently very signicant
and will not only inuence the ore reserve estimation but also contribute to elemental variations in the mineral process. Therefore, a classication between the massive ore and the semi-massive ore can clarify the true
distribution of the chemical elements from the valuable and the gangue minerals and thus improve the grade
and recovery.

5.2. Lithology of the Malmberget host rocks

Knowing the character of the host rock of the ore is important for several reasons in mining and geometallurgy.
Firstly, the mechanical properties and stability of the host rock determines the mining method, size of the tunnels and support needed. Secondly, a small part of the host rock will end up in the plant feed due to wall rock
dilution in mining. Thirdly, it is common that the ore characteristics change with the host rock.
The iron ore in the Malmberget deposits are hosted by strongly metamorphosed and deformed volcanic rocks.
Due to this strong recrystallisation and extensive alterations, the character and local stratigraphy of the host
rocks has not previously been comprehensively reconstructed. The chemical changes during alteration make
the use of mobile elements for the rock characterisation difcult, but also the use of elements that normally are
considered to be immobile may be unreliable. However, a few elements including Al, Ti and Zr have proved
to be useful for the identication of different primary lithologies (Fig. 16). A lithologic classication of the

48

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

host rock has been done based on the chemical and petrographical analyses of the host rock from several ore
bodies in the Malmberget deposit. These are metavolcanic rocks (basalt-andesite, dacite-rhyodacite-rhyolite),
felsic intrusions (deformed and undeformed aplite, granite), mac intrusions (metadiabase) and skarn (Table 6)
(Paper I).
Table 6 The petrographic description of the host rock lithology which characterises the Malmberget deposit.
Rock type

Greyish

0.1-0.4

Dacite rhyodacite - rhyolite

Pale red
to greyish
red

0.050.4

Deformed aplite

Pale red
to red

Undeformed
aplite

Pale red
to reddish

Granite

Mac
intrusions
and dykes

Altered
rocks

Grain
size
mm

Basalt-andesite
Meta
volcanic
rocks

Felsic
intrusions

Colour

Metadiabase
1+2

Metadiabase 3

Skarn

Pale red
to reddish

SiO2
%

Qtz
(%)

Amph
%

Textures

3-10

Generally these rocks exhibit a

well-developed foliation or mineral lineation and have a texture
that is grano-porphyroblastic to
slightly lepidoblastic.

0-3

Mostly a grano-porphyroblastic
texture with 1-5 mm fsp phenocrysts and in more felsic types
small amounts of qtz phenocrysts

61-72

1020

0.2-1
(1-2)

69-73

1520

1-6

0-5

0.5-2

Mostly weak to moderate foliation or mineral lineation. May

have porphyric texture with 1-5
mm large feldspar phenocrysts

0.2-2

72-76

2025

0-2

1-3

0.5-2

More or less undeformed aplite.

The rock has mostly a weak to
moderate foliation or mineral
lineation but may also be undeformed. Contain phenocrysts of
microcline (5-15) and plagioclase
(7-10 mm)

Dark grey
to greyish
brown
and greyish green

0.1-0.7
(2mm)

Green
to dark
green

0.3-15

0-3

0-25

Fe-Ti
oxides
%

52-66

5-15

5-25

Bt
%

0-5

72-76

2535

0-1

2-5

0.5-1

48-51

0-40

0-45

1-10

48-57

0-35

0-30

1-7

50-60

5-65

0-5

0-15

Generally no preserved primary

textures. Typically it has a granoblastic texture and a more or less
well-developed foliation or mineral lineation. Better preserved
dykes may show chilled contacts.
Amphibole often retrogrades
after pyroxene. Granoblastic to
granoporphyroblastic

Fig. 16 The primary character of host rock, distinguished by its geochemical signature, expressed as 100*TiO2/Al2O5
vs.1000* TiO2/Zr.
49

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Metavolcanic rocks

The ore-hosting metavolcanic rocks in Malmberget are similar to the intermediate to felsic volcanic host rock
to the Kiirunavaara deposit (Martinsson, 2004). The metavolcanic host rocks of Malmberget vary from more
a felsic composition in the western part to a mainly mac to intermediate composition in the eastern part.
The felsic rocks are also the dominant rock type in the central part of the area within the core of the syncline
structure.
Based on the SiO2 content, the mac to intermediate metavolcanic rocks are basaltic andesites and andesite-dacite. The felsic volcanic rocks consist of dacite and rhyodacite to a less common member of rhyolite.
Felsic intrusions

Both the metavolcanic rocks and the ores are intruded by several types of dykes. These include aplitic dykes
comagmatic with the volcanic rocks and a younger generation of aplite, pegmatite and granite of Lina type.
Deformed aplitic dykes are common in the host rocks but also occur in the ore. The undeformed aplite dykes
are less common in the mine but occur locally as dykes in the ore and in the host rock. Larger bodies of granite
are rare in the ore but occupy a large area to the north of the deposit.
Mac intrusions and dykes

Altered mac dykes with scapolite, amphibole and biotite have intruded the ore and the metavolcanic rocks.
They most likely represent several generations of intrusions. The metadiabase is common as up to more than
10 m wide dykes cutting the ore and the host rock. In the wider dykes the grain size often decreases toward
the contacts reecting chilled contacts. Often no primary textures are preserved, but the more coarse grained
dykes may have partly preserved magmatic textures. Based on the chemistry three different types may be distinguished based on the content of SiO2, TiO2, Zr and Th. Two types are high in TiO2 and with 48 to 51% SiO2.
The third type has lower TiO2 but higher Zr and a silica content of 48 to 57% SiO2 (Fig. 16).

5.3. Developing a framework and tools for a geological model

The geological information is highly important in the ore projects from the exploration stage through resource
delineation, the feasibility study and mining to closure of the operation. In the traditional approach much of
the geological data produced in the early stage of the project is left unused in the mining stage. The elemental
grades are the most important information in the mineral resource estimation, and frequently they are the only
one. However, a constant chemical composition can include variation in the modal mineralogy and mineral
textures which in turn cause variation in the processing properties. Descriptive mineralogical information, such
as lithology, mineral assemblages, grain sizes and textures, could be gathered relatively easy, but the data itself
has certain limitations. Descriptive mineralogical (or any other) information is difcult to include in the block
model as the properties are non-additive, and average value for a given block cannot be estimated by geostatistics. Therefore, it is more appropriate to use quantitative mineralogy for building the geometallurgical model.
The requirements for the ore characterisation techniques to be used in geometallurgy are high.They should be
capable of measuring and analysis of the different ore properties and in particular the mineralogical parameters.
The methods should be fast, inexpensive and above all practical, capable of being implemented routinely as
online tools.

50

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
5.3.1. Components of the geological model

A pure mineralogical approach in geometallurgy means that the geometallurgical model is fully based on the
mineralogy. The model uses mineral parameters, such as modal mineralogy, mineral textures, mineral association, mineral grain sizes and their relation to the liberation characteristics. Based on a particle approach modied after Lamberg (2011), a geometallurgical model of the Malmberget deposit can be established in three
sub-models (Fig. 17): a geological model, a process model and a production model.

Production
forecast

Particles
Geological
model

Unit process
model

Particle
breakage
model

Mineralogy
Textures

Simulation

Particles
behaviour

Fig. 17 The particle-based geometallurgical concept, modied from Lamberg (2011). Modal mineralogy and textures
link the geological model and the process model. In the process model minerals are treated as particles. From the geological information, the particle population is generated through the particle breakage model.

In a mineralogical approach the geological model relies on a proper ore characterisation and provides quantitative mineralogical data in such a way that elemental grades or lithology are not needed. The modal composition (mineral composition by weight percent) and the texture information (mineral association and grain
sizes) dene the geological model. A geological model is linked with the process model through particles. The
combination of a developed geological model and a process model gives a production model that is capable of
handling different process scenarios based on the included parameters from the two models.
The accurate modal mineralogy is the rst requirement for the geological model. Besides mass proportion it
also gives the chemical composition of each existing mineral in a geological unit (sample, block, domain).
The ore textures form the second part of the geological model. The traditional geological description of textures is mostly qualitative and includes parameters like grain size (coarse, moderate, ne), grain shape (euhedral,
prismatic, anhedral) and associated minerals. Descriptions such as these are insufcient in a geometallurgical
perspective, and there is a need to develop a textural analysis which gives a numeric description of the textural
properties by using additive parameters. Only then the textural information can be used both in modelling and
geostatistics.
Mineral textures and the liberation characteristics are closely associated.Therefore, an effort is made to link the
textural description and the mineral liberation distribution. This information is essential and must be included
in the geological model. This denes the 3D level that uses the mineral liberation in the process model which
allows forecasting the metallurgical functions.
The directions outlined above are the requirements for the geological model of the Malmberget deposit. The
ore characterisation showed that signicant mineralogical variation exists between and within the ore bodies.

51

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

From a production perspective the systematic elemental redistribution between magnetite and hematite is an
advantage. Magnetite in the western part of the deposit is very pure and poorer, not only in V and Ti but also
in other trace elements than those that are normally in the apatite iron ore. On the other hand, associating
hematite is rich in Ti.This magnetite-hematite association can be seen as a naturally concentration process, and
there is a possibility of using this information in production planning to produce an iron ore concentrate of
different qualities in terms of trace elements.
Texturally, the magnetite and the gangue minerals in Malmberget are relatively coarse, and the liberation size is
reasonably large. The elemental deportment of intergrowth in magnetite which is especially rich in Ti is hard
to treat and is instead a disadvantage. If Malmberget is compared to the less metamorphic and deformed iron
ore deposits in the Kiruna area, the former shows a granoblastic texture, whereas magnetite in Kiirunavaara is
columnar and skeletal. Due to this difference Kiirunavaara ore requires ner grinding and apatite oatation.
5.3.2. Developing a routine method for modal analysis

The mineralogical approach requires a quick and inexpensive modal analysis method considering the need of
producing that information in a large number (>10 000) of samples. The element to mineral conversion is a
technique where the mineral grades are calculated from chemical assay using the information on the chemical
composition of the minerals. Mathematically, the problem is a system of linear equations, and generally it is
solved with a non-negative least squares technique (Paktunc, 1998; Whiten, 2008; Lamberg et al., 1997).
To develop this method, drill core samples from Printzskld and Fabian were selected. The sample selection
was based on the specic mineralogy of each ore, and three representative sub-ore types were selected. A total
of ve predominate sub-ore types were collected: namely feldspar rich ore, apatite rich ore and amphibole rich
ore, the last one found only in Fabian ore. The biotite rich ore, typical for group 3, was not sampled.
Actual and reliable mineral lists were established for both ore bodies. The optical microscopy was used to
identify minerals (Table 7), and their elemental composition was analysed by electron microprobe (EPMA)
(Table 8). The mineralogy and the average chemical composition of each phase were found to be signicantly
different in Fabian and Printzskld (see section 5.1.1); therefore, two different mineral lists were used.
The samples were analysed for Fe, Al, Ti, Si, K, Ca, Mg, Mn, V, Na, S and P by routine X-ray uorescence
(XRF). In addition, two diagnostic analysis methods were used, a chemical titrimetric method (Fe2+) and the
Satmagan (Fe2+) magnetic balance method to determine the amount of divalent iron and the amount of magnetic material. At LKAB both of these analyses are routinely done when hematite is expected to be present in
the ore.
By knowing the chemical composition of the minerals and the chemical composition of samples, a set of linear
equations can be written. In matrix format this is A*x=b.
Where A is a matrix for the chemical composition of minerals, x is the mass proportion of minerals in the sample, and b is the chemical composition of the sample.The maximum number of minerals to be solved is limited
by the number of independent chemical assays. Therefore, some simplications were required.

52

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 7 Typical (average) modal composition of feldspar rich ore, concentrate and tailing from a Mrtsell dry
magnetic separation laboratory test.
Mineral

Orea

Formula

Concentrate

AVG
wt %

STD

Tail

AVG wt %

STD

AVG
wt %

STD

Magnetite (Mgt)

Fe3O4

54.00

1.04

90.90

1.35

2.87

0.23

Ilmenite (Ilm)

FeTiO3

0.89

0.13

0.88

0.13

0.99

0.13

Rutile (Rt)

TiO2

0.39

0.09

0.44

0.09

0.33

0.08

Quartz (Qtz)

SiO2

1.14

0.15

0.33

0.08

2.22

0.20

Calcite (Cal)

CaCO3

0.09

0.04

0.01

0.01

0.23

0.06

Albite (Ab)

NaAlSi3O8

19.07

0.61

2.83

0.22

42.54

0.90

Orthoclase (Or)

KAlSi3O8

4.77

0.29

0.96

0.12

9.54

0.42

Biotite (Bt)

K(Mg,Fe)3(AlSi3O10)(OH)2

4.49

0.25

0.66

0.10

9.11

0.36

Diopside (Di)

CaMgSi2O6

1.89

0.19

0.29

0.07

4.28

0.29

Actinolite (Act)

Ca2(Mg,Fe)5Si8O22(OH)2

7.04

0.37

1.03

0.13

15.72

0.56

Chlorite (Chl)

(Mg,Fe)5Al(Si3Al)O10(OH)8

0.46

0.09

0.27

0.07

0.75

0.12

Titanite (Tit)

CaTiSiO5

0.27

0.07

0.12

0.05

0.44

0.09

Zircon (Zr)

ZrSiO4

0.06

0.03

0.00

0.01

0.12

0.04

Apatite (Ap)

Ca5(PO4)3(F,Cl,OH)

3.97

0.24

0.28

0.07

8.72

0.36

Sulphates (anhydrite, gypsum) (Py)

CaSO4, CaSO4*2H2O

0.25

0.06

0.02

0.02

0.59

0.10

Sulphides (pyrite, chalcopyrite) (Py)

FeS2, CuFeS2

0.26

0.07

0.07

0.03

0.56

0.09

a Ore sample back calculated from concentrate and tailing.

Table 8 The average chemical composition (wt. %) of Malmberget minerals (Fabian ore) analysed by electron microprobe in 25 samples, giving the variation in the ore body. For the A-matrix the oxides converted to elements.
Mgt

Ilm

Qtz

Ab

Bt

Di

Act

Ap

Hem

590

40

60

12

36

43

59

14

100.4

72.7

TiO2

0.08

49.8

Al2O3

0.10

Cr2O3

0.02

V2O3

0.08

# analyses
SiO2

20.36

FeO

93.7

44.6

0.01

2.90

MgO

0.01

1.90
0.01

0.06

0.01
0.19

NiO

0.04

0.01

0.35

13.06

0.23

1.88

0.05

0.01

0.01

13.85

9.61

10.3

0.21

0.01

0.05

0.14

0.05

0.01

17.01

13.0

17.6

0.12

21.8

11.5

56.5

1.51

1.06

0.01
6.89

K2O

55.4
0.03

0.12

Na2O
P2O5

54.7
0.02

0.29

MnO
CaO

41.92
4.00

0.01

0.03
7.90

90.2

0.13
36.8
0.01

1.36

0.18

Cl

6.58
0.28

Sum

94.2

99.5

Fe2O3 (calc)

69.4

6.87

FeO (calc)

31.2

38.5

0.31

Recalc. sum

101.2

100.2

100.8

100.5

99.9

99.2

101.0

98.2

100.4

90.7
99.9

Pyrite and orthoclase stoichiometric

53

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Zr and C are missing in the chemical assays, so the mass proportion of neither zircon nor calcite can be
estimated. These minerals are not relevant in the concentration process as they are trace minerals. To further
reduce the number of minerals, the titanium minerals (ilmenite, rutile and titanite) were combined, and the
most voluminous one, ilmenite, represents all of the titanium minerals. The same was done for sulphur bearing
minerals: pyrite represents sulphides (pyrite, chalcopyrite) and sulphates (anhydrite, scapolite). Finally, as chlorite
is a minor mineral occurring as an alteration product of biotite, it was left out as well and is reported together
with biotite.
Two of the analysed elements by XRF, namely V and Mn, occur only as minor constituents in minerals and are
not a very good choice for mineral quantication; they are therefore not used in our calculation. All together
it gives in the end a total of eleven independent assays enabling a maximum number of eleven minerals to be
quantied.
Having eleven elements and eleven minerals, it would be possible to calculate all of the minerals simultaneously.
To avoid negative solutions and good control for some trace minerals, it was found that the best solution is
to divide the calculation into several rounds. After testing different alternatives, a plausible mineral conversion
solution for the two ore bodies was found by using a total of three calculation rounds and both LS- and NNLS
algorithms (Table 9). In the rst round the minerals which carry solely one element are solved with the LS
technique. This means that the apatite and pyrite grades are solved using phosphorous and sulphur. The second
round takes silicate minerals: albite, orthoclase, biotite, diopside, actinolite and quartz, using residual Na, Mg,
Al, Si, K and Ca, and the NNLS solution is used. Since ilmenite and magnetite contain most of Ti-Fe and Fe,
the third round uses the residual Ti and Fe (XRF). Diagnostic assay methods, SATMAGAN and Fe2+ offer a
possibility to calculate both hematite and magnetite, if present. For this alternative hematite and SATMAGAN
or Fe2+ are also included in the third calculation round. Both of the ores have individual adjustments of the
calculation rounds (Table 9). The calculated modal composition (wt.%) of the ore samples from the predominate subore types for 1D level in a process model is shown in Table 10.
Table 9 The calculation routines that are used in the modal calculation. Two alternative calculation routines; left: all
minerals calculated in one step, right: calculation divided into three rounds. The latter that gave a better result was used
(see text). The elements showed as oxides.
General

Fabian

Printzskld

Round

Mineral

Components

Methods

Round

Mineral

Mineral

Components

Methods

Apatite (Ap)

P2O5

NNLS

Apatite (Ap)

Apatite (Ap)

P2O5

LS

Pyrite (Py)

Pyrite (Py)

Pyrite (Py)

Quartz (Qtz)

SiO2

Quartz (Qtz)

Quartz (Qtz)

SiO2

Albite (Ab)

Na2O

Albite (Ab)

Albite (Ab)

Na2O

Orthoclase (Or)

MgO

Orthoclase (Or)

Orthoclase (Or)

MgO

Biotite (Bt)

K2O

Biotite (Bt)

Biotite (Bt)

K2O

Diopside (Di)

Al2O3

Diopside (Di)

Al2O3

Actinolite (Act)

CaO

Actinolite (Act)

Actinolite (Act)

CaO

Magnetite (Mgt)

Satmagan

Magnetite (Mgt)

Magnetite (Mgt)

Satmagan

Ilmenite (Ilm)

TiO2

Ilmenite (Ilm)

Ilmenite (Ilm)

TiO2

Hematite (Hem)

FeO(tot)

Hematite (Hem)

Hematite (Hem)

FeO(tot)

54

NNLS

NNLS

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 10 Calculated modal composition (wt.%) of ore samples from the predominate subore types of the classied
iron ore 2 (Fabian) and 3 (Printzskld) information for a 1D level in a process model.
Sub ore
type

Stream

Mgt %

Ilm %

Ab %

Or %

Amph%

Bt %

Qtz %

Ap %

Py %

Fa Fsp

Ore

51.38

1.06

32.69

2.7

7.1

1.65

0.37

0.28

97.24

Fa Ap

Ore

84.78

0.37

0.14

3.19

2.78

0.49

6.96

0.08

98.79

Fa Amph

Ore

64.29

1.01

2.83

22.31

6.23

1.22

0.15

98.05

Pz Fsp

Ore

42.11

1.65

20.23

13.46

7.81

6.36

6.06

0.22

97.9

Pz Ap

Ore

79.42

0.69

0.44

4.23

1.02

12.63

0.18

98.61

Total

Comparison of element to mineral conversion to QEMSCAN analyses

The element to mineral conversion technique described above is developed with emphasis to calculate the
modal mineralogy directly for ore samples. But when the accuracy and the reliability of the mineral quantication are tested, it also includes the products from the Mrtsell separation test, concentrate and tail. QEMSCAN
analyses (modal mineralogy) were produced of each size fraction of concentrate and tail (10 samples) and compared against the calculated mineralogy. The standard deviation between the methods in the mineral grades is
on average 2.4 wt. % (Fig. 18). The corresponding relative standard deviation for the individual minerals varies
being 6.5 for magnetite and higher than 15 wt. % for the gangue minerals.
7

Standard deviation (Std)

6
5
4

Mineral

Mgt
Ab
Or
Ap
Bt
Amp
Ilm
Py

3
2
1
0
0

10

20

30

40

50

60

Mineral grade (wt.%)

Fig. 18 The average mineral grades of different minerals whose grade is estimated by two methods (x-axis) against
standard deviation between the methods.

Since the minerals (variables) are correlated, a multivariate analysis of variance (MANOVA) was used to test
the statistical agreement (signicance) of the element to mineral conversion technique and the QEMSCAN,
compared in parity plots (Fig. 19). It uses the vectors of the mean of the included minerals in the two data sets
of the correlated variables involved.
The MANOVA method initially tests if there is a signicant difference between the two mean vectors. If
signicance is detected, an analysis of variance (ANOVA) for each mineral pair will identify the mineral/
group of minerals that causes the signicant difference between the methods (Johnson and Wichern, 1998).
Simultaneous condence intervals (Bonferroni) are calculated in the single variable ANOVA for each mineral.
The Bonferroni correction is used in order to keep the p-value correct due to the correlation between the
variables/minerals.

55

80

60

60

40

40

20

20

16

20

40

60

QEMSCAN

80

0
100

16

Apatite

14

14
R square=0.858

12

12

10

10

0
0 2 4 6 8 10 12 14 16 18 20 22 24 26

70

70

Albite

60

60

R square=0.955

50

50

40

40

30

30

20

20

10

10

25
20

10

20

30

40

50

QEMSCAN

60

0
70
25

Biotite
R square=0.917

20

15

15

10

10

0
0 2 4 6 8 1012 1416182022242628

QEMSCAN

Element to Mineral conversion

R square=0.979

80

Element to Mineral conversion

100

Magnetite

Element to Mineral conversion

Element to Mineral conversion

Element to Mineral conversion

100

Element to Mineral conversion

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

20

20

Orthoclase
R square=0.904

15

15

10

10

20

0
10 12 14 16 18

QEMSCAN

20

Amphibole

15 R square=0.871

15

10

10

0
0

10

20

0
30

QEMSCAN

QEMSCAN

Fig. 19 A comparison of the modal mineralogy (wt. %) of iron ore determined by the QEMSCAN (x-axis) and the
element to mineral conversion (y-axis).
The statistical evaluation - What is the accuracy of the method

Before new quantitative methods are introduced, their reliability must be evaluated, and the highest accepted error dened. According to Pitard (1989), the highest accepted error is for commercial purposes, 0.5 %
(Table 11). For resource estimate the error must be below 5%. The lowest level, the exploration and environmental estimation require an accuracy of 16 wt. % (Pitard, 1989).To meet this 5% request, it sets high requirements for the sampling, sample preparation and the analysis to avoid the sampling error being raised so high
that it limits the usefulness of the collected data (Gy, 1982; Pitard, 1989).
Both the relative standard deviation (6.5; >15 wt. %) and the MANOVA-ANOVA (p-value 0.10) comparison
of the mineral grades show a signicant difference and a lower accuracy than required for the resource estimation.
Table 11 The relative standard deviation (RSD) in the modal analysis with the element to mineral conversion technique for the Malmberget iron ore samples. Maximum accepted errors marked with an asterisk from Pitard (1989),
others modied after Lamberg et al., 2013.
Purpose of the analysis

Max accepted relative

error for Iron oxides

Max accepted relative

error for gangue
minerals

Magnetite
Malmberget

Gangue minerals
Malmberget

Exploration and environmental*

16*

16*

OK

OK

Lithology identication

15

15

OK

OK (15)

Ore type identication

10

10

OK

Not OK

Geometallurgy domain classication

7.5

10

OK (6.5)

Not OK

Geometallurgical model

7.5

Not OK

Not OK

Technical sampling, process control

and resource estimate*

5*

5*

Not OK

Not OK

Process mineralogy

Not OK

Not OK

Commercial*

0.5*

0.5*

Not OK

Not OK

56

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
5.3.3. Geometallurgical ore type classication

As shown above the element to mineral conversion introduces error in the mineral grades too large to be used
directly and simultaneously alone in the resource estimation. The error in the magnetite grade approaches the
level for the geometallurgical domain classication, but for the gangue minerals the error is too high. However,
this information can be used to develop a classication system which is more reliable than the lithological name
given in the logging. A modal classication can, for example, differentiate hard and soft ore and point out the
location of weak ore (Ghosh, 2012). The mineralogical variation can be used to create 3D mineralogical maps
over the ore body, for instance revealing the magnetite and hematite ratio, and since the classication uses the
included mineralogy, it is a generic system that also can be applied for other Fe ores.
The calculated mineral grades were used to develop preliminary geometallurgical ore types (GEM-type)
(Fig. 20). The GEM type denition aims to capture all of the important mineralogical features related to ore
processing (e.g. metallurgical meaning): 1) Fe mineral grade, 2) Fe-mineralogy, 3) gangue mineralogy and the
4) presence of sulphides (e.g. H/Mgt/Ap(Amph)).

Gangue
Mineralogy

1) Fsp

>32%

Fsp
>60%

>5%
Py

2) Amph+Ap+Bt+Py

>45%
Amph

>35%
Ap
3) Qtz

>10%

>35%
Bt
>80%

4) Unclassified

2a) Py bearing Amph- (Ap-Bt)

2b) Amph-(Ap-Bt)
2c) Ap-(Amph)
2d) Bt-(Amph-Ap)
2e) Amph+Ap+Bt
Qtz bearing Fsp+Amph

Fig. 20 A geometallurgical ore type (GEM-type) classication scheme of the gangue mineralogy of Malmberget iron
ore. For the classication the gangue mineral grades are recalculated to 100%.

As the iron grade is overall quite high, normalised gangue mineral grades were used as a basis to distinguish and
characterise the different GEM types with special reference to the silicate mineralogy (cf. silicate problem).The
Feldspar rich GEM- type (Fsp) is common in low grade ores of magnetite, and the hematite ore, in particular,
is totally hosted by the feldspar GEM-type.The amphibole GEM-type is quite widespread in the magnetite ore
and is taken as a subclass of its own because it is a harder rock type than the feldspar GEM-type.The amphibole
GEM-type is further divided into ve subtypes. Iron concentrate has restrictions to sulphur and phosphorous;
therefore, it is important to identify the variation in terms of sulphides and apatite. Thus, pyrite GEM-type (Py
bearing (Amph-(Ap-Bt)) and apatite bearing GEM-type (Ap-(Amph)) classes were established. Locally biotite
rich zones can be found, and they cause some problems in rock stability. Additionally, in comminution, biotite
tends to enrich in ne size fractions, and this is partially the reason for elevated silica content in the concentrate.

57

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Consequently, it is important to identify areas rich in biotite, and a subgroup was created accordingly, GEMtype (Bt-(Amph-Ap)). Quartz is not a typical ore gangue mineral in the ore but exists in the wall rock. The
presence of quartz in the modal composition indicates a presence of harder wall rock which may be important
to identify deserving a subclass of its own, GEM-type (Qtz bearing Fsp+Amph).
From producing the model mineralogy solely to use the grades in the geometallurgical ore type classication,
Fabian and Printzskld truly show a distinct mineral grade discrepancy within the ore in between the ore
bodies. The accuracy and reliability of this GEM-type classication system was estimated against optical microscopy. In 95% of the cases, both methods gave the same result. Preliminarily, seven different GEM-types are
dened, and these are most probably both too many and detailed (Fig. 21).
DDH 6533

DDH 6716

100%
90%

80%

80%

70%

70%

Mineral grade

90%

60%
50%
40%
30%

50%
40%
30%

20%

20%

10%

10%

0%

0%

100%

100%

90%

90%

80%

80%

70%

70%

60%
50%
40%
30%

Py %
Ap %
Qtz %
Bt %
Amph %
Fsp %
Ilm %
Hem %
Mgt %

60%

Mineral grade

Mineral grade

Mineral grade

100%

60%
50%
40%
30%
20%

20%

10%

10%
0%
100

200 400

500

0%

(m)

600

Mgt
Medium

High

700

800

300

Low

(m)

350

400

Hem

Fe-mineralogy
Low

High

450

500

Mgt
Medium

Fe-grade
Gangue mineralogy

Fig. 21 A graphic log showing the modal composition of two drill cores DDH-6533 from Fabian (left), DDH-6716
from Printzskld (right). Pictures below show the gangue mineral grades normalised to 100%.

5.3.4. Developing a method to quantify mineral textures

In geology, the ore and rock are traditionally described in terms of structures, which refer to large scale properties such as banding, bedding, and layering. Texture, on the other hand, refers to small scale properties such
as grain size, grain shape, crystallinity and mineral associations, which are generally described, while the spatial
interrelationships are often described in qualitative terms. There are also several terms used to describe grain
shapes and mutual relationships, as well as arrangements of mineral in samples. Mineralogists use the term
fabric for this, and the term texture is reserved for orientation of the minerals in a sample. However, among the
economic geologists and process mineralogists, the term texture is generally and widely used instead of fabric.
Therefore, in this study the terms mineral texture and texture are used instead of fabric.
Using drill core logging (i.e. macroscopical observations) and optical microscopy, mineral textures of the massive and the semi-massive ore were classied. Textures were described and classied according to magnetite
grain size and shape, generic grain size and the main associating mineral. The magnetite grain sizes were
measured (Table 12). The following types, arranged in the order of increasing magnetite grade, were identied:
disseminated, veiny, patchy, banded, granules, speckled and massive.

58

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 12 A classication scheme outlined for the massive and semi-massive ore from the Malmberget deposit based on
the mineralogical and textural relation.

Ore
type

Semi
massive
ore

GEMtype

Femineral

Fe
grade
*

Texture
type

Sub
type

Grain
shape

Grain
size,
mean
(m)

Texture description

Main
associating
mineral to
magnetite

Fsp

Mag

<10 %
(L)

Disseminated

Fine

Euhedral to
anhedral

44

Fine grain magnetite occur

as disseminated in the
silicate matrix

Fsp-Qtz

59-74

Mag grains, irregular

distribution in the grain
boundaries of ne grain
silica matrix or as smaller
inclusions in fsp grain

Mag

Qtz
bearing
Fsp+Amph

Mag

Amph(Ap-Bt)

Massive
ore

Fine

Fsp,
Bt-(AmphAp),
Qtz
bearing
Fsp+Amph

Amph(Ap-Bt),
Ap(Amph)

Ap(Amph)

Mag

Mag

Mag

10-80
%
(L-H)

>80 %
(H)

Veiny,
patchy,
banded,
granules

Speckled

90-100
%
(L-H)

90-100
%
(L-H)

90-100
%
(L-H)

Euhedral to
anhedral

Coarse

Euhedral to
anhedral

90-129

Coarse

Sub- to
anhedral

133

Larger magnetite grain

giving a semi-massive
appearance

FspQtzAmph

Sub- to
anhedral

100-200

Amph

Euhedral

200-400

Mag grain in a massive

appearance as a compact
equigranular matrix with
coarser grains occurring as
vein like structure

Subhedral

400-600
6002400

Mag grain occurs as

granoblastic texture with
distinct triple junctions
at grain boundaries or as
porphyroblasts (outline in
linear direction)

Bt+Amph+Ap

Mag grain occurs as

granoblastic texture with
distinct triple junctions
at grain boundaries.
Elongated grains outline
in a linear direction or
as porphyroblasts gives a
brittle appearance

Ap

Fine

Euhedral
Massive

Fsp-Qtz-BtAmph

Mag grain occur as single

grains in grain boundaries or as aggregate often
outline in veins or patches
in the matrix

Medium

Coarse

Porphyroblasts

Euhedral

20003200

8001400

Clear dependence between the magnetite grain size and magnetite (i.e. Fe) grade was found (Table 12). In the
disseminated texture magnetite is an average of about 40 microns, and grain size increases to about 140 microns
in the semi-massive ore up to 1400 microns in the massive ore (Fig. 22, Table 13).
Even though the textural classication was not based on modal mineralogy, textural classes follow and are
somewhat unique for different GEM types.
The Fsp GEM-type in the semi-massive ore consists of two different materials identied by their colour.
Melanocratic magnetite rich material brecciates the leucocratic albite-orthoclase-rich magnetite-poor matrix.
The melanocratic parts consist of magnetite, apatite and amphibole veins, bands or patches that may develop
into networks in the leucocratic rock. The mineralogy of the leucocratic brecciating rocks consists of a varying
proportion of albite, orthoclase, quartz and biotite with minor magnetite.

59

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
140

Grain size (m)

120

Texure type

Melanocratic parts

(1) Disseminted
(2) Banded coarse
(2) Banded fine
(3) Veiny coarse
(3) Veiny fine
(4) Patchy coarse
(4) Patchy fine
(5) Granules coarse
(5) Granules fine
(6) Clustered coarse
(6) Clustered fine
(8) Speckled

100
80
60
Leucocratic parts
40
20
0

10

20

30

40

50

60

Fe (wt.%)

Fig. 22 Magnetite grain size magnetite grade (Fe) shows a positive correlation in the Fsp GEM-type of Fabian ore.
Fine grained and coarse grained form a bimodal system with two end members. The absence of texture type 7 is identied for Printzskld ore body.
Table 13 Eight representative and different identied macrotextures of the classied Fsp GEM-type for different
ore bodies in Fabian and Printzskld ore bodies.
Texture type

Sub-type

(#) An increasing Fe-grade in the

geochemistry analysis

Fine grain texture

(Mgt-Fsp)
(Micro scale) (500 m)

Picture
(Macro scale) (4 cm)

Coarse grain texture

(Mgt-Fsp)
(Micro scale) (500 m)

Disseminated (1)

Non existing texture

Banded (2)

a)Waving
(3)

veins

Veiny
b)Small veins (7)
typical in Pz ore

Patchy(4)

Granules (5)

60

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 13 cont.

Clustered (6)

Speckled(8)

Non existing texture

Based on the mutual relationships of the leucocratic and melanocratic parts, seven textural types were identied (Table 13). With an increasing melanocratic content (and magnetite grade), they are: 1) disseminated, 2)
banded, 3) veiny, 4) patchy, 5) granules, 6) clustered and 8) speckled. The magnetite grains in the leucocratic
parts are smaller in all textural variants than in the melanocratic parts. This is shown in Fig. 22 where the grain
size of magnetite of seven different textural variants is shown. The average grain sizes of magnetite are shown
both for ne grain leucocratic and melanocratic coarser grain material.The magnetite grains are 40-60 microns
larger in the coarser melanocratic parts which are also richer in magnetite. In both parts the magnetite grain
size increases with magnetite grade, and in addition the overall magnetite grain size increases from type 1 to 7
(Fig. 22). The texture type 2 is deviating from the linear correlation and thus shows an uncertainty when the
Fe-grade is low.
Besides the mutual relationship of the leucocratic and melanocratic in the Fsp GEM-type, it can also be divided based on the grain size. The melanocratic parts do have a coarse grain Fsp GEM-type variant, where
the magnetite occurs as single grains (90-133 m) or as larger aggregates interstitially in grain boundaries of a
granoblastic silicate matrix.The grain shape of the magnetite varies strongly; elongated and somewhat rounded
grains occur. The grain size increases as the Fe-grade increases. In the leucocratic parts the ne grained variant
of the Fsp GEM-type magnetite varies from 44 to 75 m. Anhedral magnetite grains occur both at the grain
boundaries of a granoblastic to grano-porphyroblastic silicate matrix and as inclusions in albite and orthoclase
grains. The grain size of the magnetite grains increases as the Fegrade is increasing between the different textural types. It means that the leucocratic part is decreasing in the samples as the melanocratic parts increase due
a higher Fe-grade.
The Bt-(Amph-Ap) GEM-type is structurally controlled and is located in the footwall of the massive ore.
Texturally, the Bt-(Amph-Ap) GEM-type is unique showing a schistose appearance due to the orientation of
biotite. Tabular biotite grains, mainly (10-400 m), sometimes as larger grains up to 2000 m occur, disseminated or as a cluster that follows the lineation in the host rock either interstitially to a magnetite- or a silicate
granoblastic matrix (Table 14).
The Py bearing Amph-(Ap-Bt) GEM-type contains a signicant amount of sulphides. Single euhedral grains of
pyrite and chalcopyrite (5-200 m), occasionally (up to 4800 m), occur as dissemination in veins, at the grain
boundaries or interstitially to magnetite, and this texture occurs both in the massive ore but more commonly
in the semi-massive ore.The composition of this GEM-type regarding the mineral texture and the modal mineralogy is mostly overprinting the Amph-(Ap-Bt) GEM-type.

61

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 14 Different ore textures that are identied and dened in massive and semi-massive ore. This table can also be
used as a mini atlas during texture classication.
Fine grain clustered (9)

Amph-Px

Mag

500 m

Subhedral to anhedral magnetite grains

(56-260 m) occur in a granoblastic matrix with amphibole-pyroxene
Anhedral ne grain amphibole (25180 m) occurs interstitially to magnetite
grains, either with or without the inclusion of magnetite grains (22-70 m)
The grain size of the matrix is medium
to coarse

Coarse grain clustered (10)

Single subhedral to anhedral magnetite
grains (10-200 m) occur interstitially in
grain boundaries and as several inclusions
in coarse or ne grained amphibole-pyroxene
The coarse amphibole grains exist as a
cluster (800-3200 m) or as ne single
grains (80-100 m) interstitially to magnetite
The grain size of the matrix is medium
to coarse

Mag

Amph-Px

500 m

Schistose (11)
500 m

Tabular single biotite grains (10-400 m)

occur as larger grains (up to 2000 m),
even a cluster or aligned in lineation interstitially to magnetite
The matrix is either a granoblastic magnetite or a silicate matrix
The grain size of the matrix is medium
to coarse

Bt
Fsp

Mag

Granoblastic texture (12)

Mag

0.5 mm

Anhedral to subhedral magnetite grains

occur either as a ne grained matrix
(100-400 m) or as a medium grained
(400-800 m) matrix
The magnetite matrix forms a dense texture sometimes giving a complex appearance of complicated grain boundaries
Other mineral associations are identied
to have a porphyroblastic texture, such as
magnetite-apatite, magnetite-ilmenite

Two different GEM-types exist in the massive ore: Amph-(Ap-Bt) and Ap-(Amph). The Amph-(Ap-Bt) type
shows wide variation in the amphibole (-pyroxene) grain sizes and grain shapes, and two end members can be
identied. The coarser grained variant consists of large recrystallised amphibole (-pyroxene) grains occurring
as a cluster (0.8-3.2 mm) enclosing smaller magnetite grains (Table 14). In the ner grain variant, amphibole
grains (0.2-1 mm) occur interstitially to magnetite grains, either with or without the inclusion of small magnetite grains. In both types the magnetite grains are similar in size; the largest difference between them is that

62

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 14 cont.
Porphyroblastic texture (13)
0.5 mm

Mag

Euhedral to subhedral magnetite grains

(800-1400 m) can have an elongated shape outlining a mineral lineation
parallel to the general fabric of the rocks
or occurring either as porphyroblasts
(up to 4600 m) or outlined in veinlike
structures in the ner anhedral magnetite
matrix
The grain boundaries are simple and
straight with clear triple junctions causing
the rock to be brittle

Mag

Euhedral to anhedral (200-2000 m)

single hematite grains or as larger porphyroblasts (up to 20 mm) interstitial to
magnetite grains
Other mineral associations are identied as
having a porphyroblastic texture such as
magnetite-apatite, magnetite-ilmenite

Hem

0.5 mm

0.5 mm

Exsolution texture (14)

Mag

Hem

Magnetite grains show oriented exsolved

spinel patches (50-200 m)
Oxidation surfaces of hematite in magnetite grains or along the grain boundaries in
the magnetite matrix

0.5 mm

Lamella texture (15)

Mag

Magnetite grains show intergrowth of

ilmenite as oriented lamellas (trellis and
sandwich) approx. (2 m)
Homogenous hematite grains have inclusions of (20-100 m) Fe-Ti lamellas or as
(10-50 m) rutile needles at the rim of
the grain

Ilm

25 mm

in the coarse grained variant the magnetite grains also occur as inclusions in amphibole, whereas in the ner
grained variant, the magnetite occurs in the grain boundaries. These two textures can easily be identied during the drill core logging due to the signicant difference in amphibole grain size (Table 14). The Ap-(Amph)
type is texturally homogeneous. Equally sized apatite and magnetite are closely associated (Table 15).

63

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 15 Some specic textures of selected minerals in the Malmberget deposit.
Dissemination of sulphides in magnetite
Single euhedral grains of pyrite and chalcopyrite (5-200 m), occasionally (up to
4800 m) occur as dissemination in veins,
at the grain boundaries or interstitially to
magnetite
The pyrite is more common than the
chalcopyrite and is a texture that occurs
in the massive ore but more common in
the semi-massive ore
The grain size of the matrix is medium
to coarse

Mag

Py
Ccp

500 m

Porphyroblastic texture of apatite and magnetite

Apatite exists as single euhedral grains
(100-800 m) interstitially to magnetite
or as elongated (800-4600 m) grains
aligned in a linear direction
The grain size and shape of apatite is
always correlated to the size and shape of
the coexisting magnetite
Simple and straight grain boundaries in
the magnetite matrix
The grain size of the matrix is medium
to coarse

Mag

Ap

500 m

Fine grain texture of magnetite and feldspar

Mag

Fsp

The ne grained magnetite grain size

varies between 44 to 75 m
Anhedral magnetite grains occur both at
grain boundaries of a granoblastic to grano-porphyroblastic silicate matrix and as
inclusions in albite and orthoclase grains
The grain size of the matrix is ne to
medium

500 m

Coarse grain texture of magnetite and feldspar

500 m

Fsp

Mag

Coarse grain magnetite occurs as single

grains (90-133 m) or as larger aggregates
interstitially in grain boundaries of a
granoblastic silicate matrix
The grain shape of the magnetite varies
strongly, from elongated to rounded grains
occur
The grain size of the matrix is medium

The combinations of the different mineral/ore textures are unique for almost each GEM-type, but what is
common that each of them has both a ne and a coarse grained variant (Table 16).

64

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 16 The iron ore is classied into several classes and subtypes until unique categories of the mineral information
is distinguished to be used in a geological model. The Fsp GEM-type is used for detailed mineral texture analysis.
Classied iron ore
(see section 1.1.1)

(2) Fabian

Ore type
(see section 5.1.2)

Sub-ore type
(see section 5.3.1)

GEM-type
(see section 5.3.3)

Textures
(see table 14-15)

Semi-massive ore

Feldspar rich

Fsp

1-8, 11

Massive ore
Semi-massive ore

(3) Printzskld
Massive ore

Amphibole rich

Amph-(Ap-Bt)

9-10, 12-15

Apatite rich

Ap-(Amph)

12-15

Feldspar rich

Fsp

1-8, 11

Biotite rich*

Bt-(Amph-Ap)

11-13

Apatite rich

Ap-(Amph)

12-15

*This sub-ore type not sampled

The textural study shows that when the mineral textures are considered the modal GEM-types divide into
numerous types, and the closer one looks the more complicated and numerous the classes become (Table 16).
In a geometallurgical context the use of classes is problematic since the treating of non-numeric data in block
modelling is challenging. Therefore, to change the geological model from descriptive to practical mineral textures must be changed from qualitative to quantitative.Therefore, the mineral textures are now considered from
the mineral processing viewpoint.
In mineral processing, ore is comminuted to liberate the minerals and to make the particle size suitable for
downstream processes. Comminution is an energy intensive stage, and therefore a good balance is aimed between mineral liberation and throughput (i.e. used energy/feed tons). Full liberation is not a feasible target
since separation efciency of downstream processes tends to decrease toward very ne particle sizes (<20
microns). Therefore, after the comminution stage the targeted degree of liberation for the ore minerals is typically 90-95% and different kinds of composite particles are present. A term liberation distribution is used here
to summarise the information on mineral deportment; thus, it describes the distribution of particles by their
composition (cf. particle size distribution).
In comminution, the ore is broken into particles using multiple stages of reduction, such as blasting, crushing
and grinding, and it is also classied, with, for example, screening and hydrocycloning.The behaviour of an ore
in the comminution is dependent on machine parameters, such as nature and magnitude of the applied comminution energy (unit operation properties and operational parameters), and on the materials physical properties,
such as elasticity, hardness and strength. These together will dene in a given process the relationship between
specic energy (energy/mass) and overall size distribution of the material. The term grindability describes this,
and it is the measure of the specic energy consumption required to reduce a certain mass of material from a
given fresh and initial size down to a dened product size by grinding. Similarly, the term crushability is used
for crushing.
Comminution circuits are designed and operated to provide targeted product neness, and almost without
exception this gure is xed or changed very seldom. However, it is common that the grain size and association
of the ore minerals vary within the ore body and therefore in plant feed on a daily basis. A good geometallurgical model should therefore not only forecast the metallurgical response but also give the best operational
parameters, i.e. the target liberation degree and accordingly the target grinding neness for any given rock unit
or plant feed (blend).

65

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The liberation distribution of a comminuted ore is dependent not only on (1) nature and magnitude of applied
comminution specic energy and (2) crushability and grindability, but also on (3) modal mineralogy and (4)
mineral textures. The rst factor is independent of material. The three other factors cannot be separated fully
from each other since the physical properties controlling the comminution behaviour are highly dependent
on modal mineralogy and mineral texture. Since there are established techniques to experimentally determine
the crushability and grindability and to use them in process simulations, the third factor is taken as an isolated
parameter, and it gives together with the rst parameter the overall particle size distribution of the material
after comminution.
How does one decouple the effect of the modal mineralogy and texture into the liberation distribution?
Firstly, the mineral grain size is basically a pure textural property, but as shown above the magnetite grain size
has a positive correlation with the magnetite grade (Fig. 22). The associating mineral is controlled both by the
modal mineralogy and the texture. In the Fsp GEM-type the association of magnetite in the melanocratic and
leucocratic parts is different (Table 12). These examples show that the modal mineralogy and the mineral
textures are intimately mingled with each other, and their separation using a traditional textural denition is
practically impossible. Therefore, a new denition for mineral textures is given (Paper IV):
Two samples are texturally different if the liberation distribution by size (compensated against modal mineralogy)
is different after comminuted in similar conditions.

This separates the comminution properties and the modal mineralogy from the textural properties as well as
states that for the textural classication, the liberation distribution must be compensated against modal mineralogy and studied by size.
Textural archetypes

The liberation distribution of three different GEM-types from the two different ore bodies was studied: The
Fsp GEM-type and the Ap GEM-type from both the Fabian and Printzskld ore body and the Amph GEMtype from the Fabian ore body. Each of these samples included different textural types (Table 16).
All of these samples show a variation in the modal mineralogy by size fraction as illustrated by the magnetite
grade by size in Fig. 23. This shows that the breakage is selective, and an attempt to forecast the liberation distribution from uncrushed samples by applying the random breakage model will fail. Also, the grade distribution
varies by ore type. The Fsp GEM-types show an increase in the magnetite content by particle size. A similar
pattern is found in the Ap GEM-types with an exception that the nest size fraction has the highest magnetite grade. The Amph GEM-type shows the highest magnetite content in the middle size fractions, i.e. 75-150
and 150-300 microns. The variation in the mineral grade by size gives the rst challenge when comparing the
liberation distribution (Fig 23).
The degree of liberation of magnetite decreases in all samples by particle size (Fig. 23), which is logical and
found almost always in ore samples. However, the overall liberation degree (of magnetite) can be different in
the bulk sample even if the liberation degree by size is similar, if the overall particle size distribution if different. Therefore, the degree of liberation from the textural point of view should be studied by size. The Fsp
GEM-types from Fabian and Printzskld show that clearly (Fig. 23). They have identical liberation degrees for
magnetite in four size fractions, but in the size fraction 425-600 microns, the Fabian samples show much better
liberation. This causes the overall liberation degree to be different.
66

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
100

Magnetite grade (wt.%)

Magnetite liberation degree (%)

Fa_Fsp

90

Pz_Fsp
Fa_Amph

80
70
60

Fa_Ap

50

Pz_Ap

40
30
20
10
0
38-75 m

Fa_Fsp
Pz_Fsp

100
90
80
70
60
50
40
30
20
10
0
Bulk

75-150 m 150-300 m 300-425 m 425-600 m

Size fraction

38-75 um

75-150 um 150-300 um 300-425 um 425-600 um

Size fraction

Fig. 23 The magnetite grade plotted against the particle size in ve different GEM samples from Fabian (Fa) and
Printzskld (Pz) (left). The degree of liberation for magnetite by size fraction in the Fa_Fsp and Pz_Fsp samples.

In order to compare the liberation distribution of ve different samples, a characteristic size fraction, 150-300
microns, was selected. The magnetite grade in 150-300 microns size fraction is close to the bulk sample; the
mass proportion in crushed sample was high enough and the number of particles measured gives sound statistics.The degree of liberation of magnetite has a positive correlation with the magnetite grade, but the Fsp-type
differs signicantly from the Amph- and Ap-types. The magnetite liberation is better than the grade would

Liberation degree og magnetite (%)

suggest (Fig. 24A).

Pz_Ap

3.5

Association index (AI)

95
94

Fa_Ap

Fa_Fsp

93
92

Fa_Amp

91
90

Pz_Fsp
50

60

70

80

90

2
1.5
1
0.5
0

100

Fsp

38-75
150-300 m
425-600 m

Associatin index

7
6
5
4
3
2
1

AI Mgt-Fsp

AI Mgt-Bt

AI Mgt-Amp

Bt

Amph

10.0

1.0

0.1

AI Mgt-Ap

Ap
38-75 m
75-150 m
150-300 m
300-425 m
425-600 m

100.0

75-150
x
m
300-425 m

Magnetite deportment

Pz_Fsp
Fa_Ap

2.5

Magnetite grade (wt%)

Fa_Fsp
Fa_Amp
Pz_Ap

96

Liberated

Binary with Fsp

Binary with Bt Binary with Amp

Fig. 24 A) The degree of liberation for magnetite versus the magnetite grade of three different GEM-type textures (Fsp,
Amph and Ap) representing the size fraction, 150-300 microns (top left). B) The association index (AI) for magnetite
in size fraction 150-300 microns in the same GEM-type texture (top right). C) The association index of magnetite in
sample by size from Fa_Fsp fractions. D) The deportment of magnetite in sample by size fractions from the Fabian Fsp
GEM-type.

The comparison of mineral association for non-liberated particles is challenging. If just comparing the mass
proportion of magnetite with for example albite, the gure is affected by the magnetite liberation and the albite
grade. If the liberation degree is high, the association of magnetite with albite must be low, and similarly if the
albite grade is low the mass proportion of the binary magnetite-albite grains will also be low. Therefore, a new
Association Index (AI) was developed. It aims to describe how common it is to nd the target mineral locked
with the other minerals regardless of the liberation degree and the modal composition. The association index
is calculated for each mineral pair using the following formula:
AIAB =

Mineral A deportment with mineral B (when fully liberated grains are excluded) [wt%]
Mineral B grade (in a phase excluding mineral A) [wt%]
67

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

If the association index is 1, then the association of mineral A with B is as common as the modal composition
would suggest. If the association index is greater than 1, then the association is more common than expected,
and if it is lower than 1 then it is rarer than expected. Index value zero shows that there is no association between minerals A and B.
Examples of the calculations of the associate index are given in Table 17. The three rst factious samples show
a similar association index with different modal compositions. Sample 4 shows that even if the deportment
percentages are equally high for phase A with B and C, the association index shows that A can be found relatively more often with phase C than with B. Sample 5 shows that the association index cannot be calculated
for binary systems (C=0).
Table 17 Calculation examples for the association index. For the calculation modal [(1)-(3)] and deportment [(4)-(7)],
the analysis result is needed.
Sample

(Row)

Formula

Modal composition (wt. %)

A

50

40

50

50

50

(1)

25

30

40

40

50

(2)

25

30

10

10

(3)

Deportment of the target mineral A, mass proportion (wt. %) of A in different particle classes
Liberated A

90

80

90

90

90

(4)

In binary A-B

10

10

(5)

In binary A-C

10

(6)

In ternary A-B-C

(7)

Mass proportion of minerals in a fraction (phase) excluding the target mineral A (wt. %)
B

50

50

80

80

100

(8)

100*(2)/[(2)+(3)]

50

50

20

20

(9)

100*(3)/[(2)+(3)]

Deportment of A excluding the liberated class (wt. %)

with B

50

50

80

50

100

(10)

100*[(5)+(7)]/[100-(4)]

with C

50

50

20

50

(11)

100*[(6)+(7)]/[100-(4)]

Association index for A

with B

1.000

1.000

1.000

0.625

1.000

(12)

(10)/(8)

with C

1.000

1.000

1.000

2.500

(13)

(11)/(9)

The association indexes of magnetite with feldspars (albite+orthoclase), biotite, amphibole and apatite are
shown for the selected size fractions of ve different textural types from the Malmberget iron ore in Fig. 24b.
The Fsp GEM-types of Fa and Pz show quite similar association indexes. The association index of magnetite
with feldspar (AIMgt-Fsp) is smaller than 1, which means that magnetite is locked with feldspar less often than
the modal mineralogy would suggest. On the other hand the association index of magnetite with amphibole
(AIMgt-Amph) is higher than 1 showing that the magnetite is met more often locked with amphibole than the
modal mineralogy would suggest.The association index of magnetite with apatite shows a signicant difference
between the Fsp and Amph (and Ap) GEM-types. In the Fsp GEM-types the magnetite is rarely occurring
with apatite, but in the Amph and Ap GEM-types association is more common than the modal composition
would suggest.
Because the association index is calculated for particulate material, it carries information on both the mineral
textures (grain size, shape, associating mineral) and the ore breakage. This is clearly illustrated in Fig. 24c which
shows the variation of the association index by size. If the texture is fully homogenous and the breakage fully
random, the association indexes should be 1 for all minerals in all sizes. Deviation from 1 can be due to hete68

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

rogeneous textural or non-random breakage. The association indexes approach 1 as the particle size gets
coarser, as shown in Fig. 24d for the Fsp GEM-type. This is because in coarse particle sizes the particles start to
be identical in their modal composition and liberation distribution. This point is not reached yet at the 425600 microns particle size for Fsp GEM-type sample of Fabian because the liberation degree of magnetite is as
high as 74% (Fig. 23).
It is not possible to decouple the contribution of the texture and breakage for the association index. Some
more information is needed, and therefore the association index values are compared to the textural description
of original uncrushed ore samples. The Fsp GEM-type consists of lecocratic and melanocratic parts, and the
magnetite and amphibole content is higher in the latter and feldspar grade is higher in the former. Association
index values for magnetite with feldspar (AIMgt-Fsp) lower than 1 and higher values with amphibole (AIMgtAmph) can therefore be explained based on the texture. Addititionally, the variation in the association index
of magnetite with apatite between the GEM-types can be explained from textural differences. In the Amph
GEM-type, apatite preferentially occurs in contact with magnetite, but this is not the case in the Fsp GEMtype. The association index of magnetite with biotite shows strong variation by particle size: in ne particle
sizes the value is very high and decreases by size. Since biotite does not show a preferential association with
magnetite in the original uncrushed ore samples, this must be a breakage feature. Thus, it is signicantly more
common to nd magnetite locked with biotite in ne particle sizes than in coarse sizes, and this is not because
biotite is more common in ne size fractions. Therefore, it indicates rather that magnetite-biotite particles do
not break preferentially through grain boundaries rather the opposite.
Coming to the denition of texture, the Fsp GEM-types of Fabian and Printzskld can represent similar
textures since the liberation degree of magnetite and association indexes do not differ from each other strongly.
The Fabian Amph and Ap GEM-types are also potentially similar in texture, and the Printzskld Ap GEMtype is a textural type of its own. The question remains how small or big the difference between the liberation
distributions and the related key gures used here (degree of liberation and association index) should be to
justify saying that two samples are similar or different in their textures. In addition, another question to be
answered is how this information will be gathered and used in a geometallurgical context?
To extend the modal compensation also into liberation degree, the calculation algorithms of the particle tracking technique were applied (Lamberg and Vianna, 2007). The particle tracking technique is for mass balancing
the liberation data, but for doing that it includes steps where the particles are classied in a systematic way (basic
binning and advanced binning), and the modal composition of the liberation measurement is rened to match
with the modal composition calculated by the element to mineral conversion. Therefore, the liberation data of
ve samples was classied (basic binning) producing an identical particle population in each sample. After this,
the liberation distribution of each sample was forecasted using the textural information from another sample.
For this an algorithm given in Table 18 was used (Lamberg and Lund, 2012). For example, to forecast the
liberation distribution of sample A from sample B (called archetype), the liberation spectrum of the archetype
(sample B) was taken, and it was rened by using the modal composition of sample A.
The results of the magnetite deportment of one pair Fa_Fsp and Pz_Fsp is given in Table 19. The algorithm
forecasts the liberation distribution well in all other size fractions except in the coarsest 425-650 microns. This
is most probably due to the small number of particles measured (about 400).

69

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Table 18 Adjusting the modal composition in generating the particle population based on archetype for
given ore sample or block (Lamberg and Lund, 2012).
The particle population for a given sample is calculated by taking the particle population of the corresponding
archetype. This is now iteratively adjusted, and for the adjustment, a correction factor, k, is calculated for each
mineral (i) in a size fraction before each iteration round:
L

pj, fraction = pj, fraction x Y( r(i)p x ki, fraction )

j=1

Basically the formula above is a ratio of mineral grade from the geological model (M(i)) and the mineral grade
back calculated from the liberation data of the archetype (denominator). p refers to mass proportion of particle in
a size class and x(i) is the mass proportion of mineral in a particle.
The mass proportion of particle j (pj) is recalculated on each iteration round using the correction factor and an
equation:

ki, fraction =

M(i)fraction

n
(p(j)
j=1

x r(i)p )

Using equations given above, the particle mass proportion is iteratively adjusted until the difference between mineral grades of the geological model and archetype has reached the required tolerance.

Table 19 The magnetite deportment measured and estimated using the Pz_Fsp and the difference between the
estimated and measured.
Measured

Estimated using
Pz_Fsp

Difference

Bulk

38-75 m

75-150 m

150-300 m

300-425 m

425-600 m

89.4

96.4

95.2

92.5

85.6

74.3

Mgt-Fsp

6.6

1.4

2.9

4.7

10.8

14.2

Mgt-Bt

1.1

1.4

1.2

0.8

0.8

1.7

Mgt-Amph

2.0

0.6

0.5

1.5

1.8

5.9

Mgt-Ap

0.4

0.0

0.0

0.1

0.3

2.1

Mgt Lib

90.7

96.2

95.9

91.5

87.3

86.7

Mgt-Fsp

7.0

2.1

2.4

5.7

10.8

10.6

Mgt-Bt

0.9

0.6

1.0

1.7

0.1

0.5

Mgt Lib

Mgt-Amph

0.8

0.8

0.4

0.6

1.1

1.2

Mgt-Ap

0.1

0.1

0.2

0.1

0.0

0.0

Liberation

1.3

-0.2

0.8

-1.0

1.7

12.4

Average

0.4

0.2

0.2

0.4

0.4

2.8

Fig. 25 shows the error in forecasting the liberation distribution with all of the combinations of the ve samples. Putting a limit that the average error must be lower than 1%, then the ve samples can be simplied into
two textural archetypes: Fsp and Amph/Ap. One sample from both groups is selected to represent the group,
and this sample is called a textural archetype.
Fa_Fsp
Fa_Amp
Pz_Ap

Average error in magnetite

depormtent, %

100

Pz_Fsp
Fa_Ap

10

0.1
Fa_Fsp

Pz Fsp

Fa_Amp

Fa_Ap

Fig. 25 The average error showing the magnetite deportment in the samples.

70

Pz_Ap

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

5.4. Framework for a metallurgical model

The process model takes the information of the geological model and transfers it to information on the metallurgical performance.When developing a metallurgical model, one needs to answer questions, such as: (1) what
is the purpose of the model? (2) What is the level of the complexity of how material (ore) is described? (3)
And how detailed is the information that the model can provide on a metallurgical performance? In a geometallurgical context the purpose of the model is to return important metallurgical parameters in the resource
(block) model. Typically in the processing plants this would include: throughput, concentrate quality, recovery
of the commodities and the tailing quality. In the case of Malmberget, the metallurgical parameters considered
by the process model are (1): throughput, recovery of iron and concentrate grade in terms of iron, phosphorous
and silica. The tailing quality is not a limiting factor in Malmberget, and there is no need to include that in the
model.
In this study the geological model describes the resource on a mineralogical basis, and the process model
logically uses the same approach. Three different levels can be used (2). The unsized, 1D, level uses minerals on
an unsized basis, which means that there is no information on the particle sizes available. The sized, 2D, level
uses the mineral by size for the material description and for the process models. This enables including particle
size and the metallurgical functions can be size dependent.The liberation, 3D, level uses liberation information.
Therefore, the process must describe the behaviour of particles by their composition and size.
(3) The level of processing details means how detailed the process is described in terms of unit operations and
operational parameters. In the lowest level, one black box represents the whole process and no operational
parameters are included. In the other end of the complexity lies a model where all unit operations are described
and all effective operational parameters are included.
In this study the feasibility of the mineralogical approach is tested using a simple process: the one stage dry
magnetic separation (cobbing) test for the Fsp GEM-type of Fa and Pz ore (Fa_Fsp and Pz_Fsp). The process
model is developed for all three levels, namely 1D, 2D and 3D.
To reach the 2D level of the modal mineralogical, a mass balancing needs to be done on a size fraction level.
The calculation rules are as given in section 5.3.2, and the detailed results can be found in Paper III. The 3D
level was reach by the liberation analysis of the concentrate and tail and was then mass balanced using the
Particle Tracking technique, details in Paper IV.
5.4.1. Comminution - particle breakage model

The particle brakeage model gives the liberation distribution of a sample when the information on corresponding textural archetype and modal composition is given (Fig. 26 and 27). The model converts 1D, i.e.
unsized modal composition, to 3D (liberation distribution) using a textural archetype as a basis and a simple
algorithm in adjusting the liberation data of the archetype to match with the given modal mineralogy. The
textural archetype also includes the information on how the modal composition varies by size. The overall size
distribution model developed in a M.Sc. thesis work by Koch was used.

71

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

100

Mgt=80
Mgt=70

90

Liberation (%)

80

Mgt=60
Mgt=50
Mgt=40

70
60

Mgt=30

50
40
30
20
10
0

Bulk

0-38

38-7

50

75-1

300 00-425 25-600 0-1190 0-1680 0-3000

3
4
60
119
168

150-

Size fraction (m)

Fig. 26 The mass proportion of fully liberated magnetite in different size fractions as the magnetite grade in the ore
varies between 30 and 80 wt%. The bulk refers to combined size fractions.

Magnetite association (.%)

50
40

Mag-(Amph-Px) binaries

Fsp=90
Fsp=75
Fsp=50
Fsp=25
Fsp=10

60

Magnetite association (%)

Mag-Fsp binaries
60

30
20
10
0

Bulk

0-38

38-7

50

75-1

50
40

Fsp=90
Fsp=75
Fsp=50
Fsp=25
Fsp=10

30
20
10
0

0
0
0
-300 00-425 25-600 00-119 90-168 80-300
4
3
6
11
16

150

Size fraction (m)

Bulk

0-38

38-7

50

75-1

0
0
0
-300 00-425 25-600 00-119 90-168 80-300
3
4
6
11
16

150

Size fraction (m)

Fig. 27 The mass proportion of magnetite is shown in a magnetite-feldspar binary (left) and a magnetite-(amphibole+pyroxene) binary (right), as the ratio of feldspar to (feldspar+amphibole+pyroxene) varies between 10% and
90%. The bulk refers to combined size fractions. The magnetite grade is 50 wt%.
5.4.2. Concentration model dry magnetic separation

The process model describes quantitatively the particle behaviour in each unit process stage and therefore
returns the metallurgical response (grades for Fe, Si and P and recovery for Fe) for any given geological unit
(sample, block, domain). In the minerals processing the behaviour of particles is dictated by the particle properties; therefore, the unit models have to include the particle properties, such as size, composition and density.
The structure of the simulator binding up the unit models must be based on particles (Lamberg, 2011).
The unit operations models used in minerals processing can be divided into four types: 1) Comminution
models where particle size distribution changes. 2) Separation models where particles are distributed between
two or more output streams based on their physical properties. 3) Leaching and precipitation models where
the liquid phase is an active component and minerals dissolve and new phases are formed through chemical
reactions. 4) Simple mixers and mass distributors where material is mixed and distributed between the outputs.
In the comminution unit models (grinding mills, crushers), it is possible to use the particle breakage model
described above. Therefore, in the model the forecasting of the liberation distribution and the total size distribution can be decoupled. In the latter the traditional populations balance breakage models can be used (Weller
et al., 1996: Alruiz et al., 2009;Vogel and Peukert, 2003).
Most of the separation and leaching models used in industry are semi-empirical. In minerals processing the
fundamental process model based entirely on physics, chemistry and particle properties are still quite far away
from being practical and accurate enough for everyday use. The development of property based models in
minerals processing requires that particle properties can be measured in different parts of the process. For this

72

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

the liberation analysis is a state-of-the-art technique (Sutherland and Gottlieb, 1991; Gu, 2003; Fandrich et al.,
2007), and X-ray tomography is an emerging method (Miller et al., 2003).
The liberation measurement gives the quantitative information on the particles in a process stream, but for
modelling purposes liberation data must be mass balanced. This is done by the Particle Tracking Technique
(Lamberg and Vianna, 2007), a quantitative description of how different particle types (liberation classes) behave
in single process units and in a full process (see section 4.9).
For the Malmberget concentration process, the rst unit model was developed for the dry magnetic separation
stage, i.e. cobbing. Fig. 28 shows the distribution of minerals on an unsized basis between the concentrate and
tailing for the Fsp GEM-type from Fabian. In a particle size of P80 = 1 mm, about 30% of the material is rejected into the tail (MagTail) with about 6% magnetite losses, i.e. the recovery of magnetite into the magnetic
concentrate is 94% (Fig. 28).

Fig. 28 The processed mineral grade and recoveries in a Mrtsell dry magnetic separation (cobbing) test from for the
Fsp GEM-type of the Fabian ore type.

Studying the behaviour of the minerals by particle size shows that for magnetite the recovery is quite constantly between 92 to 96% in particle size range from <38 microns to >1.68 mm (Fig. 29). All of the gangue
minerals show a similar pattern having the recovery minimum between 38 and 106 microns (Fig. 29). Noteworthy is that the biotite shows higher recoveries than the other gangue minerals, especially in the nest
particle size fraction <38 microns. Also, minerals show a signicant difference in particle sizes coarser than 500
microns; albite shows clearly the lowest recovery, whereas apatite and biotite the highest. Excluding the nest
size fraction, the mineral recoveries can be explained by their association with magnetite (Fig. 29). Therefore,
the different patterns found for each mineral in their recovery by size are due to their liberation distribution
and the association with magnetite.

73

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Mgt
Ab
Bt
Amph
Ap

90

Recovery (%)

80
70
60
50
40
30
20

100

Mineral Recovery %

100

Mineral
Ab
Amp
80
Ap
Bt
60
Mgt
40
20

10
0

10

100

1000

10000

Size fraction (m)

10

20

30

40

50

60

70

80

90 100

Mineral association with magnetite %

Fig. 29 The cobbing test with a sample from the Fsp GEM-type of Fabian shows the recovery of minerals by (left).
The mass proportion of minerals associated with magnetite vs. mineral recovery where ve size fractions between 38
and 600 microns are shown. The fully liberated grains are shown for magnetite (right).

Entering into the liberation level, it is obvious that magnetite rich particles enter into the magnetic concentrate,
whereas particles rich in gangue minerals are found in the tailing (Fig. 20). The particle tracking technique,
however, gives quantitative information on the behaviour of the particles.
Table 20 QEMSCAN pictures of typical particles from the magnetic concentrate and tail, Fabian Fsp GEM-type samples. Brown is magnetite, orange is feldspar.
Ore type

Size fraction
(microns)

Concentrate

Tail

425-600

300-425

Fsp GEM-type,
Fabian ore body
150-300

75-150

38-75

Fig. 30 shows the recovery of magnetite-feldspar binaries into the magnetic concentrate by size and as a
function of the magnetite grade. Interestingly, the recovery curve is upward convex in ne particle sizes and
straightens toward the coarse particles.
The recovery curves are deviating only little from the linear; therefore, for simplicity a linear model was
used. For the simulation of the magnetite separation the split values for each liberated minerals were given
(Table 21). The basic assumption behind the model is that particles with a similar composition and size will
behave in the process in a similar way regardless of from which part of the ore they are derived.The simulation
was done using HSC Sim, and for multiphase particles the software uses the weighted average of the split values
given for the liberated minerals.

74

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Three cases were simulated (Table 22). In the rst one Fa-Fa, the material was from Fabian, and the Fabian process model (split values in Table 21) were used. In the second one called Pz-Fa, the material is the Printzskld,
and the Fabian model is used. In the third simulation, the Fabian sample was used as a textural archetype, and
the Printzskld modal composition was used to generate modelled Printzskld feed. Also in this simulation the
Fabian model was used (Pz(Fa)-Fa). The linear model (Fa-Fa) introduces a small error, the standard deviation
of the difference for grade is only 0.29% and for recovery 1.3% (Fa-Fa). The Fa model succeeds to forecast the

Recovery into concentrate (%)

100
90
80
70

0-38 um

60

38-75 m

50

75-150 m

40

150-300 m

30

300-425 m

20

425-600 m

10
0

10

20

30

40

50

60

70

80

90

100

The mass proportion of magnetite inparticle (wt%)

Fig. 30 The recovery of the magnetite-feldspar binary particles into the magnetic concentrate by size fraction as a function of magnetite grade in the particle.
Table 21 A model for the dry magnetic separation. Split values for minerals by size when occurring fully liberated.
Fraction

Magnetite

Feldspar

Biotite

Amphibole

Apatite

Others

38-75 um

0.961

0.030

0.109

0.035

0.031

0.331

75-150 um

0.965

0.033

0.048

0.064

0.012

0.651

150-300 um

0.971

0.054

0.088

0.090

0.104

0.554

300-425 um

0.971

0.042

0.000

0.148

0.000

0.410

425-600 um

0.987

0.112

0.000

0.371

0.047

0.856

Table 22 The observed grades and recoveries in the magnetic concentrate in the cobbing test. The results are the
sum of the size fractions from 38 to 600 size fractions. Diff: Difference (Sim-Meas), R.Diff: relative difference =
100*(Sim-Meas)/Meas.
SampleModel
Case

Meas

Sim

Diff

R.Diff%

Meas

Sim

Diff

R.Diff%

Meas

Sim

Diff

R.Diff%

Mgt wt%

89.7

90.2

0.6

0.6

87.6

88.9

1.3

1.4

87.6

87.2

-0.4

-0.5

Fsp wt%

5.63

5.28

-0.35

-6.24

7.11

5.33

-1.78

-24.99

7.11

5.59

-1.52

-21.34

Bt wt%

0.85

0.76

-0.08

-9.57

1.24

1.71

0.47

38.14

1.24

2.95

1.71

138.21

Amp wt%

2.14

2.12

-0.02

-0.95

1.83

2.06

0.23

12.70

1.83

2.54

0.71

38.71

Fa-Fa

Pz-Fa

Pz from Fa archetype - Fa

Ap wt%

0.09

0.10

0.01

5.60

0.58

0.85

0.28

47.73

0.58

0.79

0.21

36.33

Fe wt%

65.4

65.8

0.4

0.6

63.9

64.8

1.0

1.5

63.9

63.7

-0.2

-0.3

Si wt%

2.58

2.44

-0.14

-5.33

3.06

2.64

-0.43

-13.90

3.06

3.10

0.04

1.27

P%

0.01

0.02

0.00

7.66

0.09

0.14

0.04

48.08

0.09

0.13

0.03

36.69

Mgt Rec%

94.9

93.0

-1.8

-1.9

91.6

92.6

1.0

1.1

91.6

90.9

-0.8

-0.8

Fsp Rec%

13.0

11.9

-1.1

-8.6

17.0

12.7

-4.3

-25.2

17.0

13.3

-3.7

-21.6

Bt Rec%

30.5

26.9

-3.6

-11.8

8.6

11.9

3.2

37.7

8.6

20.5

11.9

137.5

Amp Rec%

20.0

19.3

-0.7

-3.5

16.9

19.0

2.1

12.4

16.9

23.4

6.5

38.3

Ap Rec%

13.4

13.8

0.4

3.2

4.9

7.2

2.3

47.2

4.9

6.6

1.8

35.9

Std grade

0.29

0.93

0.92

Std Rec

1.3

2.7

5.5

75

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

metallurgical performance for Pz when the liberation information is readily available: the standard deviation
of the difference for grades is 0.93% and for recovery 2.7% (Pz-Fa). When the liberation information is not
available and the feed sample is generated from the textural archetype, the forecast is still reasonably good; the
standard deviation of the difference in grades is 0.92% and the recovery 5.5%. These error levels clearly full
the need that the relative error should be below 5% for technical sampling and estimations.Therefore, the conclusion based on the limited testing is that the approach could be used for producing metallurgical parameters
into the block model.
As the process model is based on the particle properties, the model can be used more widely than traditional
unit models which need to be calibrated if there is a change in size distribution, modal mineralogy or liberation.

5.5. Proof of concept

The two methods that are developed give the modal mineralogy and the mineral textures quantitatively for any
given sample if the chemical analysis and textural class is available. These two pieces of data are the minimum
requirement for the geological model. To derive the parameters that describe the metallurgical performance, a
metallurgical model is required.
To test and demonstrate the feasibility of the geometallurgical framework that was created in this thesis, a small
scale test was carried out using a section from the western part of the Fabian ore body. The routine XRF analyses were collected from the drill holes (see sections 5.1.1 and 5.3.2).
Firstly, the chemical analyses were converted to modal composition using the developed element to mineral
conversion technique. The calculated modal mineralogy is then classied into the geometallurgical ore types
(see section 5.3.3). The distribution of the GEM-types in the selected section correlates well with previously dened geochemical subtypes (see section 5.1.1) (Fig. 31a). The Fsp, Amph-(Ap-Bt), Ap-(Amph) and
Bt-(Amph-Ap) are the main GEM-types, and the Qtz bearing Fsp+Amph GEM-type is subordinate as this
represents inclusions of quartz bearing country rock which generally is rare in the massive ore but may be
important to know due to its hardness.The untested Bt-(Amph-Ap) is still a probable GEM-type as the biotite
has been shown to have its own mineral properties. One preliminary GEM-type Py bearing (Amph-(Ap-Bt) is
referred to as a sub GEM-type. This one is mainly overprinting the Amph-(Ap-Bt) GEM-type, and the model
could therefore be simplied only with the extra information of sulphide rich sections. The Amph+Ap+Bt
GEM-type is probably a mix of the other GEM-types.This GEM-type is located close to the host rock contact
and is most likely responsible for a varying metallurgical response. The GEM-types units show dense variation
close to the contact to the host rock, but they seem to form more voluminous units in the inner part of the
ore section (Fig. 31b).
From the Fabian ore body, three different GEM-types have been tested. It means that a mineral composition
of all the included minerals, element to mineral conversion (EMC) rules and a textural archetype library are
available for each GEM-type. To be able to simulate the metallurgical response some of the GEM-types need
to be combined. At this point the Amph-(Ap-Bt), Py bearing (Amph-(Ap-Bt) and Amph+Ap+Bt are merged
into one GEM-type, namely Fa_Amph. The Fsp and Qtz bearing Fsp+Amph are merged into Fa_Fsp, and the
Ap-(Amph) and Bt-(Amph-Ap) are merged into a Fa_Ap GEM-type. The revised GEM-type model is shown
in Fig. 31c.

76

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

100 m

6496

6463

Ti/V

6533

6500

100 m

6496

6463

Fsp

P
100

6500

100 m

6496

6463

6533

6500

Fa_Fsp

Amph (Ap-Bt)

Fe

6533

Fa_Amph

Ap (Amph)

Fa_Ap

100

100

Bt (Amph-Ap)

200

Amph + Ap + Bt
Qtz bearing
Fsp + Amph

200

300

300

300

400

400

400

500

500

500

600

600

600

700

700

700

800

800

900

200

800

900

900

Fig. 31 A cross-section model of the western part of the Fabian ore body which shows the chemical, mineralogical and
metallurgical properties of the ore. A) The chemical variation (Ti,V, P) of different subtypes in the ore. B) A preliminary
GEM-type model is developed based on the model mineralogy. C) A revised GEM-type model now includes the modal
mineralogy, element to mineral conversion (EMC) rules and a textural archetype library for each GEM-type.

A process simulation of GEM-types for the block model

In the process simulation of the revised GEM-types, the liberation distribution of each sample is calculated using the corresponding textural archetype and the modal mineralogy of the sample. The simulation of
the mineral process is then performed on each sample giving the metallurgical response (see section 5.4).
The simulation shows that the metallurgical response is dependent on the modal mineralogy but also on the
texture. The iron recovery has a positive correlation with the head grade (iron and magnetite), but the slope
differs between the textural types (Fig. 32a). As already discussed the Amph and Ap are quite similar texturally,
and this can also be seen in their metallurgical performance (Fig. 32a). Projected silica and phosphorous contents of the concentrate show similar spread between the textural types (Fig. 32b-c). The distribution of silica
in the concentrate is more dependent on the texture than the mineralogy of the feed (Fig. 32d).
The proof of concept shows that the developed geometallurgical model is feasible, and the information provided by the geological model can be used without difculties in the process model. The test also shows that
if the metallurgical function is developed without textural information using only modal mineralogy the error
will be signicant.
In the proof of concept, some simplication was used; therefore, the received metallurgical responses are simplied and in reality the trends that are shown in Fig. 32 would be more spread and not so well dened. The
simplications are: the mineral list is reduced to include only ve minerals instead of the established eleven
and the number of textural archetypes is three. The calculation of the liberation uses only ve size fractions,
between 38 and 600 micron.The process simulations are done without the comminution model, and therefore
the model does not give any information on comminution throughput. As the process model only includes one
concentration step, the metallurgical response of the full circuit would be rather different than what the single
dry magnetic separation (cobbing) stage shows (Table 23).

77

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
100

20
Conc SiO2%

Conc Fe Rec%

90

80

70

Architype

10

60
20

30

40
50
Feed Fe%

60

70

1.0

Architype

Fa_Amp
Fa_Ap
Fa_Fsp
0

10

20
30
Feed SiO2%

40

50

100

0.8

80
Si by Amp%

Conc P%

15

Fa_Amp
Fa_Ap
Fa_Fsp

0.6

0.4
Architype

0.2

0.0
0.0

25

0.4

0.6

0.8 1.0
Feed P%

1.2

1.4

40
Architype

20

Fa_Amp
Fa_Ap
Fa_Fsp
0.2

60

1.6

Fa_Amp
Fa_Ap
Fa_Fsp
0

10

20
30
40
Feed Amp%

50

60

Fig. 32 The metallurgical response plotted against feed properties. A) The recovery of iron is dependent on the head
grade but different textural types show clearly different slope. B) Silica content of the concentrate vs. silica in the feed.
C) Concentrate phosphorous vs. feed phosphorous. D) Silica of the concentrate carried by amphibole vs. amphibole
head grade.

Table 23 The demonstration geometallurgical model includes some simplications and some restrictions
that should be considered.
Steps in the modelling of a GEM block model
Element to mineral conversion
Textural archetypes
Size fractions
Comminution model
Concentration model

Restrictions
The mineral list is simplied to ve minerals
Three textural archetypes tested and available in a library
Five size fractions, the ne (<38 microns) and the coarse (>
600 microns) particle sizes are excluded
No account to the variation in hardness, throughput etc
Includes only one separation step, coarse magnetic separation
(cobbing / Mrtsell separation)

5.6. A guideline on how to gather geological data for geometallurgical modelling

To ensure that the geometallurgical model is available in the feasibility stage, the systematic collection of the
geological data should be started already in the exploration stage. The guidelines presented here give a recipe
that is valid for any type of ore deposit.The results of the process modelling indicate that it is possible to develop a metallurgical model which takes into account the mineral liberation (Fig. 33). Once these two models are
established, the production model can be used to simulate different production scenarios.

78

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Logging DDH
Lithology, mineralogy,
mineral textures
1.
Ore types
Preliminary
2.

12.

Mineralogical study
Sampling ore types
Mineral chemistry
EMC recipe
Textural archetypes
5.
Metallurgical tests
Ore variability tests
Particle Tracking

9.

3.

Geometallurgical
ore types

4.

Geometallurgical
block model
Geo database
Production model
10.

6.

Process model
Breakage model
Grinding model

7.

11.
Simulate and forecast
the production

8.
Plant Survey
Full scale operation
Several mineral circuits

Fig. 33 A guideline for gathering and processing the geological data that are used in geometallurgical modelling. The
shadow blocks dene the development of the geological model.
Drill core logging

Drill core logging is always the fundamental basis for proper ore characterisation.The challenge in new deposits
is to get an overview of the geological variation within the deposit rather early in the feasibility study stage.
Already at this stage it requires collaboration of geologists as well as mining- and mineral engineers, as the
traditional logging focuses only in building a geological picture of the ore body; within the geometallurgical
context more information needs to be gathered, and this information should be quantitative. This study shows
that the necessary geological information to be collected in logging is: lithology, mineralogy and mineral
textures.
In a geometallurgical context, the geological information needs to be quantitative, and it is a challenge how to
make this systematically, fast, with reasonable costs and with required accuracy (SMIFU, 2011). Preferably, the
methods applied to collect the geological data would be at least partly automated. The techniques available for
drill core scanning include optical imaging, hyperspectral imaging, micro-XRF, microcomputer tomography,
SEM-imaging and laser induced breakdown spectroscopy (Pirard et al., 2007; da Costa et al., 2009; Haavisto
and Kaartinen, 2009; SMIFU 2011). Which one of these techniques that will be applied depends on the mineralogy of the ore body and must be evaluated case by case.
Ore Types

1. Each of the dened and sampled ore types is only preliminary until the metallurgical tests and process modelling conrm its metallurgical meaning (Fig. 33).
Mineralogical study

2. After the logging, the next stage is to establish a sampling scheme for the mineralogical study and the
metallurgical testing. It is important that all signicant ore types are sampled. The sampling set should include

79

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

duplicates and replicates to determine the sampling and analysis errors (Lotter et al., 2011). The included
mineral assemblages need to be analysed by their mineral chemistry to create a reliable mineral matrix of all
phases and their chemical composition. The selected methods depend on the mineralogy, but WDS analysis
is needed for minerals carrying the main commodities. For gold, LA-ICPMS analyses are required due to
the low contents. The mineralogical study must include a part which gives the routine technique to get the
modal composition. The element to mineral conversion technique is a cost effective method, but it needs to
be considered early enough to include all necessary elements in the routine analysis scheme. If the mineralogy
is complex, an additional technique may be used, e.g. Satmagan or quantitative X-ray diffraction. Frequently,
the element to mineral conversion requires some more elements to be analysed than traditional ore resource
estimates which often lack some important elements. For developing the textural archetype library, the textural
classication must be developed, and the samples need to be taken to cover a reasonable modal compositional
range in all texture types.The samples are crushed/ground, sized and the liberation is measured by size fractions.
The textural archetypes are established by comparing which of the textures are unique and on how widely they
can be applied in a modal composition range. Each textural archetype can produce a realistic liberation distribution with a wide range of mineralogy and thus be linked with the metallurgical model. A textural archetype
denes the link between the geological texture and the mineral processing model based on particles.
Geometallurgical ore types

3. The quantitative mineralogical information has been collected, and each ore type can now be named: the
geometallurgical type which includes information on the mineral composition, the chemical composition of
minerals, element to mineral conversion (EMC) rules and a textural archetype library. The liberation distribution of any sample or ore block is now able to be calculated, (arrow 6 in Fig. 33). The number of the geometallurgical types and their properties will be continuously updated as the metallurgical tests generate updated
information.
Metallurgical tests

5.The metallurgical tests are performed on every valid ore type to develop a process model. Instead of running
a large number of variability tests, it is more benecial to maximise the information from a smaller number
of samples and carefully designed tests. The laboratory tests are often simplied, but they lay the basis for the
model and test the theorem that similar particles behave in the process identically irrespective of where they
come from within the ore body. The particle tracking technique is used to track multiphase particles by their
liberation classes and to develop the property based unit process models. The validity of the geometallurgical
types should be estimated throughout the metallurgical tests. Normally the feasibility study requires some pilot
testing, and a geometallurgical model will help a great deal in selecting representative samples; this also saves
costs in the piloting. Similarly, the particle tracking should be used to validate the laboratory scale models and
to nd out scale-up rules.
Process model

7. The unit process models describe quantitatively how the particles behave in each process state. In the comminution models it is possible to decouple energy consumption/throughput, size distribution and liberation
distribution. Conventional crushing and grinding models will give the former mentioned, whereas the textural
archetypes and the breakage model produce liberation distribution. Instead of targeting a xed particle size,
the developed model structure enables making optimisation between throughput, particle size distribution and
degree of liberation. In concentration processes a similar approach is possible. Depending on the modal mineralogy and the liberation distribution, it is possible to create different strategies for different ore types.

80

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
Geometallurgical block model

4. The geometallurgical block model includes geological data and block wise information on the metallurgical
key gures. The geological model gives the elemental composition, modal composition, density, textural type,
(arrow 10 in Fig. 33). The metallurgical model provides parameters, such as concentrate composition, recovery
of the main commodities, tailing quality and plant throughput. Some of these parameters might not be additive
(e.g. throughput), and simple geostatistics may not be valid tools to generate the block values from sample data.
This is an area outside of the scope of the study and further development is needed.
Plant Survey

8. The model needs renement, and after starting the operation the process should be surveyed. This basically
includes sampling, particle tracking and rening of the model parameters, (arrow 9 in Fig. 33). The survey
results can be used in optimising and widening possible bottlenecks in the circuits.
Production model

11. The production model is a tool to be used to manage the production for the best possible result. It includes
the production schedule and economic model with product value and productions costs. 12. It is possible that
these economic parameters are returned block wise into the resource model. The other and probably more
feasible option is to build the production to seamlessly communicate with geological and process models as
well as to run continuously different production alternatives and scenarios.

6. DISCUSSION
This is the rst developed geometallurgical model of the Malmberget iron ore deposit. A comprehensive
characterisation of the host rock and the ore was done to identify the mineralogical features important for the
model.
A framework and two different methods have been developed to handle the mineralogical information in such
way that it is possible to make quantitative mineralogical assumptions and build models based on the metallurgical functions of different geological units that are suitable to be used in the mineral processing circuits.
What does it really mean to develop a geometallurgical model, both to be generic and designed for the Malmberget deposit? Does it change anything, on a long term or a daily basis in mineral processing? Does it contribute with knowledge to other related activities in the mining process and is it even practical to use. These
questions are discussed in the following three headings.

6.1. The advantage of a strategic method for geometallurgical modelling in the Malmberget deposit
The potential benets of establishing a geometallurgical model for an ore deposit are several: it lowers the risk
in operation, optimises the use of the ore resource and creates the possibility to plan the production in the long
run (Lamberg, 2010). These benets can be fully utilised only if the geometallurgical model is already available
in the feasibility study stage.
In existing mines such as Malmberget, the benet expectations are more limited. Production scheduling might
be difcult or even impossible to change. Similarly, to run the process in campaigns, i.e. one ore type for certain periods, might not be possible or as a whole not feasible. The benets can therefore come from knowing

81

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

what the limitations are of the material coming in different times. Alruiz et al. (2009) and Suazo et al. (2010)
developed a predictive geometallurgical model for the Collahuasi copper. The models are able to forecast the
throughput and copper recovery on a daily basis. Knowing the realistic daily targets can improve the performance in the long run. If the personnel do not know the variation of the plant feed in times when the feed
is difcult, they may attempt everything to obtain an average result and introduce unnecessary continuous
unbalance in the process. On the other hand when the plant feed is of good quality, they may be satised with
an average result.
In the Malmberget deposit, the mineralogical information and a geometallurgical program could, besides the
daily targets, give solutions to the so-called silica problem (Adolfsson and Fredriksson, 2011).The silica content
of the concentrate varies widely and has no direct correlation with the silica content of the feed. This study
indicates that there is a mineralogical control for the silica content of the concentrate.The control is partly due
to the modal mineralogy and partly due to the mineral textures. Unrevealing this could potentially give better
control of the silica content of the iron concentrate.
Another benet which is often overlooked is that it pushes different professionals, such as mine geologists, mining engineers and plant metallurgists to work together. By increasing the understanding of the resource, mining
and processing in the whole production chain has the potential to increase the resource efciency.

6.2. The geometallurgical model based on a mineralogical approach

The stepwise development of the geological model for Malmberget iron ore (Table 24) relies on the qualitative data provided from drill cores and ore samples. The model is based on the two developed techniques that
produce the mineral grades and mineral textures in such way that a metallurgical prediction can be made. The
requirements for this model to be useful are the capacity to handle large amounts of data fast, be inexpensive
and be of a practical applicability.
Table 24 The stepwise progress for the development of a geological model using the mineralogical approach in geometallurgy
Material
(Sample)

Data provided

Developed method

Limitations

Developed classication

Geological model

Metallurgical
validation

Fsp
Qtz bearing
Fsp+Amph

EMC

Mineral
grades

Accuracy,
complex mineralogy

Fa_Fsp

Bt-(Amph-Ap)
GEMtype*

Py bearing
(Amph-(Ap-Bt)
Amph-(Ap-Bt)

Drill
core,
Ore

Qualitative,
chemical
elements

Ap-(Amph)
Developed classication

Limitations

Developed method

Fe-grade

Mineral
textures

Texture
type
Grain size

GEMtype

Fa_Ap

Fa_Fsp
Descriptive,
semi-quantitative

Textural
archetype

Fa_Ap
Fa_Amph

Grain shape

Pz_Fsp

Mineral
association

Pz_Ap

*Preliminary GEM-type only including the modal mineralogy

82

Fa_Amph

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The developed element to mineral conversion technique was shown to be both useful and practical when
the mineral grades were calculated either from the ore samples but also directly from the drill cores.
However, when this method was tested as a general applicable technique and veried against other iron ores,
the Gruvberget and the Leveniemi ore bodies, situated in the Svappavaara ore eld, some drawbacks were
identied. The ore mineralogy in Gruvberget is rather diverse and includes more minerals, such as calcite and
garnet (Frietsch, 1966; Lindskog, 2001). The routine chemical assays did not include analysis of carbon, causing
the number of minerals to be higher than the elements; the case is being underdetermined (Paktunc, 1998;
White, 2008). A combination of the XRF and the quantitative X-ray diffraction (Rietveld) analyses was used
to test a calculation solution for more complex ore (e.g. Leveniemi ore body) (Lamberg et al., 2013). Also, for
the Gruvberget ore mineralogy, quantitative XRD can give the required additional information for successful
element to mineral conversion.
The use of modal mineralogy to forecast the mineral grades has shown to be a key parameter especially in
geometallurgy classication for some weathered iron ores previously (Paine et al., 2011; Neumann and Avelar,
2012), but a question still remains about how accurate the analysis needs to be in a geometallurgical context.
The error in the element to mineral conversion for Malmberget is too high to be used directly in resource
estimation (Pitard, 1989), but is it enough for geometallurgical modelling? Recent M.Sc. study by Koch (2013)
on Malmberget samples shows better accuracy; the relative standard deviation for magnetite grade estimate is
4-5%. However, for the gangue minerals the error is higher. How good the estimate needs to be for gangue
minerals is difcult to answer directly. One way to nd this out would be Monte Carlo simulations and
studying how much the error of the modal analysis effects the estimate for metallurgical performance, such as
iron recovery and silica grade in the magnetite concentrate.This is outside of the scope of this study, and instead
an approach where the modal mineralogy was used for GEM-type classication where the analysis error does
not play such a signicant role. However, as the metallurgical model uses the modal mineralogy, the question
on analysis accuracy could not be taken away and this remains as future work.
Mineral textures play a vital role in geometallurgy, and this area is studied very little. Hunt (2009), Bonnici
(2009) and Perez-Barnuevo et al. (2012) have developed a method based on optical microscoping, and Hunt
(2012) has used the textural information also in modelling. Using optical microscopy is more cost effective than
using a scanning electron microscope, but still the cost is expected to be per sample higher than for chemical
analysis. The method developed here to establish a library of textural archetypes relies on the assumption that
textures can be classied in drill core logging without the need of optical or electron microscopy for a large
number of samples. Whether this can be done, and how, remains in future work.
Another question which remains unsolved is how to recognise whether two samples are texturally different and
if a new item in the archetype library should be added. Here, the comparison and identication was based on
the degree of liberation by size and the association index (AI).This needs further development. A third question
for textural archetypes is how large the variation in the modal mineralogy one textural archetype can cover
and whether each magnetite ore body in Norrbotten needs textural archetypes of its own. Optical microscopy
indicates that mineral textures are unique for each ore deposit as it reveals, for the Gruvberget iron ore, a ne
grain magnetite texture that is closer to the Kiirunavaara iron ore, but still these two have signicant modal
differences which point toward the need of separate textural archetypes.

83

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

For the Malmberget iron deposit, the systematic and methods of establishing a geological model for geometallurgical purposes is presented in this thesis. Further development is needed in several areas before it really can
apply as a production tool. The proof of concept showed that the developed mineralogical approach basically
works, but several simplications were done in this demonstration.

6.3. Contribution to the genetic interpretation of the Malmberget iron ore deposit
The main objective of this PhD study, the ore characterisation, did reveal some new and important geological
features that were not used further in the geometallurgical model.
The Fabian ore body shows primary features, and it is regarded as a key ore body for unravelling the metamorphic and deformational history of the area. The eastern ore bodies of the deposit (e.g. Fabian) show similarities
to other apatite iron ores. Does this study bring some new information for the debate whether the origin of the
iron ore bodies of Malmberget is magmatic or hydrothermal or a combination of the two? Some key chemical
components like P, V, Ti, Ga are regarded to show primary characters of the ore body and not to have been
mobile in metamorphic events.The similar and strong correlation of the V and Ti content was also found in the
ViRi ore body and suggests a magmatic origin (Hildebrand, 1986; Loberg and Horndahl, 1983; Nystrm and
Henrques, 1994; Naslund et al., 2000; Martinsson 2003). The Ga that follows the patterns of V and Ti might
have a potential to be an indicator element as the Ga is found preferential in titaniferous magnetite in layered
intrusions (Vincent and Nightingale, 1974). The Koskullskulle ore body (Fig. 4), not mentioned earlier, is
geographically closely situated to the ViRi ore body and was described by Geijer (1930).This ore body shows a
similar host rock character, a developed ore breccia around the massive ore and is one of the most P poorest in
the deposit. According to its character and position it could possibly represent the lowest part or a feeder dyke
to the stratigraphically higher situated ore bodies at Malmberget.
The occurrence of the apatite could denitely be a signicant marker to distinguish the source of the mineralisation as many previous studies have shown (Park, 1973; Fritsch and Perdahl, 1995; Rhodes, 1999; Harlov et
al., 2002). The apatite shows an apparent uneven distribution, less common in the eastern part and sometimes
of a pronounced banded appearance in the western part of the deposit (Geijer, 1930).The F dominated content
is similar to the one found in the Kiirunavaara deposit (Park, 1973; Fritsch and Perdahl, 1995; Edfelt, 2007)
which reveals some indications of a magmatic origin (Treloar and Colley, 1996).
The ore breccia that encloses the massive ore bodies is different from the massive ore both in respect to the
textural and mineralogical character but also in the magnetite chemistry. The lower content of trace elements
(Ti, V) in magnetite from ore breccia suggests a hydrothermal origin, which also is supported by the presence of sulphides and anhydrite. This transition from magmatic to hydrothermal conditions producing ore of
different character has been described in the Pilot Knob magnetite deposit (Nold et al., 2013), the Gushan
magnetite-apatite deposit (Hou et al., 2011) and the deposits in Kiruna area (Martinsson, 2004; Billstrm et al.,
2010). Particularly in the structural hanging wall of the Fabian ore body, an ore breccia containing sulphides
with abundant anhydrite in close association with an amphibole-pyroxene skarn is observed.The association of
sulphides to apatite iron ores is seen also in other ore deposits in Norrbotten (Martinsson, 2003; Edfelt, 2007;
Billstrm et al., 2010), and globally (Hou et al., 2011; Nold et al., 2013).
Considering the textural and chemical composition of the coexisting magnetite and hematite, the hematite at
Malmberget was interpreted to be of several generations similar to what has been found at El Laco (Henrques

84

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

and Martin, 1978). The observed chemical differences between different types of hematite at Malmberget are
partly a result of an elemental redistribution between the magnetite and hematite during metamorphic reactions including oxidation (Annersten, 1968). To study the massive hematite ores was beyond the scope of this
thesis, but it is obvious that some of the massive hematite ores show a signicantly high amount of Ti compared
to the magnetite ores. It is therefore suggested that the hematite ore could be of a different origin. Among
other hematite ore bodies in Norrbotten it seems to exist two chemically different variants, a high Ti hematite
(Per Geijer, Ekstrmsberg) and a low Ti hematite (Lappmalmen, Gruvberget). The latter is suggested to be an
oxidation product from magnetite ore as typical oxidation textures is observed between the magnetite and
hematite ore (Frietsch, 1970; Park, 1975), while the former might represent primary hematite ores.
Geological evolution of the Malmberget deposit

Finally to summarise the results and interpretation from the host rock and the ore characterisation, there
is a potential to draw some large scale interpretation of a possible model of the evolution of the deposit in
comparison to the iron ore deposits in Kiruna area (Paper II).
The Malmberget deposit shares many features with the apatite iron deposits in the Kiruna area but there are
also important differences. Some of these might be related to primary differences but most of them are probably
the result of deformation (mineral orientation, internal structure, geometry of ore bodies) and metamorphic
recrystallization (grain size, texture, Fe-Ti oxide paragenesis) of the ore bodies at Malmberget, which have a
higher metamorphic grade.
Primary features may include large scale differences of character, mineralogical and chemical compositions and,
to some extent, differences in the average mineral chemistry of magnetite for the ore bodies at Malmberget.
The massive magnetite ore bodies that are characteristic of the eastern part of Malmberget have mineral assemblages (low to moderate amounts of apatite and amphibole as the main gangue minerals), whole rock chemical
composition (high Fe, Ti,V) and magnetite composition (high Ti,V) that are very similar to the B-type of ore
at Kiirunavaara (Park, 1973; Nystrm & Henriques, 1994; Martinsson, 2003).There are high phosphorous ores
occurring as distinct units within the massive ore at both the Fabian and the Kiirunavaara. In addition Fabian
also exhibits high Ti-V-Ga ore which is low in apatite in several well dened subunits; these indicate that the
massive ore originated from several chemically different injections of magma. These high Ti-V-Ga ore units
occur as several rather regular bodies up to 40 m wide that can be traced between the drill holes at Fabian
(Fig. 31a). At ViRi most of the ore also has this composition. In both deposits minor anhydrite and sulphides
are associated with the high Ti-V-Ga ore indicating a high volatile content of the magma.
In the eastern part of Malmberget, ore breccia typically surrounds the massive ore, although the extent and
mineralogical composition varies slightly between the different ore bodies in this area and also as compared
with similar ore breccia associated with the Kiirunavaara deposit. At Kiirunavaara the occurrence of massive
ore surrounded by ore breccia has been suggested to represent the transition from magmatic to hydrothermal
ore forming processes (Martinsson, 2004) and a similar interpretation is also suggested here for Malmberget,
where magnetite from breccia ore is chemically different from the massive ore at Fabian. The lower content
of trace elements in magnetite and the occurrence of abundant anhydrite together with amphibole-pyroxene
and minor amounts of pyrite-chalcopyrite in ore breccia is a strong indication for a hydrothermal origin. This
is further supported by the IOCG characteristic of the magnetite from the ore breccia in contrast to the AIO
character in the massive ore.

85

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

The ore bodies in the western part of the Malmberget deposit are rather different in character, mineralogical
composition and magnetite chemistry as compared to those in the eastern part. The ore bodies in the western
part are similar to the Per-Geijer ores at Kiruna, particularly regarding their banded character, the occurrence
of hematite, and the high apatite content (Park, 1973). A common feature of these ores is also the minor
occurrence of ore breccia related to the massive ore.
At the Kiruna area the Per-Geijer ores are situated stratigraphically above the Kiirunavaara ore body within the
Kiirunavaara Group in association with felsic volcanic rocks, while the Kiirunavaara ore body itself occurs at
the border of the felsic unit and underlying intermediate rocks (Geijer, 1910; Martinsson, 2004). In Malmberget
the local stratigraphy of the ore host rocks is not well constrained. However, according to preliminary results,
the host rock of the deposits in the western part of Malmberget seems to have a more felsic composition than
does the host rock of the deposits in the eastern part, which has a more intermediate to mac composition.
This might indicate a different, and most likely a higher stratigraphic position, for the Vlkomma and Hens ore
bodies compared to Fabian and ViRi. This is also supported by the decreasing trend of the average V and Ti
content in ores from east to west at Malmberget, which mimics the decreasing content of these elements from
Kiirunavaara to the stratigraphically higher Per-Geijer ores in Kiruna (Fig. 34).This pattern is also shown by the
Ekstrmsberg ore, which is situated within the felsic and upper part of the Kiirunavaara Group and has similar
low Ti-V contents and includes a signicant amount of hematite (Frietsch 1974).
The Printzskld ore body is spatially situated in the middle of the Malmberget deposit and has characteristics
which fall between those of the ore bodies in the eastern and western parts with respect to its mineral composition and to the chemistry of its magnetite.This ore body has a lenticular form, often a strongly banded structure,
contains boudinaged pegmatites parallel to banding and a strong schistosity is developed in the biotite-rich
rocks in the tectonic footwall.This might reect a position of this ore body within or close to a shear zone that
also could have caused a displacement of the ore body from its original position.
The structural and metamorphic evolution during the Svecokarelian Orogeny includes at least two major
events in northern Sweden at 1.9 and 1.8 Ga, but the conditions prevailing during metamorphic recrystallization for each of them is not known (Bergman et al., 2001, 2006). However, in areas like Malmberget that
includes large volumes of 1.8 Ga S-type Lina granites the younger metamorphic event is suggested to represent
peak metamorphic conditions (Bergman et al., 2001) at a temperature of 650-700 C and a pressure of 2-4 kbar
(Annersten, 1968). Stilbite-bearing mineral assemblage at Malmberget indicates that during post metamorphic
cooling the temperature did not exceed 150 C after 1.62-1.60 Ga. (Romer, 1996).
The present character of the Malmberget ore bodies is the result of deformation and strong metamorphic
recrystallization during the 1.9 and 1.8 Ga events that have modied and partly obliterated the primary characters including ore geometry, internal structure, mineral composition, textures and mineral chemistry. The most
obvious change is the increase in grain size and the change in oxidation conditions from the eastern (ViRi,
Fabian) to the western part (Hens,Vlkomma) of the deposit. These variations within the deposit are probably
due to more abundant granitic intrusions and a position closer to the large Lina granite intrusion to the north
for the western part. The interaction between Lina granite and the ore is also reected by the common occurrence of magnetite, hematite and apatite in Lina pegmatites within the ore deposit (Geijer, 1930).

86

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Per-Geijer

Vstra

Mag+Hem+Ap

Mag+Hem+Ap

Mag+Ap
Ryodacite

Ryodacite

P-rich

Dacite-Andesite
Printzskld
Kiirunavaara

Ore Breccia
P-rich

Ore Breccia

Fabian
P-rich

Mag+Amph+Ap

UiRi

Mag+Amph+Ap
Ore Breccia

Andesite-Basalt

Trachyandesite

Mag+Amph+Ap

Fig. 34 A geological model illustrates how the different iron ore bodies interrelations and their stratigraphically displacement could have been in the volcanic pile. The reconstructed Malmberget iron ore deposit (right) is compared to
Kiirunavaara iron ore deposit (left).

7. CONCLUSION
The work presented in this thesis focuses on the solutions of how to develop a geometallurgical model of the
Malmberget iron ore deposit. The development requires an important characterisation stage where the basic
information on geological and mineralogical variation is collected to be used in a geological model. Finally, it
was demonstrated how the developed geological model can be linked with a process model and used in process
management.
The study gives a positive answer to the hypothesis: The geological information collected traditionally by geologists is
not enough to establish a geological model in a geometallurgical context; however, this can be fully done based on a proper
mineralogical characterisation.
Other conclusions of different stages of the study are as follows:
The ore characterisation
Establishing a geometallurgical model requires a proper and correct ore characterisation that gives detailed mineralogical
information and lays the basis for parameters to be included in the model and analysis techniques to be applied. Understanding of the origin and the evolution of the deposit will help to understand, estimate and forecast geological features
in a better way.

87

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
1)

In the Malmberget iron ore deposit, the nature and local stratigraphy of the host rocks to the ore bodies is now
being constructed based on its chemical and petrographic characteristics. Despite a strong metamorphic recrystallisation and an extensive alteration that obliterates most of the primary features, the ore bodies are hosted by
metavolcanic rocks of primarily a similar character as the host rock to the Kiirunavaara deposit. These are metavolcanic rocks (basalt-andesite, dacite-rhyodacite-rhyolite), felsic intrusions (deformed and undeformed aplite,
granite), mac intrusions (metadiabase) and skarn.

2)

Viri, Fabian, Printzskld, Hens and Vlkomma ore bodies show apparent chemical, mineralogical and textural
differences which are specic to each ore body. The variation in whole rock chemistry of the magnetite ore and
the chemical composition of magnetite reect primary signatures especially for V and Ti. In the Fabian ore body,
different chemical subtypes of massive ore are identied based on their varying content of P, Ti,V and Ga. These
chemical subtypes are characterised by a specic mineralogy of apatite, amphibole, sulphides and anhydrite,
which is interpreted to be of a magmatic origin.

3)

The massive ore bodies are, in the eastern part of the deposit, surrounded by a breccia style mineralisation. It is
distinguished from the massive ore by a trace element depleted magnetite that also lacks the intergrowth of the
FeTi-oxide textures. This is particularly seen in the structural hanging wall of the Fabian ore body where there
is an extensive alteration zone of amphibole-pyroxene skarn and an ore breccia, which also contains a varying
amount of pyrite-chalcopyrite and anhydrite. These observations and the data gathered in this study indicate a
formation at a lower temperature.

4)

The Malmberget deposit has been affected by metamorphic recrystallisation and oxidation at temperatures of
at least 550 C and an oxygen fugacity close to the magnetite-hematite buffer. This has caused an elemental redistribution to various extents between the coexisting Fe-Ti oxide minerals. Most signicant is the preferential
partition of Ti into porphyroblasts of hematite. Texturally, the metamorphic overprint/feature is visible as an
increased grain size, developed into a granoblastic texture with distinct triple junctions and simple grain boundaries for magnetite and the presence of different Fe-Ti oxide mineral associations. This indicates a progressive
oxidation from east to west in the deposit with hematite occurring as an important mineral in the western part.

The technical development of quantitative methods

The ore characterisation dened the mineralogical composition of the ore and chemical composition of the
minerals, which is the basic requirement for developing methods to generate quantitative mineralogical information into a geological model.
5)

The mineralogical approach that was selected in this study was relevant as the ore characterisation revealed specic mineralogical features of the magnetite ore. In a geometallurgical context it is necessary to consider these
aspects since silica and phosphorous which are having tight requirements in the nal products are carried by minerals. Magnetite and hematite dominant ores require different treatment in the Malmberget minerals processes.

6)

The developed element to mineral conversion technique provides a practical, fast and inexpensive method to
quantify minerals from routine chemical assays. The mineralogical variation in the Fabian and Printzskld ore
bodies requires an individual mineral list to calculate the modal mineralogy. In addition to the XRF analyses,
SATMAGAN magnetic balance was used to differentiate magnetite and hematite.

7)

The geometallurgical ore type (GEM-type) classication system was developed to overcome the problem that
the accuracy of the modal mineralogy by element to mineral conversions showed, as it was not good enough to
use the mineral grades directly in ore resource estimates.

88

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model
8)

A classication scheme was developed to identify the mineral textures that are present in the ore and it was used
to verify the accurateness of the developed textural archetypes.

9)

Two different textural archetypes are dened in terms of the degree of liberation by size and the association index
(AI).This developed technique describes and distinguishes quantitatively different mineral textures by processing
the liberation data of the included particles without the inuence of the modal mineralogy.

10) Based on the modal composition and the mineral textures, three different geometallurgical ore types (GEMtype) were established for the Malmberget ore body. Each of these GEM-types describes quantitatively: the present mineral and its chemical composition, rules how to calculate the modal composition from routine chemical
assays (element to mineral conversion, EMC, rules) and a textural archetype in a library of archetypes.

The proof of concept

Finally, in order to test the accuracy of the geological model, a metallurgical model was developed and used to
forecast the metallurgical response in terms of grade and recovery.
11) The particle tracking technique using the liberation data from QEMSCAN was shown to be a crucial tool both
when the textural archetypes were developed and when the developed GEM-types were veried and tested for
a metallurgical meaning to the development of a geometallurgical model.
12) The proof of concept showed that it was possible to use the established GEM-types to calculate the modal mineralogy and liberation distribution for a geological unit/sample and receive the metallurgical response from a
process model to develop a block model of the ore.
13) A general valid guideline was developed to give necessary direction on how to gather the geological data for
a geometallurgical model. The process of sampling, testing and modelling the ore (geological model) and the
mineral process circuits (process model) is an iterative process which relies on continuously updated ore and
process parameters. The model can be used to nd the best production options among different scenarios.

8. RECOMMENDATION FOR FUTURE WORK

The following are the recommendations for future work:
1. Regarding answering the remaining question on the genesis of Malmberget iron ore, more information is needed encompassing a deeper study of the ore characterisation, including more ore bodies in
the deposit and using several different analytical methods. LA- ICPMS analyses of the FeTi-oxides can
reveal a wider pattern of the trace element redistribution and uid inclusion of the apatite and stable
isotopes of the sulphide minerals are other signicant geological features that would contribute to the
genetic discussion of the ore forming processes.
2. There should be studies including more geochronological analyses on the host rocks for verifying
the stratigraphy and geological development of the Malmberget area. U-Pb dating of zircons can, in
particular, be useful.
3. In order to solve the stratigraphy, a structural study in the area including the ore deposit is needed
which includes:

89

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

a. A detailed structural mapping of the ore bodies and the host rock.
b. Large scale geophysical characterisation of the structural features of the host rocks and the ore
deposit, e.g. by reection seismic proling and 3D modelling of gravimetric and magnetic
data.
c. Analysis of the petrophysical and rock-mechanical properties of the ore body and the host
rock.
4. The possibility of improving the accuracy of the modal analysis by the combining of XRF analysis
with quantitative X-ray diffraction (Rietveld) should be studied in detail.This kind of work has already
started (Lamberg et al., 2013).
5. Methods how to identify textures already in the drill core logging need to be developed. New drill
core logging techniques like hyperspectral imaging should be tested.
6. To develop the geological model, other GEM-types (e.g. biotite rich ore) require testing and being
veried for metallurgical purposes. As the process model only includes one concentration step, the
metallurgical response of the full circuit is necessary to study as well as the contribution of a grinding
model to the process model; this is an ongoing PhD project.
7. A crucial question for each analysis is how good the analysis needs to be and how goodwith current
techniques it is.This study was able to describe the error of different methods, but more work is needed in dening the level of accuracy needed for different analyses.
8. One important parameter of the geometallurgical model is to forecast the comminution energy and
plant throughput. More work is needed in this area since it was outside the scope of this study.

90

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

9. REFERENCES
Adolfsson, G. and Fredriksson, A., 2011, Reduction of Silica in LKAB Pellets through Different Mineral
Processing Unit Operation: Conference in mineral engineering, Lule, Sweden, 8-9 February, Proceedings,
p. 17-26.
Alruiz, O.M., Morrell, S., Suazo, C.J. and Naranjo, A., 2009, A novel approach to the geometallurgical modelling
of the Collahuasi grinding circuit: Minerals Engineering, v. 22, p. 1060-1067.
Andersen, J. C. ., Rollinson, G., Snook, B., Herrington, R. & Fairhurst. R., 2009. Use of QEMSCAN for the
characterization of Ni-rich and Ni-poor goethite in laterite ores. Minerals Engineering, 22, 1119-1129.
Andrews, J.R.G. and Mika, T.S., 1975, Comminution of a heterogeneous material: development of a model
for liberation phenomena: In Proceedings of 11th International Mineral Processing Congress, Cagliari, Italy,
20-26 April, Proceedings, p. 59-88.
Annersten, H., 1968, The mineral chemical study of a metamorphosed iron formation in northern Sweden:
Lithos, v. 1, p. 374-397.
Arvidson, B., 2013, Kaunisvaara Process Development and Process Plant Implementation: Conference in Mineral Engineering, Lule, Sweden, 5-6 February, Proceedings, p. 31-46.
Batterham, R.J., Grant, R.M. and Moodie, J.P., 1992, A perspective on Process mineralogy and Mineral
processing:The rst International Conference on Modern Process Mineralogy and Mineral Processing, Beijing, China, Proceedings, p. 3-12.
Bergman, S., Billstrm, K., Persson, P., Skild,T. and Evins, P., 2006, U-Pb age evidence for repeated Palaeoproterozoic metamorphism and deformation near the Pajala shear zone in the northern Fennoscandian shield:
GFF, v. 128, p. 7-20.
Bergman, S., Kbler, L. and Martinsson, O., 2001, Description of regional geological and geophysical maps of
northern Norrbotten County (east of Caledonian orogen): SGU Geological Survey of Sweden, v. Ba 56, p.
110.
Bonifazi, G., Massacci, P. and Mati, S., 1990, Textural and structural characteristics of a chromium ore and their
inuence on beniciation: Minerals Engineering, v. 3, p. 137-147.
Bonnici, N., Hunt, J., Berry, R., Walters, S. and McMahon, C., 2009, Quantied mineralogy and texture: Informed sample selection for communition and metallurgical testing: The Tenth Biennial SGA Meeting,
Townsville, Australia, 17th- 20th August, Proceedings, p. 679-681.
Bonnici, N., Hunt, J., Walters, S., Berry, R. and Collett, D., 2008, Relating Textural Attributes to Mineral
Processing - Developing a More Effective Approach for the Cadia East Cu-Au Porphyry Deposit:
Ninth International Congress for Applied Mineralogy, Brisbane, Australia, 8-10 September, Proceedings,
p. 415-418.
Bradshaw, D.J., 2010, Development of a new tool for process mineralogy: Process Mineralogy 10, Vineyard
Hotel, Cape Town, South Africa, 10-12 November.
Braun, G., 1986, Quantitative analysis of mineral mixtures using linear programming: Clays and Clay minerals,
v. 34 no 3, p. 330-337.
Bulled, D. and McInnes, C., 2005, Flotation plant design and production planning through geometallurgical
modeling Publication Series, pp. 809-814.: Australasian Institute of Mining and Metallurgy, v. Publication
Series, p. 809-814.
Bushell, C., 2012, The PGM otation predictor: Predicting PGM ore otation performance using results from
automated mineralogy systems: Minerals Engineering, v. 3638, p. 75-80.
Bye, A.R., 2011, Case Studies Demonstrating Value from Geometallurgy Initiatives: First AusIMM
International Geometallurgy Conference (GeoMet), Brisbane, Australia, 5-7 September, Proceedings,
p. 22.
Cliff, R.A., Rickard, D. and Blake, K., 1990, Isotope Systematics of the Kiruna Magnetite Ores, Sweden: Part
1. Age of the Ore: Economic Geology, v. 85, p. 1770-1776.
Cross, W., Iddings, J.P., Pirsson, L.V. and Washington, H.S., 1902, A quantitative chemico-mineralogical
classication and nomenclature of igneous rocks: Journal of Geology, v. 6, p. 555-690.
da Costa, G.M., Barrn, V., Mendona Ferreira, C. and Torrent, J., 2009, The use of diffuse reectance
spectroscopy for the characterization of iron ores: Minerals Engineering, v. 22, p. 1245-1250.

91

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

David, D., 2007, The importance of geometallurgical analysis in plant study, design and operational phases., v.
Australasian Institute of Mining and Metallurgy Publication Series, p. 241-247.
Debras, C., 2010, Petrology, geochemistry and structure of the host rock for the Printzskld ore body in the
Malmberget deposit: Master thesis, Lule University of Technology, Department of Chemical Engineering
and Geosciences, 49 p.
Dobby, G., Bennett, C., Bulled, D. and Kosick, X., 2004, Geometallurgial modeling The new approach to
plant design and production forecasting/planning, and Mine/Mill Optimization: Proceedings of 36th Annual Meeting of the Canadian Mineral Processors, Ottawa, Canada, 20-22 January, Proceedings, p. paper 15.
Donskoi, E., Suthers, S.P., Fradd, S.B., Young, J.M., Campbell, J.J., Raynlyn, T.D. and Clout, J.M.F., 2007,
Utilization of optical image analysis and automatic texture classication for iron ore particle characterisation:
Minerals Engineering, v. 20, p. 461-471.
Droop, G.T.R., 1987, A general equation for estimating Fe3+ concentrations in ferromagnesian silicates and
oxides from microprobe analyses, using stoichiometric criteria: Mineralogical Magazine, v. 51, p. 431-435.
Dunham, S., Vann, J. and Coward, S., 2011, Beyond Geometallurgy - Gaining Competitive Advantage by
Exploiting the Broad View of Geometallurgy: First AusIMM International Geometallurgy Conference
(GeoMet), Brisbane, Australia, 5-7 Septemberg, Proceedings, p. 1-10.
Edfelt, ., 2007, The Tjrrojkka Apatite-Iron and Cu (-Au) Deposits, Northern Sweden, Lule, Lule University of Technology, 167 p.
Fandrich, R., Gu, Y., Burrows, D. and Moeller, K., 2007, Modern SEM-based mineral liberation analysis:
International Journal of Mineral Processing, v. 84, p. 310-320.
Fandrich, R.G., Bearman, R.A., Boland, J. and Lim, W., 1997, Mineral liberation by particle bed breakage:
Minerals Engineering, v. 10, p. 175-187.
Farrell, J., Miller, A. and Gaze, R., 2011, Geometallurgical Sampling and Resource Estimation for Magnetite Deposits: First AusIMM International Geometallurgy Conference (GeoMet), Brisbane, Australia, 5-7
September, Proceedings, p. 311-319.
Forsmo, S., 2007, Inuence of green pellet properties on pelletizing of magnetite iron ore: Doctoral, Lule,
Sweden, Sustainable Process Engineering, Lule University of Technology, 106 p.
Frietsch, R. and Perdahl, J.A., 1995, Rare earth elements in apatite and magnetite in Kiruna-type iron ores and
some other iron ore types: Ore Geology Reviews, v. 9, p. 489-510.
Frietsch, R., 1984, Petrochemistry of the iron ore-bearing metavolcanics in norrbotten county northern
Sweden: Geological Survey of Sweden, v. Serie C, p. 62.
Frietsch, R., 1974,The Ekstrmsberg iron ore deposit, northern Sweden: Geological Survey of Sweden, v. Serie
C, nr 708, p. 52.
Frietsch, R., 1966, Geology and ores of the Svappavaara area, Northern Sweden: Geological Survey of Sweden,
v. Serie C, p. 282.
Gaudin, A.M., 1939, Principles of Mineral Dressing: New York, USA, McGraw-Hill Companies, 554 p.
Gay, S.L., 2004, A liberation model for comminution based on probability theory: Minerals Engineering, v. 17,
p. 525-534.
Geijer, P., 1930, Geology of the Gllivare Iron Ore eld: Geological Survey of Sweden, v. Ca 22, p. 1-115.
Geijer, P., 1910, Geology of the Kiruna district 2: Igneous rocks and iron ores of Kirunavaara, Loussavaara and
Tuollavaara: Stockholm, Kungliga tryckeriet F. A. Norstedt & Sner, 278 p.
Ghosh, R., 2012, Investigation of Borehole Stability in Malmberget: Lule University of Technology,
Department of Civil, Environmental and Natural resources engineering, 248 p.
Gomes, O., Iglesias, J., Paciornik, S. and Viera, M., 2012, Classication of Hematite Types in Iron Ores through
Circularly Polarized Light Microscopy and Image Analysis: Process Mineralogy 12, Vineyard Hotel, Cape
Town, South Africa, 7-9 November, Proceedings, p. 1-14.
Goodall, W.R., Scales, P.J. and Butcher, A.R., 2005, The use of QEMSCAN and diagnostic leaching in the
characterisation of visible gold in complex ores: Minerals Engineering, v. 18, p. 877-886.
Gottlieb, P., Wilkie, G., Sutherland, D., Ho-Tun, E., Suthers, S., Perera, K., Jenkins, B., Spencer, S., Butcher,
A. and Rayner, J., 2000, Using Quantitative Electron Microscopy for Process Mineral Applications: JOM
Journals of Metals, p. 24-25.

92

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Grip, E. and Frietsch, R., 1973, Malm i Sverige 2: Norra Sverige: Stockholm, Sweden, Almqvist & Wiksell,
295 p.
Gu,Y., 2003, Automated Scanning Electron Microscope Based Mineral Liberation analysis: Journal of Minerals
& Materials Characterization & Engineering, v. 2, p. 33-41.
Gy, P., 1982, Sampling of Particulate Materials - theory and practise: New York, US, Elsevier, 431 p.
Haavisto, O. and Kaartinen, J., 2009, Multichannel reectance spectral assaying of zinc and copper otation
slurries: International Journal of Mineral Processing, v. 93, p. 187-193.
Harlov, D.E., Andersson, U.B., Frster, H., Nystrm, J.O., Dulski, P. and Broman, C., 2002, Apatitemonazite
relations in the Kiirunavaara magnetiteapatite ore, northern Sweden: Chemical Geology, v. 191, p. 47-72.
Helle, S., Kelm, U., Barrientos, A., Rivas, P. and Reghezza, A., 2005, Improvement of mineralogical and
chemical characterization to predict the acid leaching of geometallurgical units from Mina Sur,
Chuquicamata, Chile: Minerals Engineering, v. 18, p. 1334-1336.
Henley, K.J., 1992, A Review of Recent Developments in the Process Mineralogy of Gold: Extractive
Metallurgy of Gold and Base Metals, Kalgoorlie, Proceedings, p. 177-194.
Henrques, F., Naslund, H.R., Nystrm, J.O., Vivallo, W., Aguirre, R., Dobbs, F.M. and Lled, H., 2003, New
eld evidence bearing on the origin of the El Laco magnetite, northern Chile - a discussion: Economic
Geology, p. 1497-1502.
Henriquez, F. and Martin, R., 1978, Crystal-growth textures in magnetite ows and feeder dykes, El Laco,
Chile: Canadian Mineralogist, v. 16, p. 581-589.
Hestnes, K.H. and Srensen, B.E., 2012, Evaluation of quantitative X-ray diffraction for possible use in the
quality control of granitic pegmatite in mineral production: Minerals Engineering, v. 39, p. 239-247.
Hildebrand, R.S., 1986, Kiruna-type Deposits: Their Origin and Relationship to Intermediate Subvolcanic
Plutons in the Great Bear Magmatic Zone, Nothwest Canada: Economic Geology, v. 81, p. 640-659.
Hiltunen, A., 1982, The Precambrian geology and skarn iron ores of the Rautuvaara, northern Finland:
Geological Survey of Finland (GTK), v. Bulletin 318, p. 133.
Hitzman, M.W., Oreskes, N. and Einaudi, M.T., 1992, Geological characteristics and tectonic setting of
Proterozoic iron oxide (Cu-U-Au-REE) deposits: Precambrian Research, v. 58, p. 241-287.
Hou, T., Zhang, Z. and Kusky, T., 2011, Gushan magnetite-apatite deposit in the Ningwu basin, Lower Yangtze
River Valley, SE China: Hydrothermal or Kiruna-type?: Ore Geology Reviews, v. 43, p. 333-346.
Hunt, J., Berry, R., Bradshaw, D.,Triffett, B. and Walters, S., 2012, Development of liberation/recovery domains:
examples from the Prominent Hill IOCG deposit, Australia: Process Mineralogy12, Cape Town, South
Africa, 7-9 November, Proceedings, p. 1-16.
Hunt, J., Berry, R. and Bradshaw, D., 2011, Characterising Liberation and Flotation Potential Using Image
Analysis, Simulated Fragmentation and Small-Scale Flotation: The rst AUSIMM international geometallurgy conference, Brisbane, Australia, 5-7 September, Proceedings, p. 331-333.
Hunt, J., Berry, R. and Bradshaw, D., 2011, Characterising chalcopyrite liberation and otation potential:
Examples from an IOCG deposit: Minerals Engineering, v. 24, p. 1271-1276.
Hunt, J., Berry, R., Bonnici, N., Walters, S., Kamenetsky, M. and McMahon, C., 2009, From Drill Core to
Processing - A Geometallurgical Approach to Mineralogy and Texure From Meso- to Micro-Scale: The
Tenth Biennial SGA Meeting, Townsville, Australia, 17th-20th August, Proceedings, p. 685-687.
Jackson, J., McFarlane, A. and Olson-Hoal, K., 2011, Geometallurgy - Back to the Future: Scoping and
Communicating Geomet Programs: First AusIMM International Geometallurgy Conference (GeoMet),
Brisbane, Australia, 5-7 September, Proceedings, p. 125-131.
JKTech, 2010, JKTechs monthly e-Newsletter. December 2010.
Johansson, B. and Wanhainen, C., 2010, Flotation and leach tests performed within a geo-metallurgical project
on gold in the Aitik Cu-Au-Ag-(Mo) deposit: Conference in Minerals Engineering, Lule, Sweden, 1-2
February, Proceedings, p. 61-72.
Johnson, R.A. and Wichern, D.W., 1998, Applied Multivariate Statistical Analysis, Prentice Hall, 816 p.
Jones, M.P. and Gravilovic, J., 1970, Automatic quantitative mineralogy in mineral technology: Rudy, v. 5, p.
189-197.
King, R.P. and Schneider, C.L., 1998, Mineral liberation and the batch communition equation: Minerals
Engineering, v. 11, p. 1143-1160.

93

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

King, R.P., 1979, A model for the quantitative estimation of mineral liberation by grinding: International
Journal of Mineral Processing, v. 6, p. 207-220.
Koch, P., 2013, Textural variants of iron ore from Malmberget: characterization, comminution and mineral
liberation: Master thesis, Lule, Sweden, Sustainable Process Engineering, Lule University of Technology,
101p.
Kojonen, K. and Johanson, B., 1999, Determination of refractory gold distribution by microanalysis, diagnostic
leaching and image analysis: Mineralogy and Petrology, v. 67, p. 1-19.
Kvarnstrm, B. and Oghazi, P., 2008, Methods for traceability in continuous processes-Experience from an iron
ore renement process: Minerals Engineering, v. 21, p. 720-730.
Kwitko-Ribeiro, R., 2011, New sample preparation developments to minimize mineral segregation in process
mineralogy: Proceedings, 10th International Congress for Applied Mineralogy (ICAM),Trondheim, Norway,
1-5 August, Proceedings, p. 411-417.
Lamberg, P., Parian, M.A., Mwanga, A. and Rosenkranz, J., 2013, Mineralogical mass balancing of industrial
circuits by combining XRF and XRD analyses: Conference in Mineral Engineering , Lule, Sweden, 5-6
February, Proceedings, p. 105-116.
Lamberg, P. and Lund, C., 2012, Taking liberation information into a geometallurgical model-case study,
Malmberget, Northern Sweden: Process Mineralogy12 , Cape Town, South Africa, 7-9 November,
Proceedings, p. 1-13.
Lamberg, P., 2011, Particles the bridge between geology and metallurgy: Conference in mineral engineering,
Lule, Sweden, 8-9 February, Proceedings, p. 1-16.
Lamberg, P. and Tommiska, J., 2009, HSC Chemistry 7.0 Users Guide. Mass Balancing and Data
Reconciliation: http://www.outotec.com/39476.epibrw, Outotec, 1-50 p.
Lamberg, P. and Vianna, S., 2007, A Technique for Tracking Multiphase Mineral Particles in Flotation Circuits:
7th Meeting of the Southern Hemisphere on the Mineral Technology, Ouro Preto, Brazil, Proceedings, p.
195-241.
Lamberg, P., Hautala, P., Sotka, S. and Saavalainen, S., 1997, Mineralogical balances by dissolution
methodology: Short Course on Crystal Growth in Earth Sciences, S. Mamede de Infesta, Portugal,
September 8-10, Proceedings, p. 1-29.
Lane, G.R., Martin, C. and Pirard, E., 2008, Techniques and applications for predictive metallurgy and ore
characterization using optical image analysis: Minerals Engineering, v. 21, p. 568-577.
Lawson, C.L. and Hanson, R.J., 1974, Solving Least Squares Problems, Prentice Hall, 340 p.
Liipo, J., Lang, C., Burgess, S., Otterstrom, H., Person, H. and Lamberg, P., 2012, Automated mineral
liberation analysis using INCAMineral: Process Mineralogy 12,Vineyard Hotel, Cape Town, South Africa, 7-9
November, Proceedings, p. 1-7.
Lindskog, L., 2001, Relationships between Fe-oxide Cu/Au deposits at Gruvberget - Kiruna district Northern
Sweden, Townsville, Australia, James Cook University of Northern Queensland, 123 p.
Lindsley, D.H., 1991, Oxide Minerals: Petrologic and magnetic signicance: Blacksburg, Virginia, U.S,
Mineralogical Society of America.
Lindsley, D.H., 1962, Investigations in the system FeO - Fe2O3- TiO2, Carnegie Institution of Washington
Year Book, 61, 100-106 p.
LKAB, 2013, Annual Report: Lule, Sweden, Lule Graska, 138 p.
LKAB, 2011, Annual Report: Lule, Sweden, Lule Graska, 122 p.
LKAB, 2006, Annual Report: Lule, Sweden, Lule Graska, 118 p.
Lorenzen, L. and Van Deventer, J.S.J., 1993, The identication of refractoriness in gold ores by the selective
destruction of minerals: Minerals Engineering, v. 6, p. 1013-1023.
Lotter, N.O., Kormos, L.J., Oliveira, J., Fragomeni, D. and Whiteman, E., 2011, Modern Process Mineralogy:
Two case studies: Minerals Engineering, v. 24, p. 638-650.
Lund, C., Lamberg, P. and Lindberg, T., 2013, Practical way to quantify minerals from chemical assays at
Malmberget iron ore operations An important tool for the geometallurgical program: Minerals
Engineering, v. 49, p. 7-16.

94

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Lundberg, B. and Smellie, J., 1979, Painirova and Mertainen iron ores: two deposits of the Kiruna Iron Ore type
in northern Sweden: Economic Geology, v. 74, p. 1131-1152.
Martinsson, O., berg, E. and Fredriksson, A., 2012, Apatite for extraction - Mineralogy of apatite and REE
in the Kiirunavaara Fe-deposit: XXVI International Mineral Processing Congress-IMPC 2012, New Delhi,
India, 24-28 September.
Martinsson, O., 2004, Geology and Metallogeny of the Northern Norrbotten Fe-Cu-Au Province: Society of
Economics Geologists, Guidebooks Series, v. 33, p. 131-148.
Martinsson, O. and Virkkunen, R., 2004, Apatite Iron Ore in the Gllivare, Svappavaara, and Jukkasjrvi Areas:
Society of Economics Geologists, Guidebooks Series, v. 33, p. 167-172.
Martinsson, O. and Hansson, K.E., 2004, Day seven eld guide Apatite Iron Ores in the Kiruna Area: Society
of Economic Geology, Guidebook Series, v. 33, p. 173-175.
Miller, J.D., Lin, C.L., Garcia, C. and Arias, H., 2003, Ultimate recovery in heap leaching operations as
established from mineral exposure analysis by X-ray microtomography: International Journal of Mineral
Processing, v. 72, p. 331-340.
Minz, F., Bolin, N., Lamberg, P. and Wanhainen, C., 2013, Detailed characterisation of antimony mineralogy in
a geometallurgical context at the Rockliden ore deposit, North-Central Sweden: Minerals Engineering.
Moen, K., 2006, Quantitative measurements of mineral microstructure: Doctoral Thesis, Trondheim,
Norwegian University of Science and Technology, 194 p.
Mcke, A., 2003, Magnetite, ilmenite and ulvite in rocks and ore deposits: petrography, microprobe analyses and
genetic implication: Mineralogy and Petrology, v. 77, p. 215-234.
Naslund, H.R., Aguirre, R., Dobbs, F.M., Henrques, F. and Nystrm, J.O., 2000, The Origin, emplacement,
and eruption of the magmas: IX Congreso Geologico Chileno, Santiago, Chile, Proceedings, p. 135-139.
Neumann, R. and Avelar, A.N., 2012, Renement of the isomorphic substitutions in goethite and hematite
by the Rietveld method, and relevance to bauxite characterisation and processing: Process Mineralogy 12,
Vineyard Hotel, Cape Town, South Africa, 7-9 November, Proceedings, p. 1-10.
Neumann, R. and Schneider, C.L., 2001, Prediction of monazite liberation from the silexitic rare earth ore of
Catalo I: Minerals Engineering, v. 14, p. 1601-1607.
Niiranen, K. and Bhm, A., 2012, A systematic characterization of the ore body for mineral processing at
Kiirunavaara iron ore mine operated by LKAB, Northern Sweden.: XXVI International Mineral Processing
Congress (IMPC), New Delhi, India, Proceedings, p. 1039.
Nold, J.L., Davidson, P. and Dudley, M.A., 2013, The pilot knob magnetite deposit in the Proterozoic
St. Francois Mountains Terrane, southeast Missouri, USA: A magmatic and hydrothermal replacement iron
deposit: Ore Geology Reviews, v. 53, p. 446-469.
Nordstrand, J., 2012, Mineral Chemistry of Gangue Minerals in the Kiirunavaara Iron Ore: Master Thesis,
Lule, Sweden, Department of Civil, Environmental and Natural resources engineering, 47 p.
Nystrm, J.O. and Henrques, F., 1994, Magmatic Features of Iron Ores of the Kiruna Type in Chile and
Sweden: Ore Textures and Magnetite Geochemistry: Economic Geology, v. 89, p. 820-839.
Paine, M., Knig, U. and Staples, E., 2011, Application of rapid X-ray diffraction (XRD) and cluster analysis
to grade control of iron ores: Proceedings, 10th International Congress for Applied Mineralogy (ICAM),
Trondheim, Norway, 1-5 August, Proceedings, p. 495-501.
Paktunc, A.D., 1998, MODAN: an interactive computer program for estimating mineral quantities based on
bulk composition: Computers & Geosciences, v. 24, p. 425-431.
Park, T., 1973, Rare Earth in the Apatite Iron Ores of Lappland Together With Some Data About the Sr, Th,
And U Content of These Ores: Economic Geology, v. 68, p. 210-221.
Park, T., 1975a, The origin of the Kiruna iron ores: Geological Survey of Sweden, v. C 709, p. 209.
Park, T., 1973, Kirunamalmernas bildning: Doctoral Thesis, Stockholm, Stockholm University, 222 p.
Park, C.F.,Jr, 1961, A magnetite ow in Northern Chile: Economic Geology, v. 56, p. 431-441.
Pascoe, R.D., Power, M.R. and Simpson, B., 2007, QEMSCAN analysis as a tool for improved understanding
of gravity separator performance: Minerals Engineering, v. 20, p. 487-495.
Penttinen, U., Palosaari,V. and Siura,T., 1977, Selective dissolution and determination of sulphides in nickel ores
by the bromine-methanol method: Bulletin of the Geological Society of Finland, v. 49 part 2, p. 79-80.

95

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Prez-Barnuevo, L., Pirard, E. and Castroviejo, R., 2012, Automated textural analysis of the Kansanshi Copper
ore: Process Mineralogy12, Cape Town, South Africa, 7-9 November, Proceedings, p. 1-16.
Petruk, W., 2001, Imaging of minerals, ores and related products to determine mineral characteristics: SME
Annual meeting, Denver, Colorado, U.S.
Petruk, W., 2000, Applied Mineralogy in the Mining: Amsterdam, Elsevier, 268 p.
Philander, C. and Rozendaal, A., 2011, The contributions of geometallurgy to the recovery of lithied heavy
mineral resources at the Namakwa Sands mine, West Coast of South Africa: Minerals Engineering, v. 24, p.
1357-1364.
Pirard, E., Lebichot, S. and Krier,W., 2007, Particle texture analysis using polarized light imaging and grey level
intercepts: Minerals Processing, v. 84, p. 299-309.
Pirrie, D., Butcher, A.R., Power, M.R., Gottlieb, P. and Miller, G.L., 2004, Rapid quantitative mineral and phase
analysis using automated scanning electron microscopy (QemSCAN); potential applications in forensic
geoscience: Geological Society, London, Special Publications, v. 232, p. 123-136.
Pitard, F.F., 1989, Pierre Gys sampling theory and sampling practice Vol. 2, Sampling theory and sampling
practice: Boca raton , US, CRC Press, 247 p.
Potts, P.J., Bowles, J.F.W., Reed, S.J.B. and Cave, M.R., 1995, Microprobe Techniques in the Earth Sciences:
London, UK, Chapman & Hall.
Ramdohr, P., 1980, The Ore minerals and their intergrowth: Berlin, Pergamon press, 1207 p.
Rhodes, A., Oreskes, N. and Sheets, S., 1999, Geology and Rare Earth Element Geochemistry of Magnetite
Deposits at El Laco, Chile: Evidence of Formation from Isotopically Heavy Fluids, in Skinner, B.J., ed.,
Geology and ore deposits of the Central Andes., Society of Economic Geologists, p. 299-332.
Rietveld, H.M., 1969, A Prole Renement Method for Nuclear and Magnetic Structures: Journal of Applied
Crystallography, v. 2, p. 65-71.
Romer, R.L., 1996, U-Pb systematics of stilbite-bearing low-temperature mineral assemblages from the
Malmberget iron ore, northern Sweden: Geochimica et Cosmochimica Acta, v. 60, p. 1951-1961.
Romer, R.L., Martinsson, O. and Perdahl, J., 1994, Geochronology of the Kiruna Irons Ores and Hydrothermal
Alterations: Economic Geology, v. 89, p. 1249-1261.
Schouwstra, R., de Vaux, D., Hey, P., Malysiak, V., Shackleton, N. and Bramdeo, S., 2010, Understanding
Gamsberg A geometallurgical study of a large stratiform zinc deposit: Minerals Engineering, v. 23, p.
960-967.
Schuiling, R.D. and Feenstra, A., 1980, Geochemical behaviour of vanadium in iron-titanium oxides: Chemical
Geology, v. 30, p. 143-150.
Sillitoe, R.H. and Burrows, D.R., 2002, New eld evidence bearing on the origin of the El Laco magnetite
deposit, northern Chile: Economic Geology, v. 97, p. 1101-1109.
(SMIFU), Lamberg, P., Edelbro, C., Elming, S., Osipov, E., Wanhainen, C. and Nilsson, M., 2011, Smart Mine
of the Future Work Package 12, Resource characterisation, Final Report: Lule, Sweden, Rock Tech Centre,
1-13 p.
Stamboliadis, E.T., 2008, The evolution of a mineral liberation model by the repetition of a simple random
breakage pattern: Minerals Engineering, v. 21, p. 213-223.
Stradling, A.W., 1991, Development of a mathematical model of a crossbelt magnetic separator: Minerals
Engineering, v. 4, p. 733-745.
Suazo, C.J., Kracht, W. and Alruiz, O.M., 2010, Geometallurgical modelling of the Collahuasi otation circuit:
Minerals Engineering, v. 23, p. 137-142.
Sutherland, D.N., Gottlieb, P. and Butcher, A.R., 2000, Mineral Characterisation in the 21st century: Minerals
& Metals Challenges Beyond 2000, IIM.
Sutherland, D.N. and Gottlieb, P., 1991, Application of automated quantitative mineralogy in mineral
processing: Minerals Engineering, v. 4, p. 753-762.
Treloar, P. and Colley, H., 1996, Variations in F and Cl contents in apatites from magnetite-apatite ores in
northern Chile, and their ore genetic implications: Mineralogical Magazine, v. 60, p. 285-301.
Vann, J., Jackson, J., Coward, S. and Dunham, S., 2011, The geomet curve: A model for implementation of
geometallurgy: First AusIMM International Geometallurgy Conference (GeoMet), Brisbane, Australia,
September, Proceedings, p. 35-44.

96

Mineralogical, chemical and textural characterisation of the Malmberget iron ore deposit for a geometallurgical model

Vincent, E.A. and Nightingale, G., 1974, Gallium in rocks and minerals of the Skaergaard intrusion: Chemical
Geology, v. 14, p. 63-73.
Vogel, L. and Peukert, W., 2003, Breakage behaviour of different materialsconstruction of a mastercurve for
the breakage probability: Powder Technology, v. 129, p. 101-110.
Weller, K.R., Morrell, S. and Gottlieb, P., 1996, Use of grinding and liberation models to simulate tower mill
circuit performance in a lead/zinc concentrator to increase otation recovery: International Journal of
Mineral Processing, v. 4445, p. 683-702.
Whiten, B., 2008, Calculation of mineral composition from chemical assays: Mineral Processing and extractive
metallurgy review, v. 29, p. 83-97.
Wills, B.A, 1988, Mineral Processing Technology: Exeter, Great Britain, Pergamon Press, Headington Hill Hall,
772 p.
Young, R.S., 1974, Chemical phase analysis: London, Charles Grifn & Co Ltd, 138 p.

97

Character and origin of the host rock to the Malmberget apatite

iron ore, northern Sweden
Olof Martinssona, Cecilia Lunda, Joel Anderssonb and Celine Debrasc
a

Division of Geosciences and Environmental Engineering, Lule University of Technology, 971 87 Lule, Sweden
b

Luossavaara-Kiirunavaara AB (LKAB), 983 81 Malmberget, Sweden

c
Luossavaara-Kiirunavaara AB (LKAB), 981 86 Kiruna, Sweden
* Corresponding author.Tel.: +46 920 491369. E-mail address: [email protected]

(Manuscript)

Character and origin of the host rock to the Malmberget apatite iron ore,
northern Sweden
Olof Martinssona, Cecilia Lunda, Joel Anderssonb and Celine Debrasc
a

Division of Geosciences and Environmental Engineering, Lule University of Technology, 971 87 Lule, Sweden
b
Luossavaara-Kiirunavaara AB (LKAB), 983 81 Malmberget, Sweden
c
Luossavaara-Kiirunavaara AB (LKAB), 981 86 Kiruna, Sweden
* Corresponding author.Tel.: +46 920 491369. E-mail address: [email protected]

ABSTRACT
Apatite iron ores are important sources for iron in Sweden and some other countries with the Malmberget deposit as the second largest one in Sweden. The ore bodies at the Malmberget are hosted by strongly metamorphosed and deformed rocks of felsic to mac composition. These rocks are traditionally called leptites in the
Malmberget area.The purpose with this study is to dene the primary character of the host rocks to the Malmberget deposit and compare them with other similar rocks units in northern Norrbotten.This makes it possible
to get a better understanding of the local geology at Malmberget and also for the genetic aspects of the of the
Malmberget deposit.The results from this study clearly demonstrate that the metavolcanic rocks at Malmberget
belongs to the Kiirunavaara Group and are composed of basalt, andesite-trachyandesite, trachyte-dacite and
rhyodacite. In the metavolcanic rocks and the ores occur intrusions of slightly deformed aplite and metadiabases-monzonite that corresponds to the Perthite Monzonite Suite and are co-magmatic to the extrusive rocks.
Dykes of aplite-pegmatite-granite of the Lina Suite are also common in both the ores and the metavolcaic
rocks and these are generally less deformed and lack evidence of alteration and metamorphic recrystallization
in contrast to the Perthite Monzonite Suite intrusions. The host rocks and the ore bodies might occupy a syncline structure with dominantly mac to intermediate metavolcanic rocks occurring in the outer part of the
structure and hosting the deposits in the eastern part of Malmberget, while more felsic metavolcanic rocks are
found in the core of the structure representing a stratigraphically higher position and hosting the deposits in the
western part of the Malmberget deposit. An alteration of the host rocks is dominated by albite alteration and
amphibole-pyroxene alteration. The latter is extensively developed close to the ore bodies and may at least be
related to aplitic intrusions of the Perthite Monzonite Suite. K-feldspar alteration is less frequently developed
close to the ores and is then mainly formed as a later overprinting related to the Lina Suite intrusions.

1. INTRODUCTION
Apatite iron ores are important sources for iron in Sweden and some other countries. Besides magnetite and
hematite, most of these iron ores contain signicant amounts of apatite and they are generally strongly enriched
in LREE (Martinsson, 2004). This class of deposits has been named apatite iron ores or iron ores of Kiruna
type with the Kiirunavaara deposit being the largest and best-known example. It contains more than 2000 Mt
of high-grade ore and was rst described by Geijer (1910) and interpreted by him to have formed from an iron
oxide melt. The most important deposits are Kiirunavaara and Malmberget that have been in production for
more than 100 years and accounts for more than 90 % of the total iron ore production in EU.
The ores at Malmberget are hosted by strongly metamorphosed and deformed rocks of felsic to mac composition. These rocks are traditionally called leptites in the Malmberget area (Geijer, 1930). However, by petrographic and geochemical methods the precursor to these highly recrystallized and partly strongly altered rocks
have been identied at the Printzskld ore body (Debras, 2010). The purpose with this study is to dene the
primary character of the host rocks to the Malmberget deposit and compare them with other similar rocks
units in northern Norrbotten. This makes it possible to get a better understanding of the local geology at
Malmberget and also for the genetic aspects of the of the Malmberget deposit.

2. GEOLOGICAL SETTING
The Precambrian bedrock in the northern Norrbotten region include a c. 2.8 Ga Archean granitoid-gneiss
basement, which is unconformally overlaid by greenstone, porphyry and sedimentary successions of Paleoproterozoic age. Stratigraphically lowest are rift related 2.5-2.0 Ga Karelian units, which in the Kiruna area are
represented by the Kovo Group and the overlying Kiruna Greenstone Group (Martinsson, 1997).The overlying
c. 1.9 Ga Svecofennian successions comprise the Porphyrite Group, the Kurravaara Conglomerate, the Kiruna
Porphyries and the Hauki Quartzite. Most of these Paleoproterozoic units extend outside the Kiruna area, thus
they are regionally developed in northern Norrbotten. The calcalkaline and andesite dominated Porphyrite
Group is suggested to be subduction related, while the Kiruna Porphyries have a bimodal character and a geochemical signature resembling within plate volcanics (Martinsson and Perdahl, 1995).The c. 10 km thick pile of
Paleoproterozoic volcanic and sedimentary rocks was deformed and metamorphosed during the Svecofennian
orogeny (1.9-1.8 Ga), contemporaneous with intrusion of the 1.89-1.87 Ga Haparanda and Perthite-monzonite suites. These plutonic rocks have a calc-alkaline to alkali-calcic character and are comagmatic to the
Svecofennian volcanic rocks (Witschard, 1984). The Lina Suite comprises c. 1.79 Ga minimum melt granites
and pegmatites (Skild et al., 1988) that are temporally related to TIB 1 intrusions in the Kiruna-Narvik area
(Romer et al., 1992, 1994). A second event of metamorphism and deformation is at least locally developed at
this time (Bergman et al., 2001). The youngest plutonic rocks are represented by c. 1.71 Ga TIB 2 granitoids at
the Swedish-Norwegian border (Romer et al., 1992).

2.1. Local geology

Two major units of supracrustal rock occur in the Gllivare area. Rocks of a mainly volcaniclastic origin
occur in the eastern part, while intermediate to felsic volcanic rocks dominate towards west (Fig. 1). In the
south-western part these volcanic rocks are overlain by arenitic sediments. Minor occurrences of biotite gneisses and graphitic sediments that are found in the south-eastern part may represent the oldest rocks in the area.
The supracrustal rocks are intruded by plutonic rocks varying from peridotite to granite. The volcaniclastic
rocks have been given various stratigraphic names in different areas and in different papers.The subdivision has
partly been based on petrographical differences that have been superimposed by deformation and alteration on

originally similar rock units. In this paper the Muorjevaara Group (Zweifel, 1976; Ros, 1980) is extended to
include also the stratigraphically upper parts of these rocks (i.e. Liikavaara Group) and the Aitik and Nautanen
Groups (Zweifel, 1976). The internal stratigraphy of the Muorjevaara Group is poorly known and the upper
and lower contacts are not exposed. Pelites and lithic arenites occur in the stratigraphically lower part as some
hundred meters thick and alternating units. Cross bedding and minor intercalations of gravel conglomerate
occur within the arenites. The arenites are mainly of volcaniclastic origin and have an andesitic composition.
Locally they contain signicant amounts of clastic quartz grains. Andalusite porphyroblasts are common in the
pelites. The upper part of the volcaosedimentary pile is dominated by volcaniclastic rocks of andesitic composition. Amygdaloidal andesite lava and volcanogenic conglomerates are minor constituents (Ros, 1980).

GRANITE

KIIRUNAVAARA GROUP

GABBRO

MUORJEVAARA
FORMATION

MONZONITE

METASEDIMENTS

DIORITE

SULFIDE ORE

ULTRAMAFITE

IRON ORE

QUARTZITE

FAULTS

Nautanen
Malmberget

Aitik

Puolalaki
10 km

Fig. 1 The geology of the Gllivare area from Martinsson and Wanhainen (2004).

Metavolcanic rocks belonging to the Kiirunavaara Group are restricted to the western part and constitute the
host rock to the Malmberget apatite iron ores. The composition varies from trachyandesite to rhyolite with
possible minor basaltic intercalations. Primary textures and structures are rarely preserved but amygdaloidal and
porphyritic rock has been identied at Malmberget (Geijer, 1930). In the southern part these rocks are overlain
by feldspathic quartzites comparable with the Hauki Quartzite in the Kiruna area (dman, 1940).

Minor intrusions of dioritic to granodioritic composition are common within the Muorjevaara Group. They
belong to the Haparanda Suite and have at Aitik the character of a porphyritic high level intrusion (Wanhainen
and Martinsson, 2001). Monzonitic rocks of the Perthite Monzonite Suite are restricted to the western part.
Several gabbroic to troctolitic intrusions occur in the Gllivare area. Largest is the Dundret intrusion that forms
a bowl-like structure with a steep to gentle dip towards the center of the intrusion. It belongs to a complex of
alkaline to tholeiitic mac-ultramac intrusions in the Gllivare area (Martinsson, 1994). The youngest intrusive phase is represented by large areas of the Lina Suite granites.
The Nautanen Deformation Zone (NDZ) is the most prominent tectonic structure in the area running in a
NNW direction. Strongly schistose or mylonitic rocs occur in several parallel branches of high strain within an
up to 3 km wide zone (Bergman et al., 2001). Minor shear zones with a similar direction occur also outside
the NDZ.

2.2. The Malmberget deposit

The precise age of discovery for the Malmberget deposit is not known but most likely it was found at the end
of the 17th century. Mining started in small scale in the 18th century with the main production coming from the
Kapten ore body. When the railway was built from Lule to Gllivare in 1888 ore production rapidly increased
and open pits were developed on most outcropping ore bodies. The Malmberget deposit (Fig. 2) consists of
several ore lenses with 5161 % Fe and varying contents of phosphorus (Grip and Frietsch, 1973). At the end
of 2012, approximately 640 Mt of crude ore have been produced in open pits and underground workings and
the reserves were estimated to 271 Mt with 42 % Fe (LKAB, 2012).

Tingvallskulle

N Alliansen
Vitfors

Josefina
Upland
Skne

Vlkomman

Printzskld 895 m

Hens
Baron

Dennewitz

Selet

Johannes

Koskullskulle

.Johannes
Viri
930 m

Kapten
Kapten
830 m

Fabian
830 m

Malmberget
0

1 km

Granite-pegmatite

Skarn and ore-breccia

Sillimanite gneiss
Metavolcanics rocks
(leptite)

Hematite ore
Magnetite ore

Fig. 2 Simplied geological map over the Malmberget iron ore deposit modied from (Bergman et al., 2001).

In the western and northern parts of the deposit the ore forms an almost continuous horizon with a length
of about 5 km (Fig. 2). Apatite banding is a common feature of these ores, which contain both magnetite and
hematite. The eastern part includes several more or less isolated bodies of magnetite ore, which generally is
less rich in apatite. The main gangue minerals are apatite, amphibole, pyroxene and biotite. Pyrite, chalcopyrite,
bornite and molybdenite are more rarely found. The grain size is mostly 0.5 to 2 mm for the ore minerals, but
larger porphyroblasts of magnetite may occur in hematite ore.
Various scales of brecciation are developed in the wall rocks to the ores. Especially ore bodies in the eastern
part are surrounded by extensive brecciation. Magnetite, apatite, amphibole are the main constituents of mm
to m-wide veins that develop networks and breccias. Albite occurs in some amphibole breccia and scapolite
is locally found in druses. Breccias with a high Fe-content are mainly found adjacent to the iron ores, while
breccias dominated by amphibole are developed also at distance from them. The breccias are often strongly
attened by ductile deformation and they may transform into banded ore (Geijer, 1930).

3. METHODS
193 samples were collected from 22 drill cores and 4 out crops of the ViRi, Fabian, Dennewitz, Printzskld,
Vlkomma, Baron, Marta and Josena ore bodies for the petrographic characterisation of the rocks. 116
polished thin sections of representative rock types were prepared and investigated by microscopy in transmitted
light with focus on textural and mineralogical properties but also degree of deformation. Samples for characterisation of rocks were mainly selected from drill cores and consist of 20-30 cm long pieces of structurally and
mineralogically homogenous rock types. The samples were prepared and analysed at ALS laboratories for main
and trace elements using lithium metaborate fusion and ICP-AES and ICP-MS.Three set of samples were used,
two of them based on relogging of cores and selecting of samples identied as metavolcanic, mac dykes and
aplites. The third set of samples was selected from drill cores based on the traditional nomenclature used for
rocks in the mine (pegmatite, granite, red leptite, red-grey leptite, grey-red leptite and grey leptite).

4. RESULTS
4.1. Petrography of the Malmberget host rocks
The ores are hosted by strongly metamorphosed and deformed rocks of felsic to mac composition. These
rocks are traditionally called leptites in the Malmberget area. A porphyritic texture is locally preserved in
the felsic rocks. Amygdules are occasionally encountered, suggesting a mainly extrusive origin and a primary
character similar to that of the Kiruna Porphyries. Mac rocks are mainly found adjacent to the ores as
conformable to discordant lenses. Occasionally they contain remnants of plagioclase phenocrysts and amygdules. Some of the mac rocks are probably dykes, but most of them are suggested to have formed as sills or
extrusions (Geijer, 1930). A large intrusion of Lina granite exists northwest of the deposit and the recrystallization of the host rocks increases in that direction. Dykes of granite and pegmatite are frequently found in
the ores and their host rocks. Some of the pegmatites are rich in coarse-grained hematite, apatite and titanite.
Metavolcanic rocks

The metavolcanic rocks mostly lack primary textures and is characterised by a metamorphic fabric best
described as grano-porphyroblastic (Fig. 3) expressed by rather irregular grain boundaries and slightly porphyroblastic character of some of the feldspar grains that might be relicts of altered scapolite porphyroblasts. This
is in contrast to other rocks with a granoblastic to lepodoblastic texture for metadiabase and a partly preserved
magmatic texture for the deformed aplites (Fig. 4).
5

Fig. 3 Photomicrographs showing a grano-porphyroblastic texture of the metavolcanic rocks.

Fig. 4 Photomicrographs showing the major rock types A) metaandesite, B) metadacite, C) metadiabase granoblastic texture, D) metadiabase lepidoblastic texture, E) perthitaplite, F) Lina aplite, G) feldspar altered andesite, H) skarn
altered andesite.Transmitted light and crossed nicholls. Same scale of all photos.

Basaltandesite

Metaandesitic rocks occur as the dominating host rock to the iron ores at Malmberget. They are greyish in
colour with a grain size of 0.1 to 0.4 mm and CI varying from 35 to 15 which at least partly seem to reect
a variation from basaltic andesite to andesite-dacite expressed by a silica content of 52 to 60 % SiO2. The
proportions of mac minerals show large variations with mostly 5-25 % amphibole, 0-25% biotite and 3-10
% Fe-Ti oxides occurring as a ne grained dissemination. In some cases clinopyroxene may occur in amounts
up to 10 %. Locally the amount of Fe-Ti oxides is larger (up to 20%) and in those cases the amount of biotite and amphibole is small. The Fe-Ti oxides are often dominated by ilmenite containing various amounts of
hematite lamella (Fig. 5a) and in addition magnetite occur in small to larger amounts.Titanite is common as an
accessory mineral and often forms a coating or occurs adjacent to magnetite and ilmenite (Fig. 5b), while small
amounts of disseminated pyrite and chalcopyrite are rare. Generally the rock exhibits a well-developed foliation
or mineral lineation and the texture is grano-porphyroblastic to slightly lepidoblastic. Rarely amygdules could
be seen indicating an origin as lava ows.
A

Fig 5. Photomicrographs showing Fe-Ti minerals in metaandesite A) ilmenite with lamellas of hematite, B) magnetite
(left) and ilmenite (right) overgrowth by titanite (medium gray rim).
Daciterhyodaciterhyolite

Felsic volcanic rocks occur more rarely in the mine and then mainly in the western part (Vlkomma), but seem
to be a dominant rock type in the central part of the area within the syncline structure. The rocks are mostly
pale red to greyish red and rather often they are porphyritic with feldspar phenocrysts and for the more felsic
ones also small amounts of quartz phenocrysts. The silica content varies from 61 to 72 % SiO2. The texture is
mostly grano-pophyroblastic. The rocks are generally lacking structures giving information about their origin
and they may have a well-developed mineral lineation.
Mac intrusions

Metadiabase is common as up to more than 10 m wide dykes cutting the ore and the host rock. The grain size
is generally 0.2 to 0.7 mm although locally up to 2 mm. For wider dykes the grain size may decrease towards
their borders reecting chilled contacts. The amount of mac minerals is mostly 40 to 50 % with 0 to 40 %
hornblende, 0 to 45 % biotite, 1 to 5 % of disseminated magnetite and less common 0-20 % clinopyroxene.The
color is varying from dark grey to greyish brown and greyish green. The plagioclase may be partly or totally
replaced by scapolite in some cases. Based on the content of SiO2, TiO2 and Zr three different types may be
distinguished. Two types are high in TiO2 and contain 48 to 51 % SiO2 but differ in their Zr content. The
third type has lower TiO2 but higher Zr and a silica content of 48 to 57 % SiO2, thus varying from diabase to
monzonite in composition. Often no primary textures are preserved and generally the rock has a granoblastic
texture and a more or less well developed foliation or mineral lineation. However, more coarse grained dykes
may have partly a preserved magmatic texture.

Felsic intrusions
Deformed aplite

Aplitic rocks with a mostly weak to moderate foliation or mineral lineation is common as dykes in the wall
rocks but also in the ores. They are pale red to red in colour and have a grain size of 0.2 to 1 mm. They may
exhibit a porphyritic texture a few mm large feldspar phenocrysts. The amount of mac minerals is 4 to 7
% and consists of 1 to 6 % amphibole, 0-5 % biotite, 0.5 to 2 % of disseminated magnetite, while the quartz
content is 15 to 20%. The content of SiO2 varies between 69 and 73 %. A few per cent of anhydrite may exist
as irregularly disseminated grains or small patches and pyrite and chalcopyrite is locally occurring as a weak
dissemination. Within the mine the rock is sometimes more coarse grained (1-2 mm) changing character to
ne grained granite that might exhibit two directions of foliation. The grain boundaries for the feldspars may
be rather complex but especially for more granitic types a more typical magmatic texture is shown. Titanite
might occur as a magmatic mineral with euhedral shape (Fig. 6).
A

Fig. 6 Photomicrographs showing deformed aplite with euhedral titanite crystals.

Undeformed aplite

Dykes of more or less undeformed aplite are less common in the mine but occur locally as dykes in the ore and
the host rock. The colour is pale red to reddish and the grain size is 0.2 to 2mm. The amount of dark minerals
is mostly low (2 to 4 %) and they consist of biotite, small amounts of magnetite-hematite and locally amphibole.
Titanite may occur as an accessory mineral. The aplite may show transitions to granite and pegmatite.
Granite

Larger bodies of granite are rare in the mine but occur as large intrusions to the north.The colour is pale red to
reddish and the quartz content is mostly 25 to 30 but may be up to 35%.The amount of dark minerals is rather
low with 2 to 5 % biotite and small amounts of disseminated magnetite or hematite. The rock may contain
phenocrysts of microcline (5 to 15 mm) and sometimes also plagioclase (7 to 10 mm). The rock has generally
a weak to moderate foliation or mineral lineation but may also be rather undeformed.
Alterations

Several types of alteration affects the volcanic rocks and the deformed mac and felsic intrusions, while the
undeformed felsic intrusions mostly show little evidences of alteration. Feldspar alteration is extensively developed and comprises either albite or K-feldspar and is mostly pervasive but K-feldspar alteration may also occur
as veinlets and patches. Amphibole and pyroxene may occur together and may occur disseminated, in veinlets,
stockworks or as massive patches and larger volumes dened as skarn. Scapolite is also a characteristic alteration
mineral but may have been largely been replaced by feldspar or other secondary components.

The metavolcanic host rocks to the ores at the Malmberget are mostly albite rich and the pervasive character of
the alteration is generally accompanied by disseminated magnetite, hematite or biotite as the dominant mac
mineral. K-feldspar alteration may also be pervasive but is generally less common close to the ores. Scapolite is
may be a dominating felsic mineral in mac rocks occurring evenly distributed in the matrix or as porphyroblasts or patches.
Skarn alteration is extensively developed close to many of the ores but may also be of a more regional
occurrence at Malmberget. Pyroxene and amphibole are characteristic minerals and are usually accompanied
by varying amounts of magnetite, titanite and more rarely pyrite or chalcopyrite. The skarn minerals occur
disseminated but more commonly as schlieren, veinlets or patches in the metavolcanic rocks. Also the deformed
aplites may contain amphibole or pyroxene as an irregular dissemination or as schliren-patches. A characteristic
alteration is the occurrence of feldspar-pyroxene-amphibole in zones more than 20 m wide. The feldspar is
reddish albite and pyroxene-amphibole occurs in patches.Veins and schlieren in structures sometimes showing
similarities to magmatic-hydrothermal mixing products described from Cloncurry (Mark and Forster, 2000).
Close to the ores, massive and often coarse grained amphibole-pyroxene skarn may occur in up to more than
10 m wide zones and at the Fabian ore body these skarn zones often contains abundant anhydrite and smaller
amounts of chalcopyrite and titanite. Amphibole is at least partly formed by alteration of pyroxene.

4.2. Structures
The structure of the Malmberget area has not been studied in more detail but is suggested to represent a rather
complex synform structure formed by at least two phases of folding (Bergman et al., 2001). The Malmberget
deposit is strongly affected by ductile deformation and the present shape of individual ore bodies may partly
be controlled by stretching 40 to 50 towards SSW (Martinsson and Virkkunen, 2004). Many ore lenses are
boudinaged in the plunging direction and some of the granite-pegmatite dykes exhibit a similar style of deformation (Geijer, 1930). The area is situated between the Karesuando-Arjeplog Deformation Zone (KADZ)
and the Nautanen Deformation Zone (NDZ) and may have been squeezed towards south due to east-west
compression (Fig. 7) at around 1.8 Ga (Wanhainen et al., 2005).

Fig. 7 Block diagram of the Gllivare area (from Wanhainen et al., 2005).

4.3. Geochemistry of wall rocks

Based on the petrographic investigations of the different rock types, these have been analysed and classied into
mac to intermediate metavolcanic rocks, felsic metavolcanic rocks, deformed aplites, undeformed aplites, pegmatites and granites, metadiabase including monzonitic dykes and skarn altered rocks including massive skarn
and pyroxene-amphibole rich assemblages in metavolcanic rocks.
Primary character of rocks

Using the Igneous spectra by Hughes (1973) as a rst control of chemical changes the metavolcanic rocks
exhibit a large variation in the K2O/K2O+Na2O ratio with many samples having a sodic character (Fig. 8).
This combined with a rather limited variation in total alkali (8-11 % Na2O+K2O) suggest signicant alkali
alterations.This is in contrast to the deformed aplites that have an alkali composition typical for magmatic rocks.

Fig. 8 Metavolcanic rocks, skarn altered rocks, deformed aplites and metadiabase plotted in Hughes igneous spectrum
(Hughes, 1973).

This makes the TAS diagram less useful for classication of primary compositions of the metavolcanic rocks,
giving them a strongly alkaline character for the more mac rocks and a trend crossing that of typical magmatic rocks. In the TiO2/Zr versus SiO2 diagram by Winchester & Floyd (1977) the use of more immobile
elements gives a more reliable classication and the composition varies from basalt to andesite-trachyandesite and
trachyte-dacite-rhyodacite to rhyolite. In general this chemical classication is in accordance with the petrographic classication of less altered rocks (Fig. 9). The skarn altered rocks plot mainly in the alkaline elds
which is due to a low silica content caused by dilution by Ca and Mg during skarn formation.

10

Skarn altered
Felsic
Mafic-intermediate

Rhyolite
Rhyodacite

70

Comendite
Pantellerite

Dacite
Tr
ac
hy
an
de
sit
e

Trachyte
Andesite

50

Phonolite

Basanite
Nefelinite

Al
k

Subalkaline
basalt

ali
ba
sa
lt

SiO2 %

60

(Winchester & Floyd, 1977)

40
0.001

0.01

0.1

Zr/TiO2

Fig. 9 A chemical plot showing the metavolcanic rocks in the classication diagram of Winchester & Floyd (1977).
Geochemical groups and tectonic setting

By using ratios of immobile elements the precursors to altered rocks might be possible to identify as mass
changes caused by chemical changes are eliminated and the ratios of the immobile elements are not affected
by alteration. By testing different combinations of elements generally suggested to be immobile only Ti, Al and
Zr seem to be useful, while REE,Y, Nb and in a few cases Th in various degree show evidences of mobility.
When using the ratios TiO2/Zr versus TiO2/Al2O3 most rock types denes rather distinct groups (Fig. 10)
including also strongly altered samples in which primary lithologies are not possible to identify by petrologic
methods.
20

Skarn altered
Granite-aplite
Metaaplite
Metadiorite
Metadiabase
Felsic volcanic
Mafic volcanic

100TiO2/Al2O3

16

12

40

80

Zr/Th

120

160

200

Fig. 10 A chemical plot showing the ratios of the immobile elements used for grouping the rocks.

11

In the Zr-TiO2 diagram for classication of tectonic setting (Pearce, 1982) the metavolcanic rocks largely plot
in the within plate eld (Fig. 11). Compared to metavolcanic units in Norrbotten this is similar to the Kiirunavaara Group, while the Porphyrite Group mainly plots in the Arc eld (Martinsson, 2004).

Fig. 11 A chemical plot showing the metavolcanic and skarn altered rocks in the classication diagram of Pearce (1982)
for tectonic setting.

Also in the TiO2/Al2O3 versus TiO2/Zr diagram the metavolcanic rocks at Malmberget gives the same trend
as the Kiirunavaara Group, while the Porphyrite Group is distinctly different (Fig. 12). Using the same TiO2/
Al2O3 versus TiO2/Zr diagram for the intrusive rocks it is evident that the deformed aplites and most of the
metadiabases are part of the Perthite Monzonite Suite, while the undeformed aplites and pegmatites belong to
the Lina Suite (Fig. 13).

Fig. 12 (left) A chemical plot showing the metavolcanic rocks from the Malmberget deposit compared to the Kiirunavaara Group and the Porphyrite Group (Martinsson unpublished data). Fig. 13 (right) A chemical plot showing the
intrusive rocks from the Malmberget deposit compared to the Haparanda, Perthite Monzonite and Lina Suites (Martinsson unpublished data).
12

The metavolcanic rocks dene a trend controlled by the compositional variation from basalt to rhyolite. The
deformed aplites and most of the metadiabases plot on the same trend suggesting a common magmatic origin,
while the undeformed aplites and granites plot slightly different and is also characterised by high Th contents.
This shows the altered rocks to have precursor rocks of similar character as the less altered rocks and mainly to
be of felsic composition. An exception is the massive pyroxene-amphibole skarn that plot different and most
likely have another origin (Fig. 10).

5. DISCUSSION
On the geological map sheet Gllivare SV the volcanic rocks and most of the metasedimentary rocks are
classied as Porphyry Group (Witschard, 1996) without any distinction between Malmberget area and the
Nautanen area to the east. However, the volcanic rocks of the Muorjevaara Group have an arc signature (Wanhainen and Martinsson, 1999) and in more recent publications these two domains have been separated into
Kiirunavaara Group rocks and Porphyrite Group rocks, respectively (Bergman et al., 2001). The results from
this study clearly demonstrate that the metavolcanic rocks at Malmberget belongs to the Kiirunavaara Group
and are composed of basalt, andesite-trachyandesite, trachyte-dacite and rhyodacite.
The host rock to ore bodies in the eastern part of Malmberget (ViRi, Fabian, Dennewitz, Printzskld) seem
to be mainly mac to intermediate in composition (basalt, andesite-trachyandesite), while the host rocks in the
western part of Malmberget (Vlkomma, Baron, Hens) partly have a more felsic composition (dacite-rhyodacite). No internal stratigraphy of the metavolcanic rocks have been documented at Malmberget but based on
analogies with the stratigraphy of the Kiirunavaara Group at Kiruna some suggestions can be made. At Kiruna,
the Kiirunavaara Group mainly consist of trachyandesitic lava in the lower part while pyroclastic rhyodacite
dominate the upper part and with the Kiirunavaara deposit situated at the border of these units (Martinsson,
2004).
In the possible synform structure at Malmberget the mac-intermediate metavolcanic rocks mainly occupy
the margin of the structure and locally exhibit amygdule structures indicating an origin as lava ows, while the
more felsic metavolcanic rocks seem to be more common in the core of the syncline and more likely are of
pyroclastic origin. Thus, the metavolcanic rocks at Malmberget may show the same changes as at Kiruna from
more mac-intermediate composition to felsic stratigraphically upwards in the volcanic pile and the ores might
also have a position at a similar stratigraphic position as the Kiirunavaara deposit.
The metavolcanic rocks constitutes the host rock to the ores at Malmberget and both the ores and the host
rocks are intruded by metadiabase and aplite dykes belonging to the Perthite Monzonite Suite and aplite-pegmatite-granite dykes of the Lina Suite. This demonstrates that ore emplacement must have taken place short
after deposition of the metavolcanic rocks but before the co-magmatic intrusions of aplite and diabase. The
connection between Perthite Suite magmatism and ore formation is further demonstrated by the extensive
feldspar-pyroxene-amphibole alteration which seem to be closely related to the aplites of the Perthite Monzonite Suite and partly exhibit structures resembling magmatic-hydrothermal transitions similar to what is
described from Cloncurry in Australia (Mark and Forster, 2000). However, some types of K-feldspar alteration
are more likely related to Lina Suite intrusions.

13

Extensive mobilization of ore components during metamorphic recrystallization and Lina magmatism is indicated by the occurrence of magnetite, hematite, apatite, titanite, pyroxene and amphibole in pegmatites within
the Malmberget deposit. Another example of mobilized ore components is the spectacular occurrence of large
cavities and druses containing excellent crystals of calcite, desmine, apatite, hornblende, scapolite, pyrite and
magnetite. Ages for minerals in these druses vary from 1613 to 1737 Ma and after that no major magmatic or
metamorphic event have affected the bedrock (Romer, 1996).

6. CONCLUSIONS
The host rock to the Malmberget deposit constitutes of metavolcanic rocks belonging to the Kiirunavaara
Group and varies in composition from basalt-andesite-trachyandesite to trachyte-dacite-rhyodacite-rhyolite.
Both the host rock and the ore are intruded by altered and deformed dykes of diabase-monzonite and aplitic
rocks of the Perthite Monzonite Suite. Youngest are dykes of granite-aplite-pegmatite of the Lina Suite that
might be moderately deformed expressed as mineral lineation and by boudinaged dykes. The host rocks and
the ore bodies might occupy a syncline structure with dominantly mac to intermediate metavolcanic rocks
occurring in the outer part of the structure and hosting the deposits in the eastern part of Malmberget, while
more felsic metavolcanic rocks are found in the core of the structure representing a stratigraphically higher
position and hosting the deposits in the western part of the Malmberget deposit. An alteration of the host rocks
is dominated by albite alteration and amphibole-pyroxene alteration. The latter is extensively developed close
to the ore bodies and may at least be related to aplitic intrusions of the Perthite Monzonite Suite. K-feldspar
alteration is less frequently developed close to the ores and is then mainly formed as a later overprinting related
to the Lina Suite intrusions.

7. ACKNOWLEDGMENTS
LKAB is acknowledged for initiating, nancing and the admission of publication of results.

14

8. REFERENCES
Bergman, S., Kbler, L., and Martinsson, O., 2001. Description of regional geological and geophysical maps of
northern Norrbotten county: Geol Survey of Sweden, Ba 56, pp. 110 p.
Debras, C., 2010. Petrology, geochemistry and structure of the host rock for the Printzskld ore body in the
Malmberget deposit. Master thesis 2010:052, Lule University of Technology, 40 p.
Geijer, P., 1910. Igneous rocks and iron ores of Kiirunavaara, Luossavaara and Tuolluvaara. Economic Geology
5, 697718.
Geijer, P., 1930. Geology of the Gllivare ore eld. Geological Survey of Sweden, Ca 22, 115 p. (in Swedish
with English summary).
Grip, E. and Frietsch, R., 1973. Ore deposits in Sweden 2, northern Sweden: Almqvist & Wiksell, 295 p (in
Swedish).
Hughes, C.J., 1973. Spilites, keratofyres, and igneous spectrum. Geol. Mag., 109: 513-527.
LKAB, 2012, Annual Report: Lule, Sweden, Lule Graska, 138 p.
Mark, G., Foster, D.R.W., 2000. Magmatic-hydrothermal albite-actinolite-apatite-rich rocks from the Cloncurry district, NW Queensland, Australia. Lithos 51, 223-245.
Martinsson, O., 1997, Paleoproterozoic greenstones at Kiruna in northern Sweden: a product of continental
rifting and associated mac-ultramac volcanism. In O. Martinsson: Tectonic setting and metallogeny of the
Kiruna Greenstones. Ph. D. Thesis 1997:19, Paper I, 49 p, Lule University of Technology.
Martinsson, O., 2004. Geology and metallogeny of the northern Norrbotten Fe-Cu-Au province. In: R.L
Allen, O. Martinsson and P. Weihed (Eds.), Svecofennian Ore-Forming Environments: Volcanic-associated Zn-Cu-Au-Ag, intrusion associated Cu-Au, sediment-hosted Pb-Zn, and magnetite-apatite deposits in
northern Sweden. Society of Economic Geologists, Guidebooks Series 33, 131-148.
Martinsson, O., Perdahl, J.-A., 1995. Paleoproterozoic extensional and compressional magmatism in northern
Sweden. in Perdahl, J.-A.: Svecofennian volcanism in northernmost Sweden. Ph.D. Thesis 1995:169D, Lule
University of Technology, Sweden. Paper II, 13 pp.
Martinsson, O., Virkkunen, R., 2004. Apatite iron ores in the Gllivare, Svappavaara, and Jukkasjrvi areas. In:
Allen, R.L., Martinsson, O., Weihed, P. (Eds.) Svecofennian Ore-Forming Environments: Volcanic-associated Zn-Cu-Au-Ag, intrusion associated Cu-Au, sediment-hosted Pb-Zn, and magnetite-apatite deposits in
northern Sweden. Society of Economic Geologists Guidebook Series 33, 167-172.
Martinsson, O., Wanhainen, C., 2004. Field trip day ve, Cu-Au deposits in the Gllivare area. In R.L. Allen,
O. Martinsson and P. Weihed (eds.) Svecofennian Ore-Forming Environments:Volcanic-associated Zn-CuAu-Ag, intrusion associated Cu-Au, sediment-hosted Pb-Zn, and magnetite-apatite deposits in northern
Sweden. Soc. Econ. Geol., Guidebooks Series 33, 159-162.
dman, H., 1940. Om frekomsten av Vakkosediment sder om Gllivare, Dundret. GFF Geologiska Freningen i Stockholm Frhandlingar, v62, p.45-60.
Pearce, J.A., 1982.Trace element characteristics of lavas from destructive plate boundaries. In: R.S.Thorpe (Editor), Andesites, Orogenic Andesites and Related Rocks. Wiley and Sons, Chichester, pp. 525-548.
Romer, R.L., 1996. U-Pb systematics of stilbite-bearing low-temperature mineral assemblages from the Malmberget iron ore, northern Sweden. Geochimica et Cosmochimica Acta 60, 1951-1961
Romer, R.L., Kjsnes, B., Korneliussen, A., Lindahl, I., Skysseth, T., Stendal, H. and Sundvoll, B., 1992. The Archaean-Proterozoic boundary beneth the Caledonides of northern Norway and Sweden: U-Pb, Rb-Sr and
Nd isotopic data from the Rombak-Tysfjord area: Geological Survey of Norway, Report 91, (225), 67 p.
Romer, R.L., Martinsson, O., and Perdahl, J.-A., 1994. Geochronology of the Kiruna iron ores and hydrothermal alterations: Economic Geology, v. 89, p. 1249-1261.
Ros, F., 1980. The Nautanen area, report of the work done by the Geological Survey of Sweden during 19661979: Geological Survey of Sweden, Unpublished report, PRAP 81530, 33 p. (in Swedish)
Skild, T., hlander, B.,Vocke, Jr., R.D., and Hamilton, P.J., 1988. Chemistry of Proterozoic orogenic processes
at a continental margin in northern Sweden: Chemical Geology, v. 69, p. 193-207.
Wanhainen, C., and Martinsson, O., 1999, Geochemical characteristics of host rocks to the Aitik Cu-Au deposit, Gellivare area, northern Sweden: Proceedings of the fth biennial SGA meeting and the tenth quadrennial
IAGOD Meeting, London, 22-25 August 1999, Extended abstract, p. 1443-1446.

15

Wanhainen, C., and Martinsson, O., 2001. Petrology and geochemistry of the porphyritic Aitik intrusion and its
relation to the disseminated Aitik Cu-Au deposit, northern Sweden. in Wanhainen, C.: Magmatic and Hydrothermal processes in relation to the formation of the Aitik Cu-Au deposit, northern Sweden. Licentiate
Thesis 2001:11, Paper I, 15 p. Lule University of Technology, Sweden.
Wanhainen, C., Billstrm, K., Martinsson, O., Stein, H. and Nordin, R., 2005. 160 Ma of magmatic/hydrothermal and metamorphic activity in the Gllivare area: Re-Os dating of molybdenite and U-Pb dating of
titanite from the Aitik Cu-Au-Ag deposit, northern Sweden. Mineralium Deposita 40, 435-447.
Winchester, J. A. and Floyd, P. A., 1977. Geochemical discrimination of different magma series and their differentiation products using immobile elements. Chemical geology, v. 20, pp. 325-343.
Witschard, F., 1984. The geological and tectonic evolution of the Precambrian of northern Sweden - a case for
basement reactivation? Precambrian Research 23, 273-315.
Witschard, F., 1996. Bedrock map 28K Gllivare, 1:50 000: Geological Survey of Sweden, Ai 98-101. (in Swedish)
Zweifel, H., 1976. Aitik: geological documentation of a disseminated copper deposit A preliminary Investigation: Geological Survey of Sweden, C 720, 80 p.

16

II

Origin and evolution of the metamorphosed Malmberget apatite

iron ore, northern Sweden
Cecilia Lunda, Olof Martinssona and Jens C. . Andersenb
a Division of Geosciences and Environmental Engineering, Lule University of Technology,
971 87 Lule, Sweden
b Camborne School of Mines, University of Exeter, UK
* Corresponding author. Tel.: +46 920 492354. E-mail address: [email protected]

(Submitted)

Origin and evolution of the metamorphosed Malmberget apatite iron ore,

northern Sweden
Cecilia Lunda, Olof Martinssona and Jens C. . Andersenb
a Division of Geosciences and Environmental Engineering, Lule University of Technology, 971 87 Lule, Sweden
b Camborne School of Mines, University of Exeter, UK
* Corresponding author.Tel.: +46 920 492354. E-mail address: [email protected]

ABSTRACT
The strong metamorphosed and deformed Malmberget apatite iron ore deposit, northern Sweden is one of
the larger Kiruna-type deposits in the world. Magnetite and hematite from several ore bodies have been characterised by their geochemical- and mineral composition, Fe-Ti oxide mineral textures and chemistry of the
ore minerals to get a better understanding of the primary character of the ore. Optical microscopy, whole rock
geochemistry and electron microprobe analyses (EPMA) were used. Variations in the whole rock chemistry
mainly mimic the mineral chemistry of the magnetite ores and suggest indicating a primary origin, with different signatures for the massive ore and the ore breccia. Different chemical subtypes of massive ore are identied
based on their discrepancy in levels of P, Ti,V and Ga and the mineral chemistry of the magnetite is explained
by this chemically different sub types. These subtypes are associated to a specic mineralogy of apatite, amphibole, sulphides and anhydrite. However, the chemistry of magnetite is also controlled by the local oxide mineral
assemblages, as a redistribution of V and Ti into the coexisting hematite. The textural and paragenetic relation
of Fe-Ti oxides indicates a progressive oxidation from east to west in the deposit which also is expressed by
hematite that becomes an important ore mineral in the western part of the deposit. A multiphase origin is suggested where massive ore is formed by several injections of iron oxide magma while the ore breccia may have
been formed by hydrothermal processes.

Key words
Apatite iron ore, magnetite, mineral chemistry, Fe-Ti oxide textures, metamorphic recrystallisation

1. INTRODUCTION
Sweden is the most important producer of iron ore in Europe with two large underground mines in Kiruna
and Malmberget and an open pit mine at Gruvberget operated by LKAB. These deposits have produced over
2000 million tons (Mt) of ore and have a total iron ore reserve of 1010 Mt @ ~ 46 % Fe (LKAB, 2011). Both
deposits are apatite-iron ores, an ore type that is common in northern part of Sweden but quite rare in other
parts of the world. The apatite-iron ores have been the focus of genetic discussions over 100 years. Two genetic
models have been suggested: many authors favour a magmatic model (Geijer, 1910; Park, 1961; Frietsch, 1984;
Nystrm and Henrques, 1994; Henrques et al., 2003) but a hydrothermal model has also supporters (Park,
1973; Hildebrand, 1986; Hitzman et al., 1992; Rhodes et al., 1999; Sillitoe and Burrows, 2002). As not all features of the apatite iron ores are easily explained with a single genetic model a combination of these two models, a magmatic origin with a hydrothermal overprint (Martinsson, 2004) has been suggested for the formation
of the Kiruna and Malmberget deposits at 1.89-1.88 Ga (Cliff et al., 1990; Romer, 1996). The characteristic
features of the Kiruna type deposits is that they occur in intermediate to felsic volcanic rocks, are tabular and
concordant with their host rock, or have a breccia style and include a dominating mineralogy of magnetite,
hematite, apatite, and actinolite. Apatite gives a high phosphorus content which is typical of most ore deposits
(Martinsson, 2004). Typically the vanadium content of magnetite is high and similar to orthomagmatic magnetite accumulations in mac rocks but the titanium content is intermediate to magmatic and sedimentary
iron deposits (Frietsch, 1970; Loberg and Horndahl, 1983). The two world class deposits of apatite iron ore
in Kiruna and Malmberget are suggested to have a similar origin (Geijer, 1930; Loberg and Horndahl, 1983).
However, the Malmberget iron ore deposit is more strongly affected by later metamorphism, deformation and
intrusion of granitic rocks (Bergman et al., 2001).
Two major types of apatite iron ores can be distinguished in the Kiruna area, a breccia type and a more massive
stratiform-stratabound type (Bergman et al., 2001). Ore breccia is an irregular network of ore veins which, to a
varying extent, accompany the massive ore (Geijer, 1910; Geijer, 1931; Frietsch, 1982). In the Malmberget deposit the ore bodies situated in the eastern part have cores of massive ore surrounded by extensive breccia type
of ore (Geijer, 1930). Mineralogically the ore breccia is distinguished from the massive ore by the presence of
silicates and more complex textures with irregular grain boundaries and intergrowths of magnetite and silicates
(Lund and Martinsson, 2008).
In this study we have used geochemical data and carried out electron microprobe analyses of several different
ore bodies in the Malmberget apatite iron deposit. The purpose of this study is to dene the chemical and
mineralogical differences of the ore bodies in the deposit such as the geochemical- and mineral composition,
oxide mineral textures, the mineral chemistry of the Fe-Ti oxide minerals and the included gangue minerals to
get a better understanding of the primary character of the ore and the signicance of later overprinting events.
This information is of importance for the mineral processing of the ore and the quality of ore concentrate but
has also genetic implications.

2. GEOLOGICAL SETTING AND METALLOGENY

2.1. Regional geology
The geology of the northern Norrbotten ore province includes bedrock sequences of different ages.The oldest
unit is a 2.8 Ga Archaean granitoid-gneiss basement unconformable overlain by 2.5-2.0 Ga Karelian successions followed by 1.9 Ga Svecofennian volcanic and sedimentary units. Forty apatite-iron deposits are known
from the northern Norrbotten ore province and they are hosted by and probably also genetically related to, the
volcanic rocks in the Svecofennian succession (Martinsson, 2004).The early orogenic Haparanda Suite predates
the 1.86-1.88 Ga Perthite Monzonite Suite and was affected by deformation and metamorphism at 1.88 Ga.
At 1.81-1.78 Ga a second deformation and metamorphic event occurred and was associated with intrusion of
the Lina granite-pegmatite suite (Martinsson, 2004).

2.2. Deposit geology

More than 20 different tabular to stock shaped iron ore bodies are known in the Malmberget deposit within an
area of 2.5 x 5 km (Fig 1). They have undergone at least two phases of folding and metamorphism and occupy
a large-scale fold structure where the individual ore bodies are stretched parallel to the fold axis, which plunge
40-50 towards SSW (Bergman et al., 2001).The ore bodies are recrystallized, coarse grained, and intruded by
granite, aplite and pegmatite of the Lina Suite (Martinsson and Virkkunen, 2004; Martinsson, 2004). The host
rocks to the ore consist of felsic to mac, metavolcanic rocks that are strongly recrystallized and mostly lack
primary features. Especially in the footwall of the western part of the deposit, there exists gneiss with sillimanite,
muscovite and quartz while corundum and andalusite occasionally are found in the ore (Geijer, 1930). Mac
rocks are typically rich in biotite and can be found close to the ore as conformable to discordant lenses. Some
of them are probably metamorphosed dikes while others seem to have formed as sills or extrusions (Geijer,
1930).These metavolcanic rocks are traditionally called leptites and gneisses and are mainly distinguished by the
grade of metamorphism and the grain size (Geijer, 1930). These rocks may have an origin similar to those of
trachyandesite to rhyodacite composition that hosts the Kiirunavaara deposit (Martinsson and Hansson, 2004).
The Hens (Hn) and Vlkomma (V) ore bodies are located in the western part of the deposit and form part of
an almost continuous ore horizon, 5 km in length consisting of both hematite and magnetite ore (Bergman et
al., 2001).The Hens ore body has a tabular shape, composed of separate magnetite and hematite ore lenses with
an average grade of 53.5 % Fe and 0.80 % P. A typical feature is coarse grained apatite, occurring as a regular
banding or more disseminated. The Vlkomma ore body shows almost the same mineralogical character as
Hens but magnetite and hematite occur mostly together, often with porphyroblasts of one of the minerals in a
matrix of the other (Geijer, 1930).
The Fabian (Fa) and ViRi (Vr) ore bodies occur in the eastern part of the Malmberget deposit and include a
number of more or less individual magnetite ore bodies of a massive magnetite ore which is generally poorer
in apatite (Martinsson and Virkkunen, 2004) with grades of 55.6 61.5% Fe and 0.15-0.35 % P. A characteristic
feature of these ore bodies is that the massive ores are surrounded by breccia style mineralisation that extends
up to several tens of meters into the wall rock and gradually decreases in intensity away from the ore.
The Printzskld (Pz) ore body is situated between the eastern and western part of the Malmberget deposit.
This magnetite ore body has a lenticular shape and is partly rich in apatite and biotite giving the ore a banded
structure.The grade is 50.5 % Fe and 0.75 % P. Smaller hematite bodies occur close to the magnetite and minor
amounts of hematite is also present in the magnetite ore.

Karesuando

68o

IN

Soppero

Kiruna

nd

Svappavaara
Vittangi
N

Finla

Sweden

Norw

ay

64o

60o
Malmberget
30 km

Pajala

Trend

Paleoproterozoic intrusive rocks

Porphyrite Group

High grade gneisses

Hauki and Maattavaara quartzites

Pahakurkkio Group

Kiruna Porphyries

Greenstone Group

Archaean basement
Apatite iron ore
<50 Mtone, >50 Mtone

Tingvallskulle

N Alliansen
Vitfors

Josefina
Upland
Skne

Vlkomman

Printzskld 895 m

Hens
Baron

Dennewitz

Selet

Koskullskulle

Johannes
.Johannes
Viri
930 m

Kapten
Kapten
830 m

Fabian
830 m

Malmberget
0

1 km

Granite-pegmatite

Skarn and ore-breccia

Sillimanite gneiss
Metavolcanics rocks
(leptite)

Hematite ore
Magnetite ore

Fig. 1 The geology of the northern Norrbotten ore province, apatite iron deposits marked in red. Lower gure a simplied geological map of the orebodies in the Malmberget area, modied after Bergman et al., 2001

3. SAMPLING AND ANALYTICAL METHODS

Representative drill cores were selected from the ore bodies.The drill cores were logged, sampled and analysed,
each to a different extent. The focus of the project is the Fabian and Printzskld ore bodies, which are represented by ve drill cores intersecting the ore from the hanging wall to the foot wall (Fig 2).The Vlkomma ore
body is represented by four drill cores and Hens and ViRi by hand specimens from the underground mine. Major and trace element chemical analyses were carried out on 95 drill core samples from Fabian and Printzskld
at the Activation Laboratories LTD in Canada. The major elements were analysed using the inductively coupled plasma mass spectrometry (ICP-OES) and for trace elements (ICP-MS). Approximately 100 polished thin
sections were examined and 30 of them were selected for electron microprobe analyses (EPMA) based on their
mineralogy. The analyses were performed at the Camborne School of Mines, university of Exeter, Cornwall
Campus, UK on a JEOL JXA-8200 superprobe with a technique described by Potts et al. (1995). An accelerating voltage of 15 kV and a beam current of 30 nA were used for the oxides. The probe beam diameter was
set to 1 m for the oxides and the phosphates and 5 micron for the silicates. The FeO and Fe2O3 ratio was
calculated on the basis of stoichiometry and charge balance (Droop, 1987). First, the cations other than iron
were allocated to their specic site occupancies according to their charges. After this, the Fe was proportioned
into Fe2+ or Fe3+ to balance the stoichiometry and charge. Finally oxygen was allocated accordingly to produce
FeO and Fe2O3. The recalculation of the magnetite and hematite data was prepared by the method of Jens C.
. Andersen (pers. comm.). Based on the mineral chemistry and the whole rock analyses, additional 16 samples
were selected for detailed mineralogical study by electron microprobe of Fe-Ti mineral pairs at the Finnish
Geological Survey (GTK) using a CAMECA SX 100. The accelerating voltage was 20 kV and a beam current
of 60 nA. The probe beam diameter was set to 10 m for the oxides. Standards for the different elements and
analyses in the two methods were used of natural silicates oxides and metals.

4. RESULTS
4.1. Geochemical features in the deposit
In order to trace and explain the element variation in magnetite, geochemical analyses of the magnetite ore is
used to identify variations on the macro scale. The average chemical compositions of the ores at Malmberget
demonstrate a distinct difference between massive ore and ore breccia and that there are signicant subgroups
of the massive ore. To compensate for the varying amounts of iron oxides in the ores the contents of Ti, V, P
and Ga are normalised against the Fe content (wt. %). The massive ore in Fabian could be divided into three
subgroups based on these elements. One subgroup contains higher amounts of Ti, V and Ga and another is
rich in P. The third subgroup is more heterogeneous in character but is generally low in P, Ti,V and Ga (Fig 2).
Similar subgroups could not be identied in the Printzskld ore body.

Fa
6500 Fe (%)

50 50 65 75

Ti (%)

P (%)
0 0.4

1.1

0.4

0.9

Pz
6852

V (%) Ga (ppm)
0

0.13 0.26 0

55

300

Fe (%) P (%)

Ti (%)

V (%) Ga (ppm)

30

65 75

0 0.4

0.4 0.9

0.15

40

Magnetite
Skarn
Intermediate metavolcanic rock

400

Intermediate to felsic metavolcanic rock

100

Felsic metavolcanic rock

Ore Breccia
Apatite
500

200

Amphibole / Pyroxene

Meter

Sulphides
Feldspar

600

Biotite

300

Magnetite

700

400

465
800

877

Fig. 2 Schematic drill core logs and geochemical data from Fabian and Printzskld ore bodies. The shadowed areas in
the geochemical logs are the ore sections enriched in Ti,V, P and Ga

The average V and Ti (wt. %) contents for the ore bodies in the Malmberget deposit exhibit systematic changes with decreasing contents from east to west. Compared to other apatite iron ores in Norrbotten,ViRi and
Fabian are similar to Kiirunavaara in that they have the high amounts of Ti and V, while Printzskld and Hens
are similar to the Per Geijer, Gruvberget, and Ekstrmsberg ores, which have lower contents of Ti and V (Fig
3).
1.0

Deposit

TiO2 (wt%)

0.8

0.6

Vr
Fa
Pz
Hn
Grb
KUJ
PG
Ekb

0.4

0.2

0.0
0.00

0.05

0.10

0.15

0.20

0.25

V2O3 (wt%)
Fig. 3 A chemical plot showing the average contents of V2O3 and TiO2 (wt. %) for different apatite iron deposit in the
Norrbotten area.Viri (Vr), Fabian (Fa), Printzskld (Pz), Hens (Hn) and Ekstrmsberg (Ekb). Gruvberget (Grb) and
Kiirunavaara (KUJ B-ore), LKAB unpublished data and Per Geijer (PG) ore (Park 1973)

4.2. Mineralogy
The massive ore is mineralogically simple, consisting mainly of iron oxides, amphibole-pyroxene and apatite in
different proportions and with varying textures (grain size, shape and interrelationship). Ore breccia has more
complex mineral compositions and structures and comprises different mineral assemblages (Table 1). Similar
assemblages occur locally within the massive ore as veins, but typically ore breccia is mainly developed in the
wall rocks surrounding the massive ore (Fig 4).
Table 1. The different mineral assemblages occurring in the ore.
No.

Different mineral assemblages

Magnetite+apatite+amphibole+pyroxenebiotite

Magnetite+amphibole+pyroxene+biotite+chalcopyrite+pyrite+feldsparapatite

3a

Magnetite+plagioclase+quartz+amphibole+chalcopyrite+pyritebiotit

3b

Magnetite+plagioclase+quartz+amphibole+biotitechalcopyritepyritetitanite

Magnetite+amphibole+anhydrite+chalcopyrite+pyritecalciteplagioclase

Magnetite+biotite+chlorite+feldsparamphibolepyroxene

Magnetite+plagioclase+quartz+amphibolechalcopyriteilmenitetitanitepyrite.

B
Mag

Mag

Mag

Mag

Amph/Px

F
Bt

Mag
Py

Fsp

Ccp

Amph

Fig. 4 Photomicrographs showing magnetite textures and mineral assemblages at the Malmberget deposit (reected and
transmitted light). A) Massive ore with coarse grain magnetite occurring in an equigranular matrix with triple junction
conguration. B) Massive ore with a ner grain magnetite matrix with more complex grain boundaries.
C) Ore breccia, small magnetite grains (black) existing as inclusions and along grain boundaries in the (granular) silicate
matrix (albite and quartz) with a subgranular, complicated texture. D) Ore breccia, amphibole and pyroxene occurring
as cluster together with aggregates or inclusions of magnetite (black). E) Biotite aligned in two directions and amphibole outlined in a matrix of feldspar and quartz. F) Sulphides (pyrite and chalcopyrite) occurring as dissemination in veins and at grain boundaries of magnetite

The magnetite ore can be classied into three textural types based on the magnetite grain size: ne grained
(0.01-0.4 mm), medium grained (0.4-0.8 mm) and coarse grained (0.8-1.4 mm). Coarser grains often exist
as porphyroblasts (up to 4.6 mm) and this is a characteristic texture in Hens and Vlkomma ore bodies. The
magnetite grains are euhedral to subhedral. They tend to have a granoblastic texture, or a porphyroblastic texture with gently curved grain boundaries with increasing grain size. In the latter case the grain boundaries are
simple with distinct triple junctions. Larger magnetite grains can have an elongated shape outlining a mineral
lineation parallel to the general fabric of the rocks. Small (0.01-0.2 mm) single anhedral magnetite grains occur
at grain boundaries and as inclusions in the silicate matrix, a common texture in the ore breccia. The samples
of ne grained magnetite ore typically contain coarser magnetite grains, which occur in vein-like structures.
In the ore breccia the magnetite also occurs as aggregates of larger grains (0.16-2 mm). The grain shape varies
from euhedral to anhedral and angular, although elongated and somewhat rounded grains may occur.The grain
boundaries vary from simple and straight to a more complex shape.
Hematite occurs as a minor constituent in the Fabian,Viri and Printzskld ore bodies, as single grains, exsolution lamellas or as oxidation surfaces together with rutile along the grain boundaries of magnetite. In the Hens
and Vlkomma ore bodies, hematite is instead a primary mineral together with magnetite and occur as euhedral
to anhedral (0.2-2mm) single grains or as larger porphyroblasts (>20mm). Some of the grains show a folded
and partly crenulated foliation (Fig 5).
Table 2. The gangue and associated minerals to the magnetite and hematite in the ore bodies
Mineral

Grain size

Grain shape

Ilmenite

0.1-0.2 mm,
(>1 mm)

Euhedral to subIlm exists as single grains interstitial to magnetite both in grain boundaries and
hedral, porphyas intergrowths in magnetite (0.1-0.2 mm).
roblast, lamellae

Pyrite
Chalcopyrite

0.005-0.2 mm Anhedral to
(>4.8 mm)
euhedral

Texture

Py and Ccp occur as dissemination in veins, at grain boundaries and as large

single euhedral grains. Together with Anh these sulphides occur primarily in the
ore breccia otherwise as minor minerals

Anhydrite

0.4-0.8 mm
(clusters
> 4mm)

Anhedral

Apatite

0.1-0.8 mm

Euhedral, elongated

Ap exists interstitial to magnetite or as elongated grains aligned in the direction

of the host rock lineation

0.2-1 mm
(cluster 0.83.2 mm)

Anhedral

Amphibole varies from Act to Tr and pyroxene, with a diopsidic composition

(Wo37-41En46-48Fs13-17) which occurs as clusters but also as single grains
interstitial to magnetite.

Actinolite
Tremolite
Diopside
Albite
Orthoclase
Quartz

The silicates varies from an equigranular matrix of euhedral grains of plagioclase and K-feldspar to subgranular matrix of inequigranular anhedral grains
of plagioclase, K-feldspar, quartz and sometimes amphiboles, all with complex
grain boundaries. Plagioclase and K-feldspar occur as single grains interstitial
to magnetite or as aggregates and are sometimes stretched in a linear direction.
0.01-0.12 mm Subhedral to an- The composition of plagioclase varies from albite to oligoclase (1-11 % An) and
0.2-0.8 mm
hedral, rounded K-feldspar is almost a pure orthoclase (97% Or).
0.04-1.2 mm
(< 5.4 mm)

Euhedral to
anhedral

Biotite

0.1-0.4 mm (up
Tabular
to 2 mm)

Both a Fe-rich and a Mg-rich biotite (annite, phlogopite) occur as yellow-green

to brownish grain often interstitial to magnetite or aligned in lineation or as
clusters, sometimes partly replaced by chlorite.

Titanite

0.1-0.2 mm

Anhedral

Large grains often interstitial to magnetite and diopside in samples of coarser

grains size.

Zircon

0.01 mm

Anhedral

A rare mineral that occur interstitial to magnetite

Mineral abbreviation (Kretz 1983)

Ilm

Hem

Mag
Mag
Ilm
Rt-Ilm
phase
C

Ilm

Mag

Hem
Spl

F
Hem

Hem

Mag

Fig. 5 Photomicrographs showing the oxidation textures of magnetite, ilmenite and hematite at the Malmberget iron
ore deposit (reected light). A) Exsolution lamellae of ilmenite in magnetite (Vr, Fa and Pz), (C2-C3). B) Magnetite
showing oxidation an assemblage of rutile-ilmenite phases and ilmenite with hematite lamellas (Vr, Fa and Pz), (C4-C5).
C) Hematite with exsolved ilmenite lamellas (Vr, Fa and Pz). D) Magnetite with intergrowths of spinel and rutile (Pz
and Hn). E) Magnetite and oxide hematite porphyroblasts (Hn and V). F) Hematite grains showing folded foliation,
partly crenulated (V).

The associated minerals to magnetite and hematite exhibit variations both within and between the different
ore bodies (Table 2). The main gangue mineral is apatite, existing interstitial to magnetite or as elongated
grains aligned in the direction of the host rock lineation. The apatite grain size is correlated to the grain size of
coexisting magnetite and hematite. Amphibole and pyroxene occur partly disseminated but more commonly
as patches and schlieren. The amount of these minerals together with anhydrite is gradually increasing to the
skarn (amphibole-pyroxene) alterations located at the host rock contact. The biotite occurs as dissemination or
in schlieren in the massive ore and is also common close to the host rock contacts. In the Printzskld ore body
the biotite is an abundant constituent particularly towards the footwall contact. Accessory to minor constituent
are mainly pyrite, chalcopyrite anhydrite, ilmenite and titanite. The pyrite and chalcopyrite occur locally in
small amounts disseminated and as veinlets in the massive ore in the eastern part of the deposit. These minerals
are more abundantly in ore breccia together with the anhydrite and the amphibole-pyroxene. The typical ore
breccia minerals are otherwise the feldspar and quartz occurring mainly as a matrix close to the host rock.

4.3. Fe-Ti oxide mineral textures

The Fe-Ti oxide minerals show oxidation textures with various characteristics in the different ore bodies
(Table 3), (Fig 5). Homogeneous magnetite, magnetite with intergrowths of Fe-Ti oxides (ilmenite, rutile and
spinel) and oxidation assemblages of hematite and rutile occur. Magnetite with Fe-Ti oxides contains oriented
ilmenite lamellas (trellis and sandwich) with a thickness of approximately 2 m or has oriented exsolved spinel
patches 50-200 m in size. Hematite can have remnants of magnetite and/or contain lamellas of rutile, especially at the rims of the grain or in contact with silicates. Some of the ilmenite grains contain exsolved hematite
lenses, ranging between 4.5-270 m. These lenses occur in centre of the grain. Homogenous hematite grains
have inclusions of 20-100 m Fe-Ti lamellas or have 10-50 m rutile needles exsolved at the rim of the grain.

Table 3 Representative electron microprobe analyses of the silicates, phosphate and mica
Mineral Actinolite Tremolite

Diopside Diopside Albite

Albite

Albite

Albite

Apatite Apatite Biotite Biotite

Sample
No.

V35-2 272

V13-1 623

AP1-11 V29-2
V20-3 311 V17-2 493
772
175

AP1-8
738

V35-2
270

12-2 864 8-2 536

9-2 596

AP3-14216

SiO2

55.4

57.7

54.7

54.5

71.3

69.1

69.4

71.4

44.2

38.3

Al2O3

1.88

0.94

0.23

0.64

20.6

20.4

20.4

19.6

0.02

12.8

14.0

TiO2

0.03

0.02

0.02

0.01

0.96

2.61

FeO (tot) 10.3

4.00

9.61

8.22

0.05

0.08

0.07

0.21

0.06

4.34

19.0

0.05

0.03

0.14

0.09

0.02

0.01

0.03

0.01

0.11

MnO
MgO

17.6

22.1

13.0

13.3

0.12

0.02

25.0

14.9

Na2O

1.06

0.97

1.51

1.38

7.65

9.60

9.12

8.12

0.16

0.05

K2O

0.13

0.01

0.02

7.19

5.76

CaO

11.5

12.6

21.8

22.3

0.02

0.09

0.03

0.03

56.5

57.2

0.03

0.05

P2O5

0.03

36.8

42.0

0.18

6.58

3.57

Cl

0.28

Total

98.2

98.4

101.0

100.4

99.6

99.2

98.9

99.3

101.4

102.9

94.6

94.7

Wo

38.2

39.2

En

45.8

47.1

Fs

15.8

13.6
99.8

99.4

99.8

99.8

Or

0.00

0.10

0.00

0.00

An

0.17

0.50

0.18

0.22

Ab

No. of
oxygens

23

23

32

32

32

32

26

26

22

22

7.89

2.02

2.02

12.2

12.0

12.1

12.3

6.15

5.72

0.01

0.03

4.16

4.18

4.18

3.98

2.09

2.47

0.29

Structural formulae
Si

7.79

Al
Al iv

0.21

0.11

Al vi

0.10

0.04

Ti

Fe3+

0.32

0.10

0.10

Fe2+

0.89

0.36

0.30

0.25

0.01

0.01

Mn

0.01

0.01

0.03

0.01

0.50

2.37

0.00

Mg

3.68

4.50

0.72

0.73

0.01

0.03

0.01

5.19

Na

0.29

0.26

0.11

0.10

3.32

2.54

3.24

3.08

2.72

0.04

0.01

0.02

Ca

1.74

1.85

0.86

1.28

1.10

0.88

0.02

0.01

0.01

11.3

10.6

0.01

5.84

6.15

0.08

1.04

0.52

Cl

0.18

Amphiboles recalc. from (Leake et. al 1997)

10

4.4. Mineral chemistry of magnetite

The magnetite average element composition ratios for Ni/(Cr+Mn) and Ca+Al+Mn vs. Ti+V (wt. %) (Dupuis and Beaudoin, 2011) show signicant differences within and between different ore bodies (Table 4),
(Fig 6). The massive magnetite ore bodies in the eastern part (Vr, Fa and Pz) plots in the eld of Porphyry-Cu
or AIO-eld while ore breccia from Fabian and magnetite ore from the western part of the deposit plots in
the IOCG-eld.
Table 4 The main Fe-Ti oxide mineral association and their oxidation textures show an increasing oxidation from
the eastern to the western part of the deposit exhibiting the exsolution sequence of textures C1 to C7 according to
Haggerty (1991)
Main mineral

Magnetite/Ilmenite

Magnetite/Ilmenite

Magnetite/Hematite

Ore body

C1

C2-C3

C4-C5

ViRi (East)

Magnetite with intergrowths

Homogenous mag- of FeTi-phase. Ilmenite with
netite and ilmenite intergrowth of hematite
lamellas

Hematite porphyroblasts
contain FeTi-phases, as
oxidation surface together
with rutile.

Magnetite with intergrowths

of FeTi-phase. Magnetite
Homogenous magFabian (East)
with exsolved spinel. Ilmenite
netite and ilmenite
with intergrowth of hematite
lamellas
Printzskld

Vlkomma
(West)

Hematite porphyroblasts
contain FeTi-phases, as
oxidation surface together
with rutile

Magnetite with intergrowths

Homogenous mag- of FeTi-phase, Ilmenite with Homogenous magnetite
netite and ilmenite intergrowth of hematite
and hematite
lamellas

Hens (West) -

Magnetite/Hematite

Homogenous magnetite
and hematite

Magnetite with intergrowth of

spinel and rutile. Hematite occurs
as oxidation surface along grain
boundaries or porphyroblasts with
exsolved rutile needles.

Homogenous magnetite
and hematite

Hematite occurs as oxidation

surface along grain boundaries
or porphyroblasts. Foliated and
partly crenulated magnetite and
hematite

Ca+Al+Mn (wt %)

10.0

Vr
Fa
Pz
Hn
V
Breccia

1.0

B
A
0.1

0.0
0.0

0.1

Ti+V (wt %)

1.0

10.0

Fig. 6 Plot of average element composition ratios for

Ni/(Cr+Mn) and Ca+Al+Mn vs. Ti+V (wt. %) of
magnetite from Malmberget deposit according to the
classication by Dupuis and Beaudoin (2011)
A = IOCG, B = porphyry-Cu and C = AIO

Ni/(Cr+Mn) (wt %)

10.0

Vr
Fa
Pz
Hn
V
Breccia

1.0

B
0.1

0.0
0.0

11

0.1

Ti+V (wt %)

1.0

10.0

Samples of magnetite from Fabian exhibit a large chemical variation that partly is related to the character
of the ore. In the massive ore magnetite contains rather high amounts of trace elements, especially MgO
<0.93 wt. %) and Al2O3 (<0.75 wt. %) and TiO2 (< 1.03wt. %), while magnetite from ore breccia has lower
trace element contents (Fig 7a and b).
Table 5 Representative electron microprobe analyses of magnetite, hematite, ilmenite, spinel and rutile
Ore

Vr

Mineral Det.Lim Mag

Sample
ppm
No.

Fa

Pz

Pz

Hn

Hn

Pz

Pz

Mag

Mag

Mag

Mag

Mag

Hem Hem Hem Hem

Fa

Fa

Pz

Hn

Ilm

Ilm

Spl

Rt

834
2:1

495
V40 1 S10 2 S23 1
10:1

495
TV8 1
10:1

495
S28 1
8:2

AP
TV8 3 BC6 1
81

495
S23 2
8:1

44.5

13.1

1.59

FeO

960

94.5

93.1

93.5

94.4

93.9

94.3

84.2

84.0

90.2

90.5

47.5

TiO2

88

0.21

0.47

0.33

0.01

0.01

0.00

7.87

8.39

0.35

0.31

48.32

50.5

6.16

97.7

V2O3

165

0.29

0.28

0.25

0.12

0.12

0.10

0.24

0.31

0.12

0.20

0.28

0.30

0.04

0.57

Al2O3

159

0.07

0.49

0.13

0.07

0.18

0.07

0.05

0.24

61.6

MgO

200

0.13

0.27

0.07

0.13

0.24

0.04

0.10

2.03

0.10

17.3

MnO

205

0.01

0.04

0.03

0.02

0.05

0.09

0.04

1.67

4.26

0.14

0.01

Cr2O3

180

0.01

0.01

ZnO

209

0.01

0.01

0.03

0.01

0.01

0.02

0.50

NiO

156

0.02

0.01

0.02

0.02

0.01

0.02

0.02

0.03

0.02

SiO2

CaO

107

0.01

0.02

Nb2O3

239

0.01

0.01

0.04

0.01

0.01

0.01

SO2

69

0.01

0.01

0.01

CoO

150

0.06

0.01

0.01

0.09

CuO

210

0.01

0.01

0.01

0.02

0.01

0.01

0.01

Total

95.2

94.7

94.3

94.8

94.6

94.7

92.5

92.8

90.8

91.3

99.8

99.7

99.1

99.9

Fe2O3
(calc)

69.8

68.4

68.8

70.1

69.8

70.0

86.0

85.0

99.9

100.3

10.4

3.92

FeO
(calc)

31.7

31.5

31.6

31.4

31.1

31.4

6.84

7.51

0.31

0.26

38.1

40.9

Mineral abbreviation (Kretz, 1983)

4.5. Mineral chemistry in relation to oxidation of the Fe-Ti oxides

The chemistry of magnetite shows rather systematic variation in V2O3, MgO, TiO2 and Al2O3 depending on
local mineral paragenesis and oxidation of the Fe-Ti oxides. The V2O3 content of magnetite is used to dene
two subgroups (Fig. 7d). Magnetite that coexists with ilmenite contains higher amounts of V2O3 (0.19-0.42
wt. %); this is the case for the ViRi, Fabian and Printzskld ore bodies. Magnetite that coexists with hematite,
on the other hand, shows lower V2O3 (0.09-0.19 wt. %) content; this is the case for the Printzskld, Hens
and Vlkomma ore bodies. The MgO content in magnetite shows some variation (0.00-0.31 wt. %) but the
average amount for individual ore bodies displays no signicant differences. The TiO2 content in magnetite
displays no distinct patterns (Fig. 7c). However, in ViRi, Fabian and Printzskld the magnetite coexists with
ilmenite and has generally higher amounts of TiO2 (0.05-0.80 wt. %) than does the magnetite coexists with
hematite, as occurs in Printzskld, Hens and Vlkomma (0.00-0.05 wt. %). For hematite the TiO2 content can
be used to clearly dene two subgroups (Fig. 7e). The hematite which occurs as porphyroblasts in magnetite
ore contains high amounts of TiO2 (6-12 wt. %) and slightly higher amounts of V2O3 (0.18-0.35 wt. %) (Pz1
& Hn1) compared to hematite from the hematite dominated ore and hematite occurring as a late oxidation of

12

magnetite (0.02-1.26 wt. % TiO2 and 0.08-0.22 % V2O3) (Pz2 & Hn2). Hematite from hematite dominated ore
shows instead higher amounts of Al2O3 (0.05-0.24 wt. %) (Fig 7f), which is comparable to the Al2O3 content
in magnetite (0.00-0.55 wt. %).
1.00

log MgO (wt.%)

log TiO2 (w.t%)

1.00

0.10

Ore

0.01

Breccia

0.00
0.00

0.20

0.40

0.60

0.80

0.10

Ore

0.01

Breccia

0.00
0.00

1.00

0.20

0.40

0.60

0.80

1.00

Al2O3 (wt.%)

0.80

Vr

0.70

Fa
Pz 1

0.60

Hn

0.40

Fa
Pz 1

0.40

Pz 2

0.50

Vr

0.50

MgO (wt.%)

TiO2 (wt.%)

Al2O3 (wt.%)

0.30
0.20

Pz 2
0.30

Hn
V

0.20
0.10

0.10
0.00
0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.00
0.00

0.80

0.05

0.10

Al2O3 (wt.%)

TiO2 (wt.%)

10.00
8.00
6.00

0.25

0.30

0.35

0.40

0.45

0.30
Pz 1

Pz 1
0.25

Pz 2
Hn 1
Hn 2
V

4.00

Pz 2
Hn 1

0.20

Hn 2
0.15

0.10
0.05

2.00
0.00
0.00

0.20

V2O3 (wt.%)

Al2O3 (wt.%)

12.00

0.15

0.05

0.10

0.15
0.20
0.25
V2O3 (wt.%)

0.30

0.35

0.00
0.00

0.40

0.05

0.10

0.15
0.20
0.25
V2O3 (wt.%)

0.30

0.35

0.40

Fig. 7 The chemical composition of magnetite and hematite analysed by electron microprobe of ViRi, Fabian,
Printzskld, Hens and Vlkomma ores A) The Al2O3/TiO2 variation in magnetite from massive ore and ore breccia of
Fa, expressed as log TiO2 B) The Al2O3/MgO variation in magnetite from massive ore and ore breccia of Fa, expressed
as log MgO C) The V2O3/MgO variation in magnetite from Vr, Fa, Pz, Hn and V D) The Al2O3/TiO2 variation in
magnetite from Vr, Fa, Pz, Hn and V. E) The V2O3/TiO2 variation in hematite shows two subgroups from Pz, Hn and
V. F) The V2O3/Al2O3 variation in hematite shows two subgroups from Pz, Hn and V

4.6. Thermometry and oxygen barometry

The temperature and oxygen fugacity prevailing during mineral equilibrium in the Malmberget deposit is
estimated using the magnetite-ilmenite geothermobarometry program ILMAT120 (Lepage, 2003) which is
based on the geothermometer of Andersen and Lindsley (1985). The calculated temperature (C) and oxygen
fugacity, expressed as log10fO2 are based on coexisting magnetite and ilmenite either as mineral pair or as
magnetite with intergrowths of ilmenite (Fig 8) in the For the ViRi, Fabian and Printzskld ore bodies. The
average temperature for the different ores shows approx. 500C and the average oxygen fugacity is between 20
and 21 log10fO2 which is close to the border of the magnetite-hematite stability elds.

13

T (C)
300
10

400

500

600

700

Hem + Rt
Hem + Ilm
15

-log 10 f(O2)

He

ti
ma

te

+M

n
ag

eti

te

Mgn + Rt

20

25

Mgn + Ilm
Vr
Fa
Pz

30

Fig. 8 The calculated temperature and oxygen fugacity based on the coexisting magnetite and ilmenite from ViRi,
Fabian and Printzskld ore together with the theoretical stability elds, modied from Annersten, 1968

5. DISCUSSION
5.1. Apatite iron ore general aspects
The apatite iron ore in the Malmberget deposit has been exposed to extensive regional metamorphism in upper amphibolites facies, intrusion of granites, pegmatites and aplites, strong ductile deformation and alteration
events. As a result few primary structures and textures have been preserved in the ore and the surrounding host
rocks. This is in contrast to the apatite iron ores in the Kiruna area, which do retain such features (Martinsson
and Virkkunen, 2004; Bergman et al., 2001; Geijer, 1930). Therefore a primary goal of study is to identify
chemical-textural characteristics related to primary ore forming conditions and to compare these with the
more well preserved apatite iron deposits in northern Norrbotten (e.g. Kiirunavaara, Luossavaara, Per Geijer
ores (Geijer, 1910; Nystrm and Henrques, 1994; Frietsch, 1970; Frietsch, 1982; Park, 1975), Gruvberget
(Lindskog, 2001), Tjrrojokka (Edfelt, 2007).
Many different parameters control the substitution of trace elements in ore forming minerals, including composition of the magma/ore uid, host rock composition, pressure, temperature, cooling rate, oxygen fugacity,
the local depositional environment and postdepositional recrystallization (Buddington and Lindsley, 1964, Annersten 1968, McQueen and Cross, 1998, Singoyi, 2006). In general the trace element chemistry of magnetite
from Malmberget show similarities to other apatite iron ores and matches the published data from locations
such as Kiirunavaara (Nystrm and Henrques, 1994; Frietsch, 1970; Park, 1975; Loberg and Horndahl, 1983;
Mller et al., 2003). The characteristic chemical signature for apatite iron ores is a high content of V2O3 and
TiO2 and low contents of CoO, Cr2O3, NiO and MnO (Frietsch, 1970; Helvaci, 1984; Nystrm and Henrques, 1994). The origin of an ore deposit is important; iron oxides will have different element concentrations
depending on if the deposit has formed from an iron oxide magma or from hydrothermal uids e.g. magmatic
deposits usually have higher contents of TiO2 than do deposits formed from iron oxide magma (Hildebrand,
1986; Naslund et al., 2000).

14

5.2. Chemical and mineralogical features of primary character

The trace element composition of Fe-oxides is mainly controlled by fractionation during crystallisation; changes related to subsolidus processes appear to be minimal, with only local redistribution of the trace elements
during exsolutions and oxidation processes (Buddington and Lindsley, 1964; Dare et al., 2012) (see section 5.4).
Whole rock analyses of ore may, therefore, reect primary features of the Malmberget deposit. The Fabian ore
body shows rather distinct differences between the massive ore and ore breccia but also signicant differences
within the massive ore that is also partly reected by the gangue mineral composition. The most obvious difference is the variation in P, Ti,V and Ga, which is lower in the ore breccia but also shows varying contents in
different part of the massive ore with sequences of massive ore up to 40 m thickness that have different chemical and mineralogical compositions. Sections enriched in Ti,V and Ga are separated by sections of P-rich ore
and Fe-rich sections with low content of P, Ti,V and Ga. These chemically different ore sections could be correlated between the drill holes at Fabian (Fig 9) and are suggested to reect primary variation within the ore.

100
6496

6463

6533

6500

Ti/V
P
Fe
100

200

300

400

500

600

700

800

900

Fig. 9 A cross-section through the western part of the Fabian ore body divided into geochemical subtypes of massive
ore based on variations in Ti,V, P and Ga contents

15

There is an extensive alteration zone of amphibole-pyroxene skarn and ore breccia in the structural hanging
wall of the Fabian ore body, which contains varying amount of sulphides and anhydrite.These minerals are also
present in the massive ore, where they occur mainly as disseminate grains or as smaller veins and patches and are
also mainly restricted to sections of ore rich in Ti-V-Ga. Similar skarn alteration zones and ore breccia related
to massive ore are also described at Kiirunavaara (Martinsson and Hansson, 2004) and at the Tjrrojokka apatite
iron deposit (Edfelt, 2007) although these generally lack anhydrite.
Magnetite has been shown to have variations in chemical compositions that are related to its origin. In particular the average element composition of magnetite from different types of ore deposits has been demonstrated
(Dupuis and Beaudoin, 2011). Magnetite from Malmberget plot in both the IOCG and the AIO elds using
the diagram with Ni/(Cr+Mn) or Ca+Al+Mn vs. Ti+V (wt.%). Massive ore from ViRi and Fabian plot in
the AIO eld, while the Hens and Vlkomma ore bodies and ore breccia from Fabian plot in the IOCG eld
(Fig. 6). The chemical variations in magnetite can partly be explained by the presence of ilmenite and FeTi-oxide phases due to the exsolutions and oxidation processes (see section 5.4) but are largely reecting the
geochemically trends of the ores. Generally the mineral chemistry of magnetite in the Fabian ore body shows
variations mainly related to the type of ore (massive vs. ore breccia). Magnetite from ore breccia is very low in
trace elements and magnetite from massive ore is richer in TiO2 and MgO with the highest content of TiO2
and V2O3 in ore sections enriched in Ti-V-Ga. The low Ti content of magnetite is also reected by the lack of
exsolved ilmenite lamellas.

5.3. Magnetite texture and grain size related to metamorphic recrystallization

The textural characteristic of the ore minerals in Malmberget is rather different compare to other apatite iron
ores e.g. Kiirunavaara (Nystrm and Henrques, 1994), Avnik (Helvaci, 1984) and Bafq (Bonyadi et al., 2011).
This probably is an expression of metamorphic recrystallization at different stages and conditions, and this feature is also displayed by the metavolcanic host rock that has been transformed into gneiss and a metamorphic
recrystallization of dykes cutting the ore. Within the Malmberget deposit small differences such, as the morphologic character of magnetite and hematite (grain size, shape, and textures), mineral composition, the chemical composition of the oxide minerals indicate slightly higher metamorphic conditions in the western part
compared to the eastern part.This is in agreement with Romer (1996), who drew his conclusions based on the
silicate mineralogy of the host rocks.The metamorphic recrystallization is expressed through varying grain size
of the ore minerals, where the common coarse grained texture is characterised by straight grain boundaries and
distinct triple junction congurations. Another common texture is large magnetite porphyroblasts occurring
as series of grains in vein-like textures. This selective growth of grains is typical for diffusive controlled metamorphic growth where the surrounding minerals will provide elements to those grains, acting as nucleating
agents (Vernon, 2004 references therein). A coarse grained character and occurrences of both magnetite and
hematite porphyroblasts are especially typical for the Hens and Vlkomma ore bodies. Another feature related
to recrystallization and deformation is the foliated and folded magnetite and hematite grains in the Vlkomma
ore body. This is interpreted to be caused by a late and rather brittle local deformation. Small magnetite grains
occur as inclusions in larger porphyroblasts of plagioclase and amphibole generated by the prograde metamorphism where the small grains could have been engulfed by porphyroblasts during their growth (Vernon, 2004
and references therein).

16

5.4. Textures and chemistry of Fe-Ti oxides related to oxidation of magnetite

Iron and titanium oxide minerals reveal a lot of paragenetic information about the rocks containing them and
their microtextures provide important knowledge of the prevailing oxygen pressure and temperature (Buddington and Lindsley, 1964). Magnetite and ilmenite can coexist in most parts of the stability eld for magnetite
but will change their composition in response to changes of temperature and oxygen fugacity (Lindsley, 1991).
This mineral pair represents the lowest degree of oxidation in the Malmberget deposit (Annersten, 1968).
The magnetite from ViRi, Fabian and Printzskld shows textures and mineral assemblages representing progressive oxidation from homogeneous magnetite, to exsolution lamellae of ilmenite to an intense oxidation
assemblage of hematite and rutile. Magnetite with exsolved ilmenite as trellis and sandwich lamellae indicates
an oxidation at low to moderate pressure and above 600 C (Buddington and Lindsley, 1964). Trellis texture is
dened as thin lamellae along the {111} planes in magnetite and sandwich texture is dened as thicker lamellae
along one of the {111} plane (Lindsley, 1991). These lamellae are concentrated in the centre of the magnetite
grains. Similar exolution textures in magnetite have also been described from El Laco and are interpreted to
be the result of post magmatic cooling and oxidation (Alva-Valdivia et al., 2003). Magnetite exposed to more
intense oxidation exhibit an alteration to hematite and lamelleas of rutile and could result in a complete pseudomorphic replacement by hematite (Mcke and Cabral, 2005). In the Printzskld ore body, the increasing
oxidation is expressed by the changes from magnetite-ilmenite to magnetite-hematite as the main oxide mineral assemblage. Hematite in association with magnetite contains ilmenite lamella which indicates high titanium
content in hematite at the time of original crystallisation at a high temperature (Annersten, 1968). Haggerty
(1991) describe a general exsolution sequence of textures in seven steps due to increased oxidation. Trellis and
sandwich textures correspond to step C2-C3, and magnetite as host will be replaced by hematite and rutile
(C6-C7) in areas of former ilmenite step C4-C5. Magnetite from Hens and Vlkomma ore bodies both show
an extensive martitisation of magnetite and porphyroblastic growth of hematite. The abundant occurrence of
hematite in association to iron poor silicates indicates a higher degree of oxidation (Annersten, 1968). Besides
a complete alteration of magnetite to hematite, there are different martite textures shown as incomplete martitization along grain boundaries, fractures and crystal faces which is assumed to be a late alteration product
(Annersten, 1968).
The magnetite and hematite from the different ore bodies (ViRi, Fabian, Printzskld, Hens and Vlkomma)
display some systematic chemical variations for the trace elements V2O3,TiO2, Al2O3 and MgO.The oxide mineral assemblages and their textures are expressions of the local oxidation conditions of the ore bodies, which has
controlled the redistribution of elements among magnetite and hematite during metamorphic recrystallization.
In ViRi, Fabian and part of the Printzskld ore body magnetite-ilmenite is the dominating mineral pair while
in Hens,Vlkomma and the part of Printzskld the dominated oxide mineral pair is magnetite-hematite.
In ViRi, Fabian and part of the Printzskld ore body, where magnetite-ilmenite is the predominate mineral
pair, the magnetite shows a varying but higher content of TiO2 compare to magnetite in Hens,Vlkomma and
the part of Printzskld were the predominate oxide mineral pair is magnetite-hematite. This is interpreted to
be related to the increased oxidation westwards expressed by the oxidation textures, which causes the magnetite to begin to exsolve TiO2 due to intensive diffusion and a decreasing temperature (Ramdohr, 1980), which
ultimately results in trellis and sandwich textures (Mcke, 2003). Magnetite is nearly the pure end member
in the Hens and Vlkomma ore bodies, and also in the part of Printzskld ore body which contains hematite;

17

this is also illustrated by the absence in these locations of the earlier described Fe-Ti lamellas. In the Hens and
Printzskld ores the coexisting hematite occurs as porphyroblasts with a high TiO2 content, which is in marked
contrast with the TiO2 poor hematite formed by martitisation. The high Ti-content of hematite increases its
stability at a lower oxygen pressure (Lindsley, 1962) than would characterise ore with coexisting magnetite-hematite. At Vlkomma hematite has a rather low TiO2 content and this might be the result of insufcient
supply of TiO2 during the oxidation of magnetite to hematite since most of the magnetite has been replaced
by hematite.
Magnetite grains that coexist with ilmenite will be depleted in MgO compared to magnetite grains surrounded only by magnetite as a result of chemical redistribution during sub-solidus exsolution lamellae/oxidation
processes (Dupuies and Beaudoin, 2011). In the Printzskld and Hens ore bodies, where the coexistent mineral
pair is magnetite and hematite, those magnetite grains which are MgO enriched contain exsolved spinel textures. This is probably due to the same redistribution mechanism as for the TiO2 by intensive diffusion and a
decreasing temperature (Ramdohr, 1980). There is also some variation in the Al2O3 content of magnetite and
hematite, with slightly higher contents in magnetite from ViRi and Fabian compared to Vlkomma, and hematite from Vlkomma has higher contents compared to the porphyroblastic hematite at Hens.
The vanadium content in magnetite coexisting with ilmenite or hematite at equilibrium is dependent on the
prevailing oxygen fugacity and temperature (Schuiling and Feenstra, 1980). The ilmenite coexisting with magnetite in the ViRi and Fabian ore bodies is enriched in V2O3. Schuiling and Feenstra (1980) have shown that
ilmenite in magnetite-ilmenite pairs which forms fairly close to the magnetite-hematite buffer will contain
more V2O3 than magnetite because at these conditions V is more common as V4+ and will thus substitute for
Ti4+. However, in the Fabian and ViRi the amount of ilmenite is small and the magnetite has a rather high V2O3
content. In magnetite-hematite pairs, when the hematite-magnetite boundary is approaching, the hematite will
concentrate more V, since there are only 2/3 as many available sites in the coexisting magnetite where the V4+
can substitute for Fe3+. Therefore, low content of V2O3 in magnetite in Hens and Vlkomma ore compared to
ViRi and Fabian ore is probably an expression of the redistribution of vanadium during extensive formation of
hematite as porphyroblasts; this feature is particularly obvious in the Printzskld ore, where both mineral pairs
are present (g 7d&f).

5.5. Thermometry and oxygen barometry

Norrbotten is a low to intermediate pressure province, metamorphosed under variable conditions; the few
temperature-pressure determinations which have been made indicate ranges of 500-800 C, and 2-6kbar
(Bergman et al., 2001). At the Malmberget deposit the metamorphic temperature is estimated to 500-550 C
using the coexisting iron oxide and Mg-Fe silicates. Temperature and oxygen pressure have been calculated for
the last mineral equilibrium that was close to the stability eld of co-existing magnetite and hematite (Annersten, 1968). Our data using the magnetite-ilmenite pair give very similar results for both temperature and
oxygen fugacity. Although Sauerzapf et al. (2008) showed from experimental work that there are some uncertainties in using this thermo-oxybarometer for temperatures below 600 C; the metamorphic conditions at the
Malmberget deposit have, most probably, reached at least 550 C. Likewise, the oxygen fugacity has likely been
close to the magnetite-hematite buffer, as illustrated by the occurrence of several generations of magnetite and
hematite growth.

18

5.6. Evolution of the Malmberget deposit

The Malmberget deposit shares many features with the apatite iron deposits in the Kiruna area but there are
also important differences. Some of these might be related to primary differences but most of them are probably
the result of deformation (mineral orientation, internal structure, geometry of ore bodies) and metamorphic
recrystallization (grain size, texture, Fe-Ti oxide paragenesis) of the ore bodies at Malmberget, which have a
higher metamorphic grade.
Primary features may include large scale differences of character, mineralogical and chemical compositions and,
to some extent, differences in the average mineral chemistry of magnetite for the ore bodies at Malmberget.
The massive magnetite ore bodies that are characteristic of the eastern part of Malmberget have mineral assemblages (low to moderate amounts of apatite and amphibole as the main gangue minerals), whole rock chemical
composition (high Fe, Ti,V) and magnetite composition (high Ti,V) that are very similar to the B-type of ore
at Kiirunavaara (Park, 1973; Nystrm & Henriques, 1994; Martinsson, 2003).There are high phosphorous ores
occurring as distinct units within the massive ore at both the Fabian and the Kiirunavaara. In addition Fabian
also exhibits high Ti-V-Ga ore which is low in apatite in several well dened subunits; these indicate that the
massive ore originated from several chemically different injections of magma. These high Ti-V-Ga ore units
occur as several rather regular bodies up to 40 m wide that can be traced between the drill holes at Fabian
(Fig. 9). At ViRi most of the ore also has this composition. In both deposits minor anhydrite and sulphides are
associated with the high Ti-V-Ga ore indicating a high volatile content of the magma.
In the eastern part of Malmberget, ore breccia typically surrounds the massive ore, although the extent and
mineralogical composition varies slightly between the different ore bodies in this area and also as compared
with similar ore breccia associated with the Kiirunavaara deposit. At Kiirunavaara the occurrence of massive
ore surrounded by ore breccia has been suggested to represent the transition from magmatic to hydrothermal
ore forming processes (Martinsson, 2004) and a similar interpretation is also suggested here for Malmberget,
where magnetite from breccia ore is chemically different from the massive ore at Fabian. The lower content
of trace elements in magnetite and the occurrence of abundant anhydrite together with amphibole-pyroxene
and minor amounts of pyrite-chalcopyrite in ore breccia is a strong indication for a hydrothermal origin. This
is further supported by the IOCG characteristic of the magnetite from the ore breccia in contrast to the AIO
character the massive ore.
The ore bodies in the western part of the Malmberget deposit are rather different in character, mineralogical
composition and magnetite chemistry as compared to those in the eastern part. The ore bodies in the western
part are similar to the Per-Geijer ores at Kiruna, particularly regarding their banded character, the occurrence
of hematite, and the high apatite content (Park, 1973). A common feature of these ores is also the minor occurrence of ore breccia related to the massive ore.
At the Kiruna area the Per-Geijer ores are situated stratigraphically above the Kiirunavaara ore body within the
Kiirunavaara Group in association with felsic volcanic rocks, while the Kiirunavaara ore body itself occurs at
the border of the felsic unit and underlying intermediate rocks (Geijer, 1910; Martinsson, 2004). In Malmberget
the local stratigraphy of the ore host rocks is not well constrained. However, according to preliminary results,
the host rock of the deposits in the western part of Malmberget seems to have a more felsic composition than
does the host rock of the deposits in the eastern part, which has a more intermediate to mac composition.
This might indicate a different, and most likely a higher stratigraphic position, for the Vlkomma and Hens ore

19

bodies compared to Fabian and ViRi. This is also supported by the decreasing trend of the average V and Ti
content in ores from east to west at Malmberget, which mimics the decreasing content of these elements from
Kiirunavaara to the stratigraphically higher Per-Geijer ores in Kiruna (Fig. 3).This pattern is also shown by the
Ekstrmsberg ore, which is situated within the felsic and upper part of the Kiirunavaara Group and has similar
low Ti-V contents and includes a signicant amount of hematite (Frietsch, 1974).
The Printzskld ore body is spatially situated in the middle of the Malmberget deposit and has characteristics
which fall between those of the ore bodies in the eastern and western parts with respect to its mineral composition and to the chemistry of its magnetite.This ore body has a lenticular form, often a strongly banded structure,
contains boudinaged pegmatites parallel to banding and a strong schistosity is developed in the biotite-rich
rocks in the tectonic footwall.This might reect a position of this ore body within or close to a shear zone that
also could have caused a displacement of the ore body from its original position.
The structural and metamorphic evolution during the Svecokarelian Orogeny includes at least two major
events in northern Sweden at 1.9 and 1.8 Ga, but the conditions prevailing during metamorphic recrystallization for each of them is not known (Bergman et al., 2001, 2006). However, in areas like Malmberget that
includes large volumes of 1.8 Ga S-type Lina granites the younger metamorphic event is suggested to represent
peak metamorphic conditions (Bergman et al., 2001) at a temperature of 650-700 C and a pressure of 2-4 kbar
(Annersten, 1968). Stilbite-bearing mineral assemblage at Malmberget indicates that during post metamorphic
cooling the temperature did not exceed 150 C after 1.62-1.60 Ga. (Romer, 1996).
The present character of the Malmberget ore bodies is the result of deformation and strong metamorphic
recrystallization during the 1.9 and 1.8 Ga events that have modied and partly obliterated the primary characters including ore geometry, internal structure, mineral composition, textures and mineral chemistry. The most
obvious change is the increase in grain size and the change in oxidation conditions from the eastern (ViRi,
Fabian) to the western part (Hens,Vlkomma) of the deposit. These variations within the deposit are probably
due to more abundant granitic intrusions and a position closer to the large Lina granite intrusion to the north
for the western part. The interaction between Lina granite and the ore is also reected by the common occurrence of magnetite, hematite and apatite in Lina pegmatites within the ore deposit (Geijer, 1930).

6. CONCLUSION
At the Malmberget apatite iron deposit the Viri, Fabian, Printzskld, Hens and Vlkomma ore bodies show
chemical, mineralogical and textural characteristics that are specic to each ore body. Variations in the whole
rock chemistry of the ores are primarily due to primary features of the ores, with different signatures for the
massive ore and the ore breccia. However, there are also indications of different subtypes of massive ore by their
variations in levels of P,Ti,V and Ga.These ore sections are also characterised by the specic gangue mineralogy
of apatite, amphibole, sulphides and anhydrite. Magnetite from the ore breccia has a low content of trace elements similar to magnetite from IOCG type deposits and may have formed by hydrothermal processes, while
magnetite in the massive ore generally has a higher trace element content and shows chemical characteristics
typical for apatite iron ores. The massive ore may have a magmatic origin and, based on the occurrence of
chemical subtypes; it may have formed by multiphase injections of iron oxide magma.
Interpreted primary features at Malmberget (whole rock chemistry of the ore, magnetite chemistry, mineral
composition and large scale characteristics of the ores) show many similarities to the apatite iron ores in the

20

Kiruna area (Kiirunavaara and Per Geijer ores). Based on analogies with the Kiruna area, and the possible differences in host rock composition from the eastern and western part of the Malmberget deposit, it cannot be
excluded that the Malmberget ore bodies occur at different stratigraphic positions.
The Malmberget deposit has been affected by metamorphic recrystallization and oxidation at temperatures of
at least 550 C and an oxygen fugacity close to the magnetite-hematite buffer. This has caused an increase in
grain size, changes in textures, and Fe-Ti oxide assemblages. As a result the chemistry of magnetite has, to various extents, been modied due to the elemental redistribution between the coexisting Fe-Ti oxide minerals.
Most signicant is the preferential partition of Ti, and to some extent V, into porphyroblasts of hematite. Also
the formation of ilmenite and rutile during oxidation has affected the chemistry of magnetite and resulted in a
signature resembling magnetite from IOCG type deposit, while the more primary magnetite has a composition
typical for apatite iron ores. The increased grain size caused by metamorphic recrystallization has developed a
granoblastic texture of the ore with distinct triple junctions and simple grain boundaries for magnetite and the
Fe-Ti oxide associations indicates a progress oxidation from east to west in the deposit with hematite occurring
as a main mineral in the western part. As a result the textures of ores and the chemistry of magnetite are, in
part, very different from the Kiirunavaara deposit; this has importance for mineral processing and metallurgical
aspects of the deposit.

7. ACKNOWLEDGEMENT
Hjalmar Lundbohm Research Centre (HLRC) is greatly acknowledged for the funding of this project. The
authors want to express their gratitude to Dr Robert Fairhurst, Camborne School of Mines, UK and Bo
Johansson at GTK in Espoo, Finland for excellent support during microprobe analyses. People at LKAB are
indebted for their contribution of data and knowledge. Our colleague at the LTU, Riia Chmielowski is greatly
acknowledged for proof reading the manuscript.

21

8. REFERENCES
Annersten, H., 1968, The mineral chemical study of a metamorphosed iron formation in northern Sweden:
Lithos, v. 1, p. 374-397.
Andersen, D.J. and Lindsley, D.H., 1985, New (and nal!) models for the Ti-magnetite/ilmenite geothermometer and oxygen barometer: American Geophysical Union, AGU Spring Meeting Eos Transactions, 66
(18) 416 p.
Alva-Valdivia, L., Rivas, M.L., Goguitchaichvili, A., Urrutia-Fucugauchi, J., Gonzalez, J.A., Morales, J., Gmez, S., Henrquez, F., Nystrm, J.O. and Naslund, R.H., 2003, Rock-Magnetic and Oxide Microscopic
Studies of the El Laco Iron Ore Deposits, Chilean Andes, and Implications for Magnetic Anomaly Modeling: International Geology Review, v. 45, p. 533-547.
Bergman, S., Billstrm, K., Persson, P., Skild, T. and Evins, P., 2006, U-Pb age evidence for repeated Palaeoproterozoic metamorphism and deformation near the Pajala shear zone in the northern Fennoscandian
shield: GFF, v. 128, p. 7-20.
Bergman, S., Kbler, L. and Martinsson, O., 2001, Description of regional geological and geophysical maps of
northern Norrbotten County (east of Caledonian orogen): SGU Geological Survey of Sweden, v. Ba 56, p.
110.
Bonyadi, Z., Davidson, G.J., Mehrabi, B., Meffre, S. and Ghazban, F., 2011, Signicance of apatite REE depletion and monazite inclusions in the brecciated SeChahun iron oxideapatite deposit, Bafq district, Iran:
Insights from paragenesis and geochemistry: Chemical Geology, v. 281, p. 253-269.
Buddington, A.F. and Lindsley, D.H., 1964, Iron-Titanium Oxide Minerals and Synthetic Equivalents: Journal
of Petrology, v. 5, p. 310-357.
Cliff, R.A., Rickard, D. and Blake, K., 1990, Isotope Systematics of the Kiruna Magnetite Ores, Sweden: Part
1. Age of the Ore: Economic Geology, v. 85, p. 1770-1776.
Dare, S.A.S., Barnes, S. and Beaudoin, G., 2012,Variation in trace element content of magnetite crystallized
from a fractionating sulde liquid, Sudbury, Canada: Implications for provenance discrimination: Geochimica et Cosmochimica Acta, v. 88, p. 27-50.
Droop, G.T.R., 1987, A general equation for estimating Fe3+ concentrations in ferromagnesian silicates and
oxides from microprobe analyses, using stoichiometric criteria: Mineralogical Magazine, v. 51, p. 431435.
Dupuis, C. and Beaudoin, G., 2011, Discriminant diagrams for iron oxide trace element ngerprinting of
mineral deposit types: Mineralium Deposita, v. 46, p. 319-335.
Edfelt, ., 2007, The Tjrrojkka Apatite-Iron and Cu (-Au) Deposits, Northern Sweden, Lule, Lule University of Technology, 167 p.
Frietsch, R., 1984, Petrochemistry of the iron ore-bearing metavolcanics in norrbotten county northern
Sweden: Geological Survey of Sweden, v. Serie C, p. 62.
Frietsch, R., 1982, On the Chemical Composition of the Ore Breccia at Luossavaara, Northern Sweden:
Mineralium Deposita, v. 17, p. 239-243.
Frietsch, R., 1974, The Ekstrmsberg iron ore deposit, northern Sweden: Geological Survey of Sweden, v.
Serie C, nr 708, p. 52.
Frietsch, R., 1970, Trace elements in magnetite and hematite mainly from northern Sweden.: Geological
Survey of Sweden, v. Serie C, p. 136.
Geijer, P., 1931, The iron ores of the Kiruna type, geographical distribution, geological characters and origin:
Geological Survey of Sweden, v. Serie C, p. 1-39.
Geijer, P., 1930, Geology of the Gllivare Iron Ore eld: Geological Survey of Sweden, v. Ca 22, p. 1-115.
Geijer, P., 1910, Geology of the Kiruna district 2: Igneous rocks and iron ores of Kirunavaara, Loussavaara and
Tuollavaara: Stockholm, Kungliga tryckeriet F. A. Norstedt & Sner, 278 p.
Haggerty, S.E., 1991, Oxide textures - A mini-atlas, in Lindsley, D.H., ed., Reviews in Mineralogy: Washington, D.C, U.S, Mineralogical Society of Amerika, p. 129.
Helvaci, C., 1984, Apatite-Rich Iron Deposits of the Avnik (Bingl) Region, Southeastern Turkey: Economic
Geology, v. 79, p. 354-371.

22

Henrques, F., Naslund, H.R., Nystrm, J.O.,Vivallo, W., Aguirre, R., Dobbs, F.M. and Lled, H., 2003, New
eld evidence bearing on the origin of the El Laco magnetite, northern Chile - a discussion: Economic
Geology, p. 1497-1502.
Hildebrand, R.S., 1986, Kiruna-type Deposits: Their Origin and Relationship to Intermediate Subvolcanic
Plutons in the Great Bear Magmatic Zone, Nothwest Canada: Economic Geology, v. 81, p. 640-659.
Hitzman, M.W., Oreskes, N. and Einaudi, M.T., 1992, Geological characteristics and tectonic setting of Proterozoic iron oxide (Cu-U-Au-REE) deposits: Precambrian Research, v. 58, p. 241-287.
Leake, B.E., Woolley, A.R., Arps, C.E.S., Birch, W.D., Gilbert, M.C., Grice, J.D., Hawthorne, F.C., Kato, A.,
Kisch, H.J., Krivovichev,V.G., Linthout, K., Laird, J., Mandarino, J.A., Maresch, W.V., Nickel, E.H., Rock,
N.M.S., Schumacher, J.C., Smith, D.C., Stephenson, N.C.N., Ungaretti, L., Whittaker, E.J.W. and Youzhi,
G., 1997, Nomenclature of amphiboles: Report of the Subcommittee on Amphiboles of the International
Mineralogical Association, Commission on New Minerals and Minerals Names: American Mineralogist, v.
82, p. 1019-1037.
Lepage, L.D., 2003, ILMAT: an Excel worksheet for ilmenite-magnetite geothermometry and geobarometry:
Computers & Geosciences, v. 29, p. 673-678.
Lindskog, L., 2001, Relationships between Fe-oxide Cu/Au deposits at Gruvberget - Kiruna district Northern Sweden, Townsville, Australia, James Cook University of Northern Queensland, 123 p.
Lindsley, D.H., 1991, Oxide Minerals: Petrologic and magnetic signicance: Blacksburg,Virginia, U.S, Mineralogical Society of America.
Lindsley, D.H., 1962, Investigations in the system FeO - Fe2O3- TiO2, Carnegie Institution of Washington
Year Book, 61, 100-106 p.
LKAB, 2011, Annual Report: Lule, Sweden, Lule Graska, 122 p.
Loberg, B.E.H. and Horndahl, A., 1983, Ferride Geochemistry of Swedish Precambrian Iron Ores: Mineralium Deposita, v. 18, p. 487-504.
Lund, C. and Martinsson, O., 2008, A characterising of the ore minerals due to mineralogical, chemical and
textural properties in Malmberget.: Conference in Minerals Engineering, Lule, Sweden, Proceedings, p.
71-80.
Martinsson, O., berg, E. and Fredriksson, A., 2012, Apatite for extraction - Mineralogy of apatite and REE
in the Kiirunavaara Fe-deposit: XXVI International Mineral Processing Congress-IMPC 2012, New Delhi, India, 24-28 September.
Martinsson, O., 2004, Geology and Metallogeny of the Northern Norrbotten Fe-Cu-Au Province: Society of
Economics Geologists, Guidebooks Series, v. 33, p. 131-148.
Martinsson, O. and Virkkunen, R., 2004, Apatite Iron Ore in the Gllivare, Svappavaara, and Jukkasjrvi Areas:
Society of Economics Geologists, Guidebooks Series, v. 33, p. 167-172.
Martinsson, O. and Hansson, K.E., 2004, Day seven eld guide Apatite Iron Ores in the Kiruna Area: Society
of Economic Geology, Guidebook Series, v. 33, p. 173-175.
Martinsson, O., 2003, Characterisation of iron mineralisations of Kiruna type in the Kiruna area, northern
Sweden: Proceedings of the Seventh Biennial SGA Conference Mineral Exploration and Sustainable Development, Athens, Greece, 24-28 August 2003, Proceedings, p. 1087-1090.
McQueen, K.G. and Cross, A.J., 1998, Magnetite as a geochemical sampling medium: application to skarn
deposits: Geological Society of Australia Special Publication, v. 20, p. 194-199.
Mcke, A., 2003, Magnetite, ilmenite and ulvite in rocks and ore deposits: petrography, microprobe analyses
and genetic implication: Mineralogy and Petrology, v. 77, p. 215-234.
Mcke, A. and Cabral, A.R., 2005, Redox and nonredox reactions of magnetite and hematite in rocks: Chemie der Erde Geochemistry, v. 65, p. 271-278.
Mller, B., Axelsson, M.D. and hlander, B., 2003, Trace elements in magnetite from Kiruna, northern Sweden, as determined by LA-ICP-MS: GFF, v. 125, p. 1-5.
Naslund, H.R., Aguirre, R., Dobbs, F.M., Henrques, F. and Nystrm, J.O., 2000, The Origin, emplacement,
and eruption of the magmas: IX Congreso Geologico Chileno, Santiago, Chile, Proceedings, p. 135-139.
Nystrm, J.O. and Henrques, F., 1994, Magmatic Features of Iron Ores of the Kiruna Type in Chile and Sweden: Ore Textures and Magnetite Geochemistry: Economic Geology, v. 89, p. 820-839.

23

Park, T., 1975, Kiruna Iron Ores Are Not Intrusive-Magmatic Ores of the Kiruna Type: Economic Geology, v. 70, p. 1242-1258.
Park, T., 1975a, The origin of the Kiruna iron ores: Geological Survey of Sweden, v. C 709, p. 209.
Park, T., 1973, Kirunamalmernas bildning: Doctoral Thesis, Stockholm, Stockholm University, 222 p.
Park, C.F.,Jr, 1961, A magnetite ow in Northern Chile: Economic Geology, v. 56, p. 431-441.
Potts, P.J., Bowles, J.F.W., Reed, S.J.B. and Cave, M.R., 1995, Microprobe Techniques in the Earth Sciences:
London, UK, Chapman & Hall.
Ramdohr, P., 1980, The Ore minerals and their intergrowth: Berlin, Pergamon press, 1207 p.
Rhodes, A.L. and Oreskes, N., 1999, Oxygen isotope composition of magnetite deposits at El Laco, Chile:
Evidence of formation from isotopically heavy uids, in Skinner, B.J., ed., Geology and ore deposits of the
central Andes, Society of Economic Geologists, Special Publication 7, p. 333-351.
Romer, R.L., 1996, U-Pb systematics of stilbite-bearing low-temperature mineral assemblages from the
Malmberget iron ore, northern Sweden: Geochimica et Cosmochimica Acta, v. 60, p. 1951-1961.
Rumble, D., 1973, Fe-Ti Oxide Minerals from Regionally Metamorphosed Quartzites of Western New
Hampshire: Contributions to Mineralogy and Petrology, v. 42, p. 181-195.
Sauerzapf, U., Lattard, D., Burchard, M. and Engelmann, R., 2008, The TitanomagnetiteIlmenite Equilibrium: New Experimental Data and Thermo-oxybarometric Application to the Crystallization of Basic to
Intermediate Rocks: Journal of Petrology, v. 49, p. 1161-1185.
Schuiling, R.D. and Feenstra, A., 1980, Geochemical behaviour of vanadium in iron-titanium oxides: Chemical Geology, v. 30, p. 143-150.
Sillitoe, R.H. and Burrows, D.R., 2002, New eld evidence bearing on the origin of the El Laco magnetite
deposit, northern Chile: Economic Geology, v. 97, p. 1101-1109.
Singoyi, B., Danyushevsky, L., Davidson, GJ., Large, R., Zaw, K., 2006, Determination of trace elements in
magnetites from hydrothermal deposits using the LA ICP-MS technique: SEG Keystone Conference,
Denver, US, Proceedings, CD-ROM
Vernon, R.H., 2004, A practical guide to Rock Microstructure: Cambridge, UK, Cambridge University
Press.

24

III

Practical way to quantify minerals from chemical assays at

Malmberget iron ore operations An important tool for
the geometallurgical program
Cecilia Lunda, Pertti Lambergb, Therese Lindbergc
a Division of Geosciences and Environmental Engineering, Lule University of Technology,
971 87 Lule, Sweden
b Division of Sustainable Process Engineering, Lule University of Technology,
971 87 Lule, Sweden
c LKAB, Research & Development, 983 81 Malmberget, Sweden

Minerals Engineering 49 (2013) 716

Contents lists available at SciVerse ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Practical way to quantify minerals from chemical assays at Malmberget

iron ore operations An important tool for the geometallurgical program
Cecilia Lund a,, Pertti Lamberg b, Therese Lindberg c
a
b
c

Division of Geosciences and Environmental Engineering, Lule University of Technology, 971 87 Lule, Sweden
Division of Sustainable Process Engineering, Lule University of Technology, 971 87 Lule, Sweden
LKAB, Research & Development, 983 81 Malmberget, Sweden

a r t i c l e

i n f o

Article history:
Received 30 November 2012
Accepted 9 April 2013

Keywords:
Geometallurgy
Iron ore
QEMSCAN
Modal mineralogy
Geological model

a b s t r a c t
This is the rst step in establishing a geometallurgical program for the Malmberget iron ore deposit,
northern Sweden. Geometallurgy captures geological and metallurgical (processing) information into a
spatially-based predictive model of mineral processing characteristics. This paper describes the development of a practical, fast and inexpensive technique to quantify minerals from routine chemical assays.
Ore samples and process samples from two different orebodies were used in the process of developing
this element to mineral conversion technique that involved electron microprobe (EPMA), X-ray uorescence (XRF) and SATMAGAN analyses. The method was validated against QEMSCAN analyses. From the
calculated modal mineralogy an ore classication system was established based on the iron mineralogy,
iron mineral grades and gangue mineralogy to create a preliminary geological/geometallurgical model of
the ore. However, in a geometallurgical context the modal composition is not sufcient and the geological
model requires information on mineral textures, too.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Geometallurgy combines geological and metallurgical information to create a spatially-based predictive model for a mineral processing plant to be used in production management (Lamberg,
2011). Industrial application is called a geometallurgical program
and its development normally goes through the following steps
(Bulled and McInnes, 2005; David, 2007): (1) Collection of geological data. (2) An ore sampling program for metallurgical testing. (3)
Laboratory testing of these samples. (4) Developing new ore-type
denitions called geometallurgical domains. (5) Developing mathematical relationships for the estimation of important metallurgical parameters across the geological database. (6) Developing a
metallurgical model of the process. (7) Plant simulation using the
metallurgical process model and the distributed metallurgical
parameters as the data set. (8) Calibration of the models via benchmarking for existing operations.
There exist basically two different approaches for linking the
steps listed above. The rst one relies on geometallurgical testing
and the other approach is based on mineralogy (Lamberg, 2011).
The mineralogy approach requires proper quantitative mineral
characterisation of the ore, but also dening the process model
based on minerals.
Corresponding author. Tel.: +46 920 492354; fax: +46 920 97 364.
E-mail address: [email protected] (C. Lund).
0892-6875/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.mineng.2013.04.005

This paper focuses on developing a practical technique to calculate the modal mineralogy by using chemical analyses of iron ores.
This is the rst step in establishing a geometallurgical model in the
Malmberget deposit, a particle-based approach consisting of three
sub-models, proposed by Lamberg (2011) (Fig. 1). The geological
model is dened by the modal mineralogy and mineral textures
which include mineral associations and grain sizes. In our case a
quick and inexpensive modal analysis method is a necessity considering the need of producing modal composition in a big amount
(>10 000) of ore samples.
Modal mineralogy, i.e. mineral grades, has traditionally been
determined by optical microscopy with point counting. It is time
consuming (Petruk, 2000) and the quality is dependent on the skill
of the mineralogist (Henley, 1992). Image analysis with optical
microscopy makes the method faster but positive identication
of minerals is not always easy especially if a process sample comprised of very ne grained particles. Recent advances have improved the quality of quantitative of X-ray diffraction (XRD) but
still the technique (Rietveld) is sensitive to preferred orientation,
crystal size, crystallinity and variation in the chemical composition
of the several phases present in the mixture (Rietveld, 1969). Automated mineralogy, i.e. scanning electron microscopy based image
analysis (QEMSCAN, MLA, PTA, IncaMineral) provides more rapid
quantitative analysis of mineral grades and textures (Jones and
Gravilovic, 1970; Sutherland and Gottlieb, 1991; Gu, 2003; Moen,
2006; Liipo et al., 2012). Although this technique has not only

C. Lund et al. / Minerals Engineering 49 (2013) 716

Fig. 1. The particle-based geometallurgical concept, modied from Lamberg (2011). Modal mineralogy and textures links the geological model and the process model. In the
process model minerals are treated as particles. From the geological information, the particle population is generated through the particle breakage model.

largely replaced optical microscopy and quantitative XRD and also

greatly expanded the use of quantitative mineralogy, it still has
some drawbacks which limit its usage. Firstly, it is basically a standardless analysis method. Secondly, for accurate analysis from particulate materials, the sample should have a narrow size
distribution to avoid classication in sample preparation and to
eliminate the need for changing magnication during image collection (Kwitko-Ribeiro, 2011). This means that samples need to be
sized by sieving which is time consuming and also will multiply
the number of samples to be analysed. Thirdly, there are some limitations on the particle size. For sound statistics one needs to analyse thousands of particles and, in coarse size fractions, multiple
resin mounts will be required to contain the necessary numbers
of such relatively large particles. In very ne particle and mineral
grain sizes (<5 lm) the resolution starts to be a problem in mineral
identication as the electron beam diameter approaches the size of
individual grains. Finally, minerals with very similar chemical
composition are difcult to identify in SEM based image analysis.
Magnetite and hematite are an example of such a mineral pair
(Fandrich et al., 2007; Gomes et al., 2012).
The use of multicomponent chemical mass balancing in solving
mineral quantities from chemical analysis (e.g. XRF) has been successfully used for decades by a number of researchers and laboratories (Braun, 1986; Paktunc, 1998; Whiten, 2008; Lamberg et al.,
1997; Lamberg, 2011). Normative calculations used by geologists
for classication purposes, for example CIPW (Cross, Iddings, Pirsson and Washington) (Cross et al., 1902), differ from mass balancing method by calculating hypothetical and not actual mineral
composition for samples, using stoichiometric compositions for
minerals (e.g. CIPW uses volatile free minerals) and reporting mineral grades as normative per cent, not as weight proportions.
Therefore, for quantitative geometallurgical purposes they are
not suitable. Instead, the mass balancing method, described here
as element-to-mineral conversion, is a mathematical mass balancing where the bulk elemental composition of a sample is transformed to mineral quantities using information on the chemical
composition of the actual observed minerals and the mineral quantity are reported as weight percentages enabling study of even the
trace elemental distribution between the minerals. Even if the
element-to-mineral conversion technique has been known for
decades it is an underused technique and according to the literature it is systematically applied in mining operations very rarely
(e.g. Kemi chromite mine, Lamberg, 2011).
The purpose of this study is to develop a practical modal analysis technique for the iron ore and process samples of the Malmberget deposit and processing plant. The developed element to
mineral conversion technique uses, in addition to the chemical
assays, the mineral diagnostic method, i.e. Satmagan analysis.
Three data sets are used. The rst and the second data sets are used
for developing and validating the technique. Reliability is estimated with Monte Carlo simulation and the developed method is

validated against the mineralogical information analysed by QEMSCAN. The third, and the largest data set, has been used to show
how modal mineralogy can be used in developing a mineralogical
classication system for a geometallurgical program.

2. Malmberget iron operations

The Kiirunavaara and Malmberget iron ores mined by LKAB
contribute about 90% of the iron ore production in Europe. The annual production in Malmberget is about 14 Mt of ore at 43.1% Fe
and in Kiirunavaara 22.7 Mt at 42.8% Fe (LKAB, 2011). In the
Malmberget ore eld, more than 20 different tabular to stock
shaped ore bodies of both magnetite and hematite are spread over
an area of 2.5  5 km2 situated in a Svecofennian succession of volcanic rocks (Martinsson and Virkkunen, 2004).
The two ore bodies of Malmberget considered in this study, Fabian and Printzskld are irregularly shaped magnetite ore lenses.
Fabian is larger and more compact in shape. The ore contacts shift
from sharp to graded, generating in the latter case a low grade Feore type, traditionally called ore breccia in this deposit (Bergman
et al., 2001). Besides existing in the rims of the ore body, the ore
breccia can also be found as inclusions in the massive ore. A common skarn breccia, with distinct modal mineralogy (i.e. amphibole/
diopside present), occurs in the host rock near the ore contact. The
Printzskld ore body shows similar features but the ore body is
longer and narrower and has a banded appearance. A strong alteration of the host rock has contributed to a larger amount of biotite
at the footwall and in the ore. Small bodies of hematite are present
in the footwall and hematite occurs also as an accessory mineral in
the ore. The main gangue minerals in the both the breccia and massive ore are albite, K-feldspar, apatite, quartz, amphiboles (actinolitetremolite), clinopyroxene, biotite, sulphides (pyrite and
chalcopyrite) and anhydrite. During mining both the ore breccia
and the massive ore are being mined together.
The Malmberget concentrator has two different lines, one for
magnetite dominated ore and the other for hematite dominated
ore. The magnetite line in Malmberget consists of dry and wet concentration sections (Fig. 2). The dry process stage uses screens,
cone crushers, autogeneous crushers (VSI) and low-intensity magnetic drum separators producing two different pre-concentrates: a
Sinter Feed (FAR) with higher iron grade and a slightly lower grade
Pellet Feed (PAR), both approximately <10 mm. In the wet concentration plant the Sinter Feed is processed further by screening to
<2 mm followed by wet low intensity magnetic separation
(WLIMS). The Pellet Feed is further upgraded in three steps of ball
mill grinding and WLIMS. In the nal grinding step the Pellet Feed
is ground to approx. 68% < 45 lm to reach the optimum size distribution for downstream processing (Kvarnstrm and Oghazi, 2008).
The nal concentrates have high requirements for purity, e.g.
the Sinter Feed (MAF) has maximum limits of 0.8% SiO2, 0.45%

C. Lund et al. / Minerals Engineering 49 (2013) 716

FAR

ROM

Sinter Fines

PAR

To Pellet Pla
n

Dry process
To Tailing Area

Wet process
To Tailings Pond

Fig. 2. Simplied ow sheet of Malmberget concentrator showing separate lines for magnetite and hematite dominated ores, modied from berg and Plsson (2008).

TiO2 and 0.025% P (LKAB, 2011). If requirements are not met, the
undesired impurities from gangue minerals (Si, P) or elevated trace
element contents in the magnetite (Ti) cause problems in downstream processing e.g. pelletization process or in the blast furnace
(Adolfsson and Fredriksson, 2011).
Currently Malmberget does not have an integrated geometallurgical program to be used in production. Based on iron mineralogy (magnetite versus hematite) ore bodies are mined selectively
and fed to two different lines. It is known that the ore bodies vary
in their iron mineral chemistry and, at times, for the production of
some special products (e.g. low/high vanadium or titanium), mining is done selectively from certain places. Otherwise, the modal
mineralogy and variation in the grain size of magnetite is not used
in production planning.
3. Samples and applied methods
The samples for the rst data set were collected from ve drill
cores representing two ore bodies, Fabian and Printzskld. Over
100 mineralogical samples were selected for characterising the
mineralogy, mineral chemistry and the texture of the ore. Polished
thin sections were prepared for the analyses of mineral chemistry
by electron microprobe (EPMA), at the Camborne School of Mines,
University of Exeter, Cornwall Campus, UK, using a JEOL JXA-8200
superprobe with a technique described by Potts et al. (1995). An
accelerating voltage of 15 kV and a beam current of 30 nA were
used for the oxides. The probe beam diameter was set to 1 lm
for the oxides and the phosphates and 5 lm for the silicates.
The second dataset includes metallurgical samples that were
collected from several drill cores. The drill cores were carefully
logged and three dominant ore types were identied and accordingly three composite samples were created (Lund et al., 2010).
The mineral processing tests were all done at the LKABs mineral
technical laboratory, Malmberget. The ore types were crushed to
3 mm so that in situ textures and mineral grain sizes were preserved. Dry magnetic separation tests were done with Sala Mrtsell
magnetic separation using a standard test procedure by LKAB
(Lund et al., 2010; Mattsby, 2010). Concentrate and tail from the
tests were weighed and sized into ten size fractions using sieves
of 38, 75, 150, 300, 425, 600, 1190, 1680 and 3000 lm. The chemical analyses were all performed at LKABs chemical laboratory at
Malmberget, using their standardised methods for X-ray uorescence (XRF), Titrimetric method (ISO 9035) and SAturation MAgnetic Analyser model 135. Bulk samples and size fractions were
analysed with XRF for Fe, Al, Ti, Si, K, Ca, Mg, Mn, V, Na, S and P.

The amount of divalent iron was analysed with wet chemical titration and the amount of magnetic material was determined with a
SATMAGAN magnetic balance (Stradling, 1991). Polished resin
mounts were prepared from size fractions for optical microscopy
and QEMSCAN analysis.
The quantitative modal analyses were performed at the LKAB,
Lule with QEMSCAN E430 Pro for sized samples. The system is
based upon a Carl Zeiss scanning electron microscope with four energy dispersive X-ray detectors (EDS). The measurement mode
used was PMA (particle mineral analysis). To obtain quantitative
analysis of the material, a Species Identication Protocol (SIP-le)
was developed to be able to discriminate the different minerals
at this particular deposit (Gottlieb et al., 2000; Pirrie et al., 2004).
In the analyses a total of 16 minerals were identied and quantied (Table 1). From each sample about 5000 particles were analysed and modal mineralogy was reported. In this study this data
set was used to verify the element to mineral conversion against
the QEMSCAN modal mineralogy.
The third data set includes only chemical analyses from eleven
drill cores. Modal mineralogy of the samples was calculated using
the methodology developed for the previous data sets. The third
data set was used to demonstrate how the method can be used
in a geometallurgical context.
4. Element to mineral conversion
4.1. Theory
Element to mineral conversion is based on a set of linear algebraic equations where the bulk chemical composition of a sample
is converted to mineral grades using a set of least-squares equations (Paktunc, 1998; Whiten, 2008; Lamberg et al., 1997). This
can be expressed as an equation

Axb

where A is a table giving the weight fractions of elements in the

minerals (e.g. electron microprobe analyses, known), the x vector
is the weight fractions of the minerals in a sample (unknown) and
the b vector is the weight fraction of elements in the sample
(known). For a simple magnetite albite binary with Fe and Si analyses the equation would be (chemical composition of minerals from
Table 2):


 
 

 0:73 0:00   xMgt   0:60 




 0:00 0:33   xAb   0:06 

10

C. Lund et al. / Minerals Engineering 49 (2013) 716

Table 1
Typical (average) modal composition of Malmberget; ore, concentrate and tailing from a Mrtsell dry magnetic separation laboratory test.
Mineral

Magnetite (Mgt)
Ilmenite (Ilm)
Rutile (Rt)
Quartz (Qtz)
Calcite (Cal)
Albite (Ab)
Orthoclase (Or)
Biotite (Bt)
Diopside (Di)
Actinolite (Act)
Chlorite (Chl)
Titanite (Tit)
Zircon (Zr)
Apatite (Ap)
Sulphates (anhydrite, gypsum) (Py)
Sulphides (pyrite, chalcopyrite) (Py)
a

Orea

Formula

Fe3O4
FeTiO3
TiO2
SiO2
CaCO3
NaAlSi3O8
KAlSi3O8
K(Mg,Fe)3(AlSi3O10)(OH)2
CaMgSi2O6
Ca2(Mg,Fe)5Si8O22(OH)2
(Mg,Fe)5Al(Si3Al)O10(OH)8
CaTiSiO5
ZrSiO4
Ca5(PO4)3(F,Cl,OH)
CaSO4, CaSO42H2O
FeS2, CuFeS2

Concentrate

Tail

Avg. (wt.%)

STD

Avg. (wt.%)

STD

Avg. (wt.%)

STD

54.00
0.89
0.39
1.14
0.09
19.07
4.77
4.49
1.89
7.04
0.46
0.27
0.06
3.97
0.25
0.26

1.04
0.13
0.09
0.15
0.04
0.61
0.29
0.25
0.19
0.37
0.09
0.07
0.03
0.24
0.06
0.07

90.90
0.88
0.44
0.33
0.01
2.83
0.96
0.66
0.29
1.03
0.27
0.12
0.00
0.28
0.02
0.07

1.35
0.13
0.09
0.08
0.01
0.22
0.12
0.10
0.07
0.13
0.07
0.05
0.01
0.07
0.02
0.03

2.87
0.99
0.33
2.22
0.23
42.54
9.54
9.11
4.28
15.72
0.75
0.44
0.12
8.72
0.59
0.56

0.23
0.13
0.08
0.20
0.06
0.90
0.42
0.36
0.29
0.56
0.12
0.09
0.04
0.36
0.10
0.09

Ore sample back calculated from concentrate and tailing.

Table 2
The average chemical composition (wt.%) of Malmberget minerals (Fabian ore) analysed by electron microprobe in 25 samples, giving the variation in the ore body. For the Amatrix the oxides converted to elements. Mineral abbreviations as in Table 1.

# analyses
SiO2
TiO2
Al2O3
Cr2O3
V2O3
FeO
MnO
MgO
CaO
Na2O
K2O
P2O5
NiO
F
Cl
Sum
Fe2O3 (calc)
FeO (calc)
Recalc. sum

Mgt

Ilm

Qtz

Ab

Bt

Di

Act

Ap

Hem

590

40

7
100.4

60
72.7

43
55.4
0.03
1.88

14

49.8

36
54.7
0.02
0.23

59

0.08
0.10
0.02
0.08
93.7
0.01
0.01

12
41.92
4.00
13.06

13.85
0.05
17.01

9.61
0.14
13.0
21.8
1.51

10.3
0.05
17.6
11.5
1.06
0.13

0.21
0.01
0.12
56.5

20.36
0.29
44.6
2.90
1.90
0.01

0.01
0.19
0.04
0.01
94.2
69.4
31.2
101.2

0.06

0.01
0.01

0.01

0.01
6.89

0.01

0.03
7.90

0.35
0.05
0.12
90.2

36.8
0.01
1.36
99.5
6.87
38.5
100.2

100.5

99.9

99.2

0.18
101.0

98.2

6.58
0.28
100.4

90.7
99.9
0.31
100.8

Pyrite and orthoclase stoichiometric.

There are two constraints for the solution. Mineral grades (x)
cannot be negative and the sum of minerals must to be smaller
than or equal to 100%. The solution can be found by searching
for the minimum value for the residual vector (R).

R jb  A  xj

If there are more elements than minerals the case is over-determined and there is more than one solution. If the number of minerals is higher than elements the case is under-determined and there
is no unique solution (Paktunc, 1998; Whiten, 2008). The multiple
solutions can be solved e.g. by Gaussian elimination and multiple
regression techniques to minimize the sum of square of the
deviations.
As minerals in nature are seldom stoichiometric the actual elemental composition of minerals analysed by electron microprobe
should be used (A matrix). Quite often simple bulk chemical assays
of a sample do not provide enough information for estimating
mineral grades. For example in porphyry copper deposits there
may be several copper sulphide minerals present in the sample
and simple Cu, Fe and S assays do not enable all of them to be
quantied. For metallic mineral deposits a number of selective

analysis techniques have been developed. Most of them are traditional wet chemical methods. For gold they have been known for
decades (Lorenzen and Van Deventer, 1993). Also for copper they
are widely used (Lamberg et al., 1997), but they are quite practical
also for nickel (Penttinen et al., 1977), molybdenum and zinc
(Young, 1974). In this study two different techniques were used
to independently determine the magnetite and hematite grades:
titrimetric divalent iron analysis and SATMAGAN magnetic balance. For each minerals paragenesis one should evaluate whether
diagnostic methods are available and could be used since they will
improve the accuracy of the element to mineral conversion.
Here the element to mineral conversion was done with HSC
Chemistry software version 7.1 (Lamberg et al., 1997). The software includes a reference table in its mineral database for the dissolution of minerals in several selective analysis techniques listed
above. In solving Eqs. (1) and (2) HSC includes several mathematical options and here the Least Square and the Non-negative Least
Squares solutions were used (Lawson and Hanson, 1974). For further information on mathematical methods the reader is referred
to Braun (1986), Paktunc (1998) and Whiten (2008) and the HSC
manual (Lamberg and Tommiska, 2009).

11

C. Lund et al. / Minerals Engineering 49 (2013) 716

4.2. Method for the Malmberget iron ore

The optical microscopy revealed that in the metallurgical samples (the second data set) magnetite is the only iron ore mineral,
i.e. no hematite was observed. The main gangue minerals are plagioclase (albite), K-feldspar (orthoclase) and amphiboles (actinolitetremolite) together with apatite, biotite, chlorite, diopside,
and quartz. A small amount of anhydrite, calcite, chalcopyrite,
ilmenite, pyrite, rutile, scapolite, titanite and occasionally zircon
was found. The main difference between the two ore bodies is that
biotite and apatite grades are higher in the Printzskld ore. The
average chemical composition of the minerals analysed in 25 samples from Fabian used to generate the A-matrix is given in Table 2.
Twelve elements are routinely analysed by XRF at the LKAB
chemical laboratory. In addition, two diagnostic analysis methods,
titrimetric method (Fe2+) and Satmagan (Fe2+) are used in the
Malmberget laboratories when it is anticipated that both magnetite and hematite are present in the ore or process sample, giving
one additional element. The titrimetric method gives the weight
percentage of divalent iron which in Malmberget ores is found in
magnetite (24%), and in mac minerals (amphibole, biotite). The
Satmagan is a magnetic balance which basically gives the mass
proportion of magnetic material. At LKAB Satmagan has been calibrated against the Fe2+ assay, i.e. if the sample is 100% magnetite
both Fe2+ content and the Satmagan should report 24%.
Two of the analysed elements by XRF, namely V and Mn, occur
only as minor constituents in minerals, and are not a very good
choice for mineral quantication and are therefore not used in
our calculation. All together it gives in the end a total of eleven
independent assays enabling a maximum number of eleven minerals to be quantied. Due to the lack of Zr and C analyses the mass
proportion of neither zircon nor calcite can be estimated. They are
trace minerals, dont have any big relevance for the concentration
process and therefore this simplication is justied. To further reduce the number of minerals, titanium minerals (ilmenite, rutile
and titanite) were combined and the most voluminous one, ilmenite represents all the titanium minerals. The same was done for sulphur bearing minerals: pyrite represents sulphides (pyrite,
chalcopyrite) and sulphates (anhydrite, scapolite). Finally as chlorite is a minor mineral occurring as an alteration product of biotite,
it was dropped out as well and is reported together with biotite.
Having eleven elements and eleven minerals, it would be possible to calculate all the minerals simultaneously. However, with
this way, the Least Squares solution (LS) for some samples gave a
negative value for some of the minerals. When using a Non-Negative Least Squares solution (NNLS without weighting) the algorithm put very little notice on some of the trace elements and
the corresponding minerals, resulting in zero grades e.g. for pyrite
and ilmenite and even the residual for titanium and sulphur was
positive (Table 3). Therefore, the best solution found was to

divide the calculation into several rounds. After testing different

alternatives a plausible mineral conversion solution for the two
ore bodies was found by using a total of three calculation rounds
and both LS- and NNLS-algorithms (Table 3). In the rst round the
minerals which carry solely one element are solved with the LS
technique. This means that the apatite and pyrite grades are
solved using phosphorous and sulphur. The second round takes
silicate minerals: albite, orthoclase, biotite, diopside, actinolite
and quartz, using residual Na, Mg, Al, Si, K and Ca; and the NNLS
solution is used. Since ilmenite and magnetite contain most of Ti
Fe and Fe, the third round uses the residual Ti and Fe (XRF). Diagnostic assay methods, SATMAGAN and Fe2+ give a possibility to
calculate both hematite and magnetite, if present. For this alternative also hematite and SATMAGAN or Fe2+ are included in the
third calculation round.
Mathematically the method seems to give a good result as the
calculation gives low nal residuals and the sum of the minerals
is close to 100%. However, to have some reliable estimate on the
precision an error analysis was carried out using Monte Carlo simulation. The standard deviation of the chemical analyses was received from the LKAB chemical laboratory and for the minerals it
was calculated from the electron microprobe analyses. For the
XRF analysis the relative standard deviation (RSD) of iron is very
low, 0.1% and for the other main elements the RSD is between 1%
and 2%. For magnetite the relative standard deviation of the iron
assay by electron microprobe within the ore body (i.e. in the sample set) is close to 5%, i.e. 72.9 3.6. For silicates the relative standard deviation of the main elements is quite high, about 10% (RSD).
Table 4 shows the standard deviation for the mineral grades resulting from the Monte Carlo simulation. Although the standard deviations of the mineral grades are quite high, one must remember
that there is still a good chemical mass balance since mineral
grades from back calculations give a similar chemical composition
for the samples as the chemical assays.
Hematite is reported to be zero as there is no detected content
in the samples (Table 5). Both the ores have individual adjustments
of the rounds of calculations. In Printzskld ore the calculation
gave that the diopside content is higher than that of actinolite,
which according to optical microscopy is not true. Therefore diopside was excluded and the second round used ve minerals and six
elements (over-determined). In calculating magnetite from Fe2+ or
SATMAGAN in some samples little iron was left over for allocation
to hematite, showing the uncertainty of the calculation.
4.3. Comparison of element to mineral conversion with QEMSCAN
analyses
The accuracy and reliability of mineral quantication was estimated by comparing the modal mineralogy by the element to mineral conversion method with the QEMSCAN analyses for 20 process

Table 3
Calculation routines used in modal calculation. Two alternative calculation routines; left: all minerals calculated in one step, right: calculation divided into three rounds. The
latter that gave a better result was used (see text). The elements showed as oxides.
Round

Mineral

Components

Methods

Round

Mineral

Components

Methods

Apatite (Ap)
Pyrite (Py)

P2O5
S

NNLS

Apatite (Ap)
Pyrite (Py)

P2O5
S

LS

Quartz (Qtz)
Albite (Ab)
Orthoclase (Or)
Biotite (Bt)
Diopside (Di)
Actinolite (Act)

SiO2
Na2O
MgO
K2O
Al2O3
CaO

Quartz (Qtz)
Albite (Ab)
Orthoclase (Or)
Biotite (Bt)
Diopside (Di)
Actinolite (Act)

SiO2
Na2O
MgO
K2O
Al2O3
CaO

NNLS

Magnetite (Mgt)
Ilmenite (Ilm)
Hematite (Hem)

Satmagan
TiO2
FeO(tot)

Magnetite (Mgt)
Ilmenite (Ilm)
Hematite (Hem)

Satmagan
TiO2
FeO(tot)

NNLS

12

C. Lund et al. / Minerals Engineering 49 (2013) 716

Table 4
Modal composition of ore, concentrate and tailing of Mrtsell dry magnetic separation laboratory test calculated according to element to mineral conversion.
Orea

Mineral

Magnetite (Mgt)
Ilmenite (Ilm)
Quartz (Qtz)
Albite (Ab)
Orthoclase (Or)
Biotite (Bt)
Diopside (Di)
Actinolite (Act)
Apatite (Ap)
Sulphides (pyrite, chalcopyrite) (Py)
Hematite (Hem)
a

Concentrate

Tail

Avg. (wt.%)

STD

Avg. (wt.%)

STD

Avg. (wt.%)

STD

57.96
1.29

19.42
5.46
4.55
3.18
3.03
3.54
0.21

1.07
0.16

0.62
0.30
0.28

0.23
0.06

87.00
1.01

5.34
1.67
1.09
1.17
0.85
0.34
0.06

1.32
0.14

0.31
0.17
0.15

0.08
0.03

6.97
1.62

47.30
10.78
9.71
8.03
5.89
7.86
0.51

0.37
0.18

0.95
0.44
0.41

0.34
0.10

Ore sample back calculated from concentrate and tailing. Standard deviation from Monte Carlo simulation, compare with Table 1.

Table 5
Calculated modal composition (wt.%) of the ore and Mrtsell dry magnetic separation laboratory test products calculated with element to mineral conversion. Mineral
abbreviations as in Table 1.
Ore

Fraction

Stream

Mgt (%)

Ilm (%)

Qtz (%)

Ab (%)

Or (%)

Bt (%)

Di (%)

Act (%)

Ap (%)

Py (%)

Hem (%)

Total

SG

Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Pz
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa
Fa

3875 lm
3875 lm
3875 lm
75150 lm
75150 lm
75150 lm
150300 lm
150300 lm
150300 lm
300425 lm
300425 lm
300425 lm
425600 lm
425600 lm
425600 lm
3875 lm
3875 lm
3875 lm
75150 lm
75150 lm
75150 lm
150300 lm
150300 lm
150300 lm
300425 lm
300425 lm
300425 lm
425600 lm
425600 um
425600 lm

Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail
Ore
Concentrate
Tail

47.29
92.10
5.87
48.47
90.84
5.68
51.26
86.93
7.87
54.71
85.40
8.42
60.11
82.81
9.86
60.26
92.34
4.52
64.92
90.95
4.81
65.66
88.00
5.61
62.44
82.89
7.84
64.51
77.77
9.19

1.90
1.02
2.73
1.50
0.92
2.08
1.42
0.94
2.00
1.50
1.06
2.16
1.43
1.18
1.98
1.17
0.92
1.59
0.85
0.80
0.97
0.82
0.84
0.75
1.05
1.10
0.91
1.22
1.28
0.97

18.70
1.56
34.55
18.67
2.86
34.62
15.98
4.46
29.98
12.41
4.35
24.54
11.41
5.07
25.39
24.79
2.95
62.72
22.65
4.61
64.29
23.14
6.26
68.34
24.72
9.34
65.66
21.76
11.91
62.89

11.51
1.59
20.70
10.06
1.66
18.57
8.71
2.85
15.81
8.03
3.39
15.06
8.17
4.07
17.28
2.21
0.35
5.45
1.26
0.18
3.78
1.11
0.63
2.31
1.52
0.87
3.21
2.00
1.12
5.68

4.65
0.79
8.18
7.60
1.05
14.19
9.07
1.19
18.67
8.78
1.43
19.90
6.35
1.54
16.91
1.74
0.79
3.40
1.96
0.78
4.68
1.67
0.76
4.17
1.87
1.12
3.90
1.80
1.48
3.14

7.67
1.25
18.82
6.47
1.29
18.41
5.45
1.84
14.99
6.14
3.08
14.25
6.09
4.23
13.83

6.81
1.08
12.13
5.07
0.88
9.31
5.63
1.47
10.71
6.85
2.16
13.86
5.96
2.87
12.93

7.70
0.60
14.24
7.48
0.52
14.51
6.67
0.50
14.16
6.45
0.55
15.37
5.09
0.73
14.70
0.52
0.07
1.30
0.45
0.06
1.35
0.43
0.09
1.35
0.36
0.12
0.98
0.26
0.18
0.62

0.22
0.04
0.38
0.17
0.04
0.29
0.15
0.05
0.27
0.17
0.05
0.33
0.15
0.06
0.34
0.34
0.08
0.79
0.22
0.06
0.59
0.19
0.05
0.56
0.23
0.07
0.66
0.24
0.09
0.84

98.79
98.77
98.78
99.01
98.77
99.24
98.87
98.38
99.47
98.89
98.40
99.65
98.67
98.34
99.39
98.69
98.75
98.59
98.78
98.73
98.89
98.46
98.58
98.07
98.33
98.64
97.43
97.88
98.06
97.15

3.60
4.91
2.89
3.63
4.85
2.90
3.71
4.70
2.96
3.81
4.65
3.00
3.92
4.57
3.00
3.90
4.93
2.86
4.01
4.86
2.86
4.02
4.75
2.84
3.94
4.57
2.87
4.02
4.43
2.89

samples. The accuracy requirement in the geometallurgical context

is different from process mineralogy. In a process mineralogical
study, the aim typically would be to identify differences in mineral
recovery of 0.5% absolute, corresponding to about 1 wt.% accuracy
(1r) in mineral grades. However, to establish geometallurgical domains and identify signicant geometallurgical ore types, an accuracy of 5 wt.% is adequate (1r) (Pitard, 1989).
The mineral grades calculated by element to mineral conversion
versus QEMSCAN are compared in parity plots (Fig. 3). To test the
statistical agreement (signicance) of the two methods, a multivariate analysis of variance (MANOVA) was used due to the fact that
the minerals (variables) are correlated. It uses the vectors of the
mean of the included minerals in the two data sets of the correlated variables involved. The MANOVA method initially tests if
there is a signicant difference between the two mean vectors. If
signicance is detected, an analysis of variance (ANOVA) for each
mineral pair will identify the mineral/group of minerals that

causes the signicant difference between the methods (Johnson

and Wichern, 1998). Simultaneous condence intervals (Bonferroni) are calculated in the single variable ANOVA for each mineral.
The Bonferroni correction is used in order to keep the p-value correct due to the correlation between the variables/minerals. The
comparison shows that the element to mineral conversion can be
used in a geometallurgical context to classify the ore types (p-value
0.10) but in process mineralogy some more development is needed
to improve the accuracy (precision). This is true especially for apatite and amphibole for which the two methods show a signicant
difference.

5. Geometallurgical ore type classication

To develop a system for ore type classication the third data
set was collected from eleven drill cores of both Printzskld and

13

R square=0.979

80

80

60

60

40

40

20

20

0
0

20

40

60

80

0
100

70

R square=0.955

60

50

40

40

30

30

20

20

10

10

0
0

10

14
R square=0.858

12

10

10

Element to Mineral conversion

Element to Mineral conversion

12

20

30

40

50

60

70

20

20
Orthoclase
R square=0.904

15

15

10

10

0
0

QEMSCAN
16

Apatite

14

60

50

QEMSCAN
16

70

Albite

Element to Mineral conversion

100

Magnetite

25

25
Biotite

20

R square=0.917

20

15

15

10

10

0 2 4 6 8 10 12 14 16 18 20 22 24 26

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28

QEMSCAN

QEMSCAN

10 12 14 16 18

QEMSCAN
Element to Mineral conversion

100

Element to Mineral conversion

Element to Mineral conversion

C. Lund et al. / Minerals Engineering 49 (2013) 716

20

20

Amphibole
R square=0.871

15

15

10

10

0
0

10

20

30

QEMSCAN

Fig. 3. A comparison of the modal composition (wt.%) of iron ore determined by QEMSCAN analyses (x-axis) and the element to mineral conversion (y-axis). Line shows the
regression with R squared values and samples as in Table 5.

Fabian. The routine chemical assays (XRF) were done for the samples collected in logging; supplementary Fe2+ analyses were only
acquired when hematite was identied mesoscopically during
logging. Over 300 XRF analyses were converted to mineral grades
and an ore type classication was developed based on modal
mineralogy. A preliminary geometallurgical ore type (GEM-type)
denition aims to capture all the important mineralogical features related to ore processing (see Section 2): (1) Fe mineral
grade, (2) Fe-mineralogy, (3) gangue mineralogy and (4) the presence of sulphides.
As the iron grade is overall quite high, normalised gangue mineral grades were used as a basis to distinguish and characterise the
different GEM-types with special reference to the silicate mineralogy (cf. silicate problem). When developing the system for geometallurgical purposes it is important to keep the classication simple
and at the same time connect the metallurgical meaning to all the
GEM-types. Originally seven different GEM-types were dened
shown in the ow sheet in Figs. 4 and 5. The preliminary types
were most probably too many and the classication was too detailed and needed revision. Firstly, a simplication was done by
the combination of K-feldspar and albite to feldspar since they
are similar in their metallurgical behaviour. Thereafter the main
gangue minerals of the ores are feldspar and amphibole.
The feldspar rich GEM- type (Fsp) is common in hematite and
low grade ores of magnetite (Fig. 6). The amphibole GEM-type is
quite widespread in the magnetite ore and is taken as a class of
its own because it is a harder rock type than the feldspar GEM-type
(Fig. 4). The amphibole GEM-type is further divided into ve subtypes. Iron concentrate has restrictions to sulphur and phosphorous, therefore it is important to identify the variation in terms
of sulphides and apatite, and therefore a pyrite GEM-type (Py bearing (Amph-(ApBt)) and an apatite bearing GEM-type (Ap-(Amph))
were established. Locally biotite rich zones can be found and they
cause some problems in rock stability. Also in comminution biotite
tends to enrich in ne size fraction and this is partially the reason

Fig. 4. A geometallurgical ore type (GEM-type) classication scheme of the gangue

mineralogy of Malmberget iron ore. For the classication the gangue mineral grades
are recalculated to 100%.

for elevated silica content in the concentrate. Consequently it is

important to identify areas rich in biotite and a subgroup was created, GEM-type (Bt-(Amph-Ap)).
Quartz is not a typical gangue mineral in the ore but exists in
the wall rock. The presence of quartz in the modal composition
indicates a presence of harder wall rock which may be important
to identify and therefore deserves a subclass of its own, GEM-type
(Qtz bearing Fsp + Amph). This developed classication scheme is a
preliminary one and needs to be revised in practice.

C. Lund et al. / Minerals Engineering 49 (2013) 716

Amph+Ap+Bt(22)

Qtz bearing Fsp+Amph(2)

Ap-(Amph)(62)

Bt-(Amph-Ap)(12)

Amph-(Ap-Bt)(71)

Py bearing Amph-(Ap-Bt)(36)

Fsp%

Fsp(74)

Qtz bearing Fsp+Amph(2)

Amph+Ap+Bt(22)

Bt-(Amph-Ap)(12)

Ap-(Amph)(62)

Amph-(Ap-Bt)(71)

Py bearing Amph-(Ap-Bt)(36)

Fsp(74)

Qtz bearing Fsp+Amph(2)

Amph+Ap+Bt(22)

Bt-(Amph-Ap)(12)

Ap-(Amph)(62)

Amph-(Ap-Bt)(71)

Py bearing Amph-(Ap-Bt)(36)

100
80
60
40
20
0

100
80
60
40
20
0

(h)
40
30
20
10
0
Qtz bearing Fsp+Amph(2)

Amph+Ap+Bt(22)

Bt-(Amph-Ap)(12)

Ap-(Amph)(62)

Amph-(Ap-Bt)(71)

Py bearing Amph-(Ap-Bt)(36)

40
30
20
10
0
Fsp(74)

Qtz bearing Fsp+Amph(2)

Amph+Ap+Bt(22)

Bt-(Amph-Ap)(12)

Ap-(Amph)(62)

Amph-(Ap-Bt)(71)

Py bearing Amph-(Ap-Bt)(36)

50
40
30
20
10
0
Fsp(74)

50
40
30
20
10
0

Qtz %

(g)

80
70
60
50
40
30
20
10
0

(f)
100
80
60
40
20
0
Fsp(74)

Amph %

100
80
60
40
20
0

Qtz bearing Fsp+Amph(2)

Amph+Ap+Bt(22)

0
Bt-(Amph-Ap)(12)

10

Ap-(Amph)(62)

10
Amph-(Ap-Bt)(71)

20

20

Py bearing Amph-(Ap-Bt)(36)

30

80
70
60
50
40
30
20
10
0

Amph+Ap+Bt(8)

Bt-(Amph-Ap)(3)

Ap-(Amph)(51)

Py bearing Amph-(Ap-Bt)(12)

Amph-(Ap-Bt)(14)

100
80
60
40
20
0
Fsp(34)

Amph+Ap+Bt(22)

Qtz bearing Fsp+Amph(2)

Ap-(Amph)(62)

Bt-(Amph-Ap)(12)

Amph-(Ap-Bt)(71)

Py bearing Amph-(Ap-Bt)(36)

(c)

100
80
60
40
20
0

(e)

30

Fsp(74)

Py %

(d)

Bt %

Hem %

(b)
100
80
60
40
20
0

100
80
60
40
20
0
Fsp(74)

Mgt %

(a)

Ap %

14

Fig. 5. Box and Whisker diagram illustrating the variation of mineral grades (wt.%) in different GEM-types of Malmberget, classied according to Fig 4.

The accuracy and reliability of this GEM-type classication system was estimated against optical microscopy. In 95% of the cases
both methods gave the same result.
Examinations of selected samples by optical microscopy show
that texture varies within the GEM-types. For example the feldspar-type includes both ne and coarse grained magnetite. This
indicates that modal mineralogy is not sufcient for comprehensive geometallurgical classication and additional information on
mineral textures is required.

6. Discussion
This study is the rst step in developing a geometallurgical
model for the Malmberget iron ore deposit. There are principally
two different approaches to create spatially-based predictive models for a geometallurgical program, geometallurgical testing and
mineralogy.
Geometallurgical tests are small scale laboratory tests which
aim to measure directly metallurgical response of the samples

but require also lot of efforts to get reliability. Examples of such

are GeM Comminution Index test (JKTech, 2010), JK Mineral Separability Indicator test (Bradshaw, 2010) and Davis tube test (Niiranen and Bhm, 2012). These tests are commonly done without the
mineralogical information and the metallurgical response is reported on an elemental basis, therefore the geometallurgical variations in the ore body and domains are based on elemental grades.
The mineralogical approach relies on a complete quantitative
mineral characterisation and the established process models based
on minerals can then be used in three different levels. The rst level takes mineral quantities and describes the behaviour of each
mineral in each process stage quantitatively (unsized by mineral).
The second level is sized by mineral; this level allows consideration
of variation in mineral size distribution, arising from variation in
particle size distribution. The third level uses the liberation information and each unit process model must be capable to use it.
Modal mineralogy (mineral grades) was shown to be the key
parameter in the geometallurgical classication in some weathered iron ores (Paine et al., 2011; Neumann and Avelar, 2012). Besides, the lithology, i.e. modal mineralogy, can be a direct link to

C. Lund et al. / Minerals Engineering 49 (2013) 716

15

Fig. 6. A graphic log showing the modal composition of two drill cores DDH-6533 from Fabian (left), DDH-6716 from Printzskld (right). Pictures below show the gangue
mineral grades normalised to 100%.

grindability, as shown by Casali et al. (2001). However, modal composition is seldom used as a basis of a geometallurgical model and
most probably is not sufcient since also the mineral textures
(grain size, associations) play a signicant role in the processing
of the ore.
The use of modal mineralogy as a systematic predictive tool in
industrial mineral processes is rather rare even if it is frequently
used in the owsheet development and problem solving (Lotter,
2007; Benedictus et al., 2008; Schouwstra et al., 2010; Bowell
et al., 2011). In the development of the process models based on
modal mineralogy the modal analyses have been made mainly by
automated mineralogy (Lotter, 2007) and X-ray diffraction (Rietveld) (Hestnes and Srensen, 2012). However, the element to mineral conversion method is a practical, fast and inexpensive
technique to calculate the modal mineralogy from the routine
chemical assays. It does not suffer from low detection limits, and
doesnt have limitation on particle size and the crystallinity of
phases. Additionally the chemical composition of the minerals is
taken from the ore, i.e. not using stoichiometric compositions
and the method is easy to automate to become a routine tool. In
complex ores additional selective analysis techniques can be used,
as the use of Satmagan here demonstrates.
Element to mineral conversion gives modal composition for any
ore types and one could question whether there is any need for
establishing any additional geometallurgical ore types or domains.
The established geological model based on modal mineralogy has
the potential to illustrate geometallurgical variation better than
the modal mineralogy alone. For example it differentiates hard
and soft ore, points out the location of weak ore (Ghosh, 2012),
and reveals the magnetite and hematite ratio. Furthermore it is a
method that can be applied in other Fe-ores of the area.
The purpose of the comminution stage is to liberate ore minerals and the operation of the comminution circuit is measured by
the degree of liberation. Generally, in Malmberget the degree of
liberation of a mineral (in a given size) is directly related to its
grain size which in turns commonly correlates with the mineral

grade. However, the degree of liberation is not enough to forecast

the behaviour of a mineral in the process since different types of
mixed particles have behaviour of their own. For example it has
been shown that in lead otation galenasphalerite binaries behave in a quite different way than galenaquartz binaries (Lamberg
and Vianna, 2007). Therefore, besides liberation the association of
the mineral is important and besides modal mineralogy at least
mineral textures should be included in the geometallurgical model.
Consequently the modal composition and preliminary modal classication is most likely not enough for an accurate geometallurgical model.
7. Conclusions
In this study an important and useful tool for a mineral based
geometallurgical program at the Malmberget iron ore deposit
was developed. Geometallurgy requires a practical, fast, inexpensive modal analysis method and this study shows that this can
be reached by the element to mineral conversion method. Even if
element to mineral conversion cannot provide the accuracy of
automated mineralogy, in a geometallurgical context the quality
is adequate for identifying geometallurgical ore types and domains
as well as to develop process models. The use of an additional
selective analysis technique (Satmagan) increases the number of
minerals that can be estimated and this is often needed in complex
ores. The established geometallurgical ore type (GEM-type) illustrates the mineralogical variation in an ore but does not fully explain the metallurgical response in a mineral process. Additional
information on mineral textures is needed. This topic will be addressed in a coming paper.
Acknowledgements
We thank HLRC (Hjalmar Lundbohm Research Centre) and
CAMM (Centre of Advanced Mining and Metallurgy) for nancial

16

C. Lund et al. / Minerals Engineering 49 (2013) 716

support in this project. Our appreciation goes also to Eva berg,

ke Sundvall, Lars Drugge and Kari Niiranen in LKAB who contributed with knowledge and comments to this study.
References
Adolfsson, G., Fredriksson, A., 2011. Reduction of silica in LKAB pellets through
different mineral processing unit operation. In: Conference in Mineral
Engineering, Proceedings, Lule, Sweden, 89 February, pp. 1726.
Benedictus, A., Berendsen, P., Hagni, A.M., 2008. Quantitative characterisation of
processed phlogopite ore from Silver City Dome District, Kansas, USA, by
automated mineralogy. Minerals Engineering 21, 10831093.
Bergman, S., Kbler, L., Martinsson, O., 2001. Description of regional geological and
geophysical maps of northern Norrbotten County (east of Caledonian orogen).
SGU Geological Survey of Sweden Ba 56, 110.
Bowell, R.J., Grogan, J., Hutton-Ashkenny, M., Brough, C., Penman, K., Sapsford, D.J.,
2011. Geometallurgy of uranium deposits. Minerals Engineering 24, 1305
1313.
Bradshaw, D.J., 2010. Development of a new tool for process mineralogy. In: Process
Mineralogy 10, Vineyard Hotel, Cape Town, South Africa, 1012 November.
Braun, G., 1986. Quantitative analysis of mineral mixtures using linear
programming. Clays and Clay minerals 34 (3), 330337.
Bulled, D., McInnes, C., 2005. Flotation plant design and production planning
through geometallurgical modeling Publication Series. Australasian Institute of
Mining and Metallurgy, v. Publication Series, pp. 809814.
Casali, A., Gonzalez, G., Vallebuona, G., Perez, C., Vargas, R., 2001. Grindability softsensored based on lithological composition and on-line measurements.
Minerals Engineering 14, 689700.
Cross, W., Iddings, J.P., Pirsson, L.V., Washington, H.S., 1902. A quantitative chemicomineralogical classication and nomenclature of igneous rocks. Journal of
Geology 6, 555690.
David, D., 2007. The Importance of Geometallurgical Analysis in Plant Study Design
and Operational Phases. Australasian Institute of Mining and Metallurgy
Publication Series, pp. 241247.
Fandrich, R., Gu, Y., Burrows, D., Moeller, K., 2007. Modern SEM-based mineral
liberation analysis. International Journal of Mineral Processing 84, 310320.
Ghosh, R., 2012. Investigation of Borehole Stability in Malmberget. Lule University
of Technology, Department of Civil, Environmental and Natural resources
engineering, 248p.
Gomes, O., Iglesias, J., Paciornik, S., Viera, M., 2012. Classication of hematite types
in iron ores through circularly polarized light microscopy and image analysis.
In: Process Mineralogy 12, Proceedings, Vineyard Hotel, Cape Town, South
Africa, 79 November, pp. 114.
Gottlieb, P., Wilkie, G., Sutherland, D., Ho-Tun, E., Suthers, S., Perera, K., Jenkins, B.,
Spencer, S., Butcher, A., Rayner, J., 2000. Using quantitative electron microscopy
for process mineral applications. JOM Journals of Metals, 2425.
Gu, Y., 2003. Automated scanning electron microscope based mineral liberation
analysis. Journal of Minerals and Materials Characterization and Engineering 2,
3341.
Henley, K.J., 1992. A review of recent developments in the process mineralogy of
gold. In: Extractive Metallurgy of Gold and Base Metals, Proceedings, Kalgoorlie,
pp. 177194.
Hestnes, K.H., Srensen, B.E., 2012. Evaluation of quantitative X-ray diffraction for
possible use in the quality control of granitic pegmatite in mineral production.
Minerals Engineering 39, 239247.
JKTech, 2010. JKTechs monthly e-Newsletter. December 2010.
Johnson, R.A., Wichern, D.W., 1998. Applied Multivariate Statistical Analysis.
Prentice Hall, 816p.
Jones, M.P., Gravilovic, J., 1970. Automatic quantitative mineralogy in mineral
technology. Rudy 5, 189197.
Kvarnstrm, B., Oghazi, P., 2008. Methods for traceability in continuous processes
Experience from an iron ore renement process. Minerals Engineering 21, 720
730.
Kwitko-Ribeiro, R., 2011. New sample preparation developments to minimize
mineral segregation in process mineralogy. In: Proceedings, 10th International
Congress for Applied Mineralogy (ICAM), Trondheim, Norway, 15 August, pp.
411417.
Lamberg, P., 2011. Particles the bridge between geology and metallurgy. In:
Conference in Mineral Engineering, Proceedings, Lule, Sweden, 89 February,
pp. 116.
Lamberg, P., Vianna, S., 2007. A technique for tracking multiphase mineral particles
in otation circuits. In: 7th Meeting of the Southern Hemisphere on the Mineral
Technology, Proceedings, Ouro Preto, Brazil, pp. 195241.

Lamberg, P., Hautala, P., Sotka, S., Saavalainen, S., 1997. Mineralogical balances by
dissolution methodology. In: Short Course on Crystal Growth in Earth Sciences,
Proceedings, S. Mamede de Infesta, Portugal, September 810, pp. 129.
Lamberg, P., Tommiska, J., 2009. HSC Chemistry 7.0 Users Guide. Mass Balancing
and Data Reconciliation. <http://www.outotec.com/39476.epibrw>, Outotec,
pp. 150.
Lawson, C.L., Hanson, R.J., 1974. Solving Least Squares Problems. Prentice-Hall,
340p.
Liipo, J., Lang, C., Burgess, S., Otterstrom, H., Person, H., Lamberg, P., 2012.
Automated mineral liberation analysis using INCAMineral. In: Process
Mineralogy 12, Proceedings, Vineyard Hotel, Cape Town, South Africa, 79
November, pp. 17.
LKAB, 2011. Annual Report. Lule, Sweden, Lule Graska, 122p.
Lorenzen, L., Van Deventer, J.S.J., 1993. The identication of refractoriness in gold
ores by the selective destruction of minerals. Minerals Engineering 6, 1013
1023.
Lotter, N.O., 2007. Distribution modelling of the nickel assay grades in nal tailings
at Raglan, Qubec. Minerals Engineering 20, 10671074.
Lund, C., Lindberg, T., Martinsson, O., 2010. Mineralogicaltextural characterisation
of different apatiteiron ore bodies, Malmberget deposit, Sweden, treated in a
sorting process in laboratory scale. In: Process Mineralogy10, Proceedings, Cape
Town, South Africa, pp. 19.
Martinsson, O., Virkkunen, R., 2004. Apatite iron ore in the Gllivare, Svappavaara,
and Jukkasjrvi areas. Society of Economics Geologists, Guidebooks Series 33,
167172.
Mattsby, C., 2010. A process mineralogical examination of the Vlkomma ore body,
Malmberget deposit, Sweden. Bachelor Thesis. Ume, Sweden, Ume University.
Moen, K., 2006. Quantitative measurements of mineral microstructure. Doctoral
Thesis. Trondheim, Norwegian University of Science and Technology, 194p.
Neumann, R., Avelar, A.N., 2012. Renement of the isomorphic substitutions in
goethite and hematite by the Rietveld method, and relevance to bauxite
characterisation and processing. In: Process Mineralogy 12, Proceedings,
Vineyard Hotel, Cape Town, South Africa, 79 November, pp. 110.
Niiranen, K., Bhm, A., 2012. A systematic characterization of the ore body for
mineral processing at Kiirunavaara iron ore mine operated by LKAB, Northern
Sweden. In: XXVI International Mineral Processing Congress (IMPC),
Proceedings, New Delhi, India, p. 1039.
berg, E., Plsson, B.I., 2008. Anvndning av processimulering fr att identiera
askhalsar i malmbehandlingssystem. In: Conference in Minerals Engineering,
Proceedings, Lule, Sweden, 56 February, pp. 114.
Paine, M., Knig, U., Staples, E., 2011. Application of rapid X-ray diffraction (XRD)
and cluster analysis to grade control of iron ores. In: Proceedings, 10th
International Congress for Applied Mineralogy (ICAM), Trondheim, Norway, 15
August, pp. 495501.
Paktunc, A.D., 1998. MODAN: an interactive computer program for estimating
mineral quantities based on bulk composition. Computers and Geosciences 24,
425431.
Penttinen, U., Palosaari, V., Siura, T., 1977. Selective dissolution and determination
of sulphides in nickel ores by the brominemethanol method. Bulletin of the
Geological Society of Finland Part 2 49, 7980.
Petruk, W., 2000. Applied Mineralogy in the Mining. Elsevier, Amsterdam, 268p.
Pirrie, D., Butcher, A.R., Power, M.R., Gottlieb, P., Miller, G.L., 2004. Rapid
quantitative mineral and phase analysis using automated scanning electron
microscopy (QemSCAN); potential applications in forensic geoscience.
Geological Society, London, Special Publications 232, 123136.
Pitard, F.F., 1989. Pierre Gys Sampling Theory and Sampling Practice, Sampling
Theory and Sampling Practice, vol. 2. CRC Press, Boca Raton Fl, US, 247p.
Potts, P.J., Bowles, J.F.W., Reed, S.J.B., Cave, M.R., 1995. Microprobe Techniques in the
Earth Sciences. Chapman & Hall, London, UK.
Rietveld, H.M., 1969. A prole renement method for nuclear and magnetic
structures. Journal of Applied Crystallography 2, 6571.
Schouwstra, R., de Vaux, D., Hey, P., Malysiak, V., Shackleton, N., Bramdeo, S., 2010.
Understanding Gamsberg a geometallurgical study of a large stratiform zinc
deposit. Minerals Engineering 23, 960967.
Stradling, A.W., 1991. Development of a mathematical model of a crossbelt
magnetic separator. Minerals Engineering 4, 733745.
Sutherland, D.N., Gottlieb, P., 1991. Application of automated quantitative
mineralogy in mineral processing. Minerals Engineering 4, 753762.
Whiten, B., 2008. Calculation of mineral composition from chemical assays. Mineral
Processing and extractive metallurgy review 29, 8397.
Young, R.S., 1974. Chemical Phase Analysis. Charles Grifn & Co Ltd, London, 138p.

IV

Incorporating ore and mineral textures in a geometallurgical model

of Malmberget iron ore

Cecilia Lunda*, Pertti Lambergb and Therese Lindbergc

a

Division of Geosciences and Environmental Engineering, Lule University of Technology, 971 87 Lule, Sweden
b
Division of Sustainable Process Engineering, Lule University of Technology, 971 87 Lule, Sweden
c
LKAB, Research & Development, 983 81 Malmberget, Sweden
* Corresponding author.Tel.: +46 920 492354. E-mail address: [email protected]

(Manuscript)

Incorporating ore and mineral textures in a geometallurgical model

of Malmberget iron ore

Cecilia Lunda*, Pertti Lambergb and Therese Lindbergc

a

Division of Geosciences and Environmental Engineering, Lule University of Technology, 971 87 Lule, Sweden
b
Division of Sustainable Process Engineering, Lule University of Technology, 971 87 Lule, Sweden
c
LKAB, Research & Development, 983 81 Malmberget, Sweden
* Corresponding author.Tel.: +46 920 492354. E-mail address: [email protected]

ABSTRACT
A geometallurgical model was developed in three steps using the Malmberget iron ore deposit, northern
Sweden as a case study. It is based on a mineralogical-particle approach which means that the mineral information is in the focus. Firstly, the geological model describes quantitatively the variation within the ore body
in modal composition and mineral textures. Traditional geological textural descriptions are qualitative and too
vague and therefore a method that describes and distinguishes quantitatively different mineral textures and
creates different types called textural archetypes was developed.
The second part of the geometallurgical model is a sub-model which forecasts how ore will break and which
kind of particles will be generated. A simple algorithm was developed to estimate the liberation distribution for
the progenies of each textural archetype. The model enables numerical prediction of the liberation spectrum
with varying modal mineralogy.The third step includes a process model describing quantitatively how different
particles behave in each unit process stage. As a whole the geometallurgical model takes the spatial information of the geological model in terms of modal composition and textural type. The particle breakage model
forecasts the liberation distribution of the corresponding feed to the concentration process and the process
model returns the metallurgical response in terms of product quality (grade) and effectively (recovery).

Key words
Geometallurgical model, iron ore, mineral textures, Particle Tracking, textural archetypes

1. INTRODUCTION
The geometallurgy embraces geological and metallurgical information to create spatially-based predictive
model (3D) for mineral processes (Lamberg, 2011). The industrial application is called geometallurgical program which is a structured effort to bridge all the adherent knowledge of the resource for production planning
and management.
Geometallurgical programs are needed for better resource management and to lower the risk in the process
operation related to geological variations within the ore body. It is a vital part of the protability of the operation. The mine needs to be capable to adjust the concentration process and the product qualities to meet the
requirements of a changeable global marked e.g. by a more effective utilisation of the ore resources or handle
larger volumes of lower grade ore. Today there exist different kinds of geometallurgical models depending on
the ore, its quality and the mineral processing circuit (Alruiz et al., 2009; Suazo et al., 2010; Hunt et al., 2012).
Most of the geometallurgical programs are established by using certain steps and relies on metallurgical and
geometallurgical testing (Dobby et al., 2004; Bulled and McInnes, 2005; David, 2007; Lamberg, 2011). A series
of representative ore samples are collected and are then tested to measure the metallurgical response directly
with a standard methodology (e.g. standard otation test).There are high expectations on the representativeness
of the samples and tests since they link the ore and metallurgical response. As the sample set should include
all variability in the ore this is often called a variability test. Based on the test results, a mathematical model is
created to explain the metallurgical response based on the sample characteristics. The model parameters are
almost always chemical components not minerals.
Iron mines are big volume operations and the production is driven by throughput. Lots of the mining companies produce high volumes of relatively low grade iron ore products with a Fe grade between 62% and 64%.
Examples of such production are direct shipping of hematite ores in Australia and Brazil. The LKAB, Swedish
iron ore producer, represents totally another end of production strategy. They produce customer tailored high
grade iron ore pellets (>67% Fe) and nes for blast furnaces and direct reduction (LKAB, 2011).
Good understanding on the raw material and its variability are essential in both the strategies. In the direct
shipping ores the production chain is very short; normally it includes only crushing and screening, and therefore if the ore is not suitable for the product requirements the tools in mineral processing are very few. In the
high grade products; where the tolerable grade of impurities like silica, aluminium, phosphorous and sulphur,
are very low; the success of processing is very dependent on the ore quality and its identication before it
comes to the plant.
In the literature there is very little information on the existing geometallurgical programs of iron ores, but a
few can be found, i.e. from Kiruna, Sweden (Niiranen and Bhm, 2012) and NW Australia (Paine et al., 2011).
Technique that is frequently used in evaluating the metallurgical variation in magnetite bearing iron ores is
Davis tube (Farrell et al., 2011; Niiranen and Bhm, 2012). It is a small scale magnetic separation test and the
ore samples are normally ground to the liberation size before testing. The corresponding concentrate and tail
are chemically analysed and the distribution of elements is then calculated.The iron distribution (recovery) and
the concentrate quality are used in predicting the metallurgical response in full scale operation.
The problem with using only the chemical components is that they are not the primary neither the direct
reason for the metallurgical response. As chemical components are bond in minerals it is more appropri-

ate to use mineralogy for building the metallurgical functions and the geometallurgical domains. However,
minerals do not occur independently in the processes they occur in particles which is varying in size, shape and
composition.
Lund et al. (2013) developed a practical, fast and inexpensive method to derive modal composition from
routine chemical assays using an element to mineral conversion technique which formed the rst part of a
geometallurgical model of the Malmberget deposit. However, the modal mineralogy is not sufcient to solely
answer the ore behaviour when processing the ore. The mineral textures play a signicant role and need to be
considered when a geometallurgical model is developed.
The texture characterisations are usually very subjective (Bonnici et al., 2008) and traditionally more
related to ore characterisation than process mineralogy (Perez-Barnuevo et al., 2012). In a mineral process, the
mineral textures and the liberation are closely associated. The included textures are one important parameter
that determines the limitation in an upgrading process (Butcher, 2010).The purpose of the comminution stage
is to liberate ore minerals appropriately for the concentration process to enable reaching required concentrate
quality with adequate recovery.
Geometallurgical models referred earlier generally almost solely ignore liberation information and the use of
textural information is quite undeveloped.
In mineral processing the relationship between the mineral (micro) textures and liberation has been a separate
research subject for a long time. Basically the aim has been to forecast the liberation distribution from a twodimensional picture of an ore.This is generally called the liberation model.The principle was presented already
in early 1900 by Gaudin (1939). Andrews and Mika (1975) developed a graphical presentation and this was
further developed by King (1975) and King and Schneider (1998). These models assume random breakage
which is unfortunately very rare especially in grinding. King and Schneider (1998) developed the model
further and included a kernel function which overcomes the problem of random breakage. Gay (2004) used
slightly different approach. Hunt et al. (2011a and 2011b) used chess-board pattern for crushed samples and
reduced the effect of the random breakage assumption. All these methods require two dimensional microphotograph of an ore sample. In a geometallurgical context this means a preparation of thin sections, their
photographing and image processing for a large number of samples.This is not very practical and an alternative
way is needed.
This case study aims to nd a solution how to incorporate mineral texture information into a particle
based approach (Fig. 1) modied from a concept by (Lamberg, 2011). This is done through a case study of
Malmberget iron ore deposit, in Northern Sweden. It focuses on mineral parameters, such as modal
mineralogy, mineral textures, mineral association, mineral grain sizes and their relation to the liberation
characteristics.
The nal purpose is to deliver a geological model which can offer quantitative rather than descriptive data
to be used in a process model. Firstly, the geological model is complemented with textural information.
Secondly, it is demonstrated how such geological model can be linked with a process model capable to forecast the
metallurgical response like grade and recovery for any given geological unit (sample, block, or domain).

Production
forecast

Particles
Geological
model

Unit process
model

Particle
breakage
model

Mineralogy
Textures

Simulation

Particles
behaviour

Fig 1. The particle-based geometallurgical concept modied from Lamberg (2011). Modal mineralogy and texture
links the geological model and process model. In the process model minerals are treated as particles. From the geological
information the particle population is generated through the particle breakage model.

2. SAMPLING, EXPERIMENT AND ANALYTICAL WORK

The samples in this study were collected from ve drill cores representing two ore bodies, Fabian (Fa) and
Printzskld (Pz) in the Malmberget deposit. Sampling, sample preparation and analytical methods used for the
mineral analyses and the element to mineral conversion are in detailed described by Lund et al. (2013).
Three dominate ore types were identied during logging, namely i) Feldspar (albite and orthoclase) rich ore
(Fsp) of Fa and Pz, ii) Apatite rich or of Fa and Pz and iii) Amphibole rich ore of Fa, generating a total of >100
kg in ve different composite samples.
The composite samples were processed and analysed at the LKAB, Malmberget laboratories. They were rst
crushed to a particle size below 3 mm. This was followed by a dry magnetic separation tests with Sala Mrtsell
magnetic separation using standard test procedure by LKAB producing two outputs: concentrate and tailings
(Lund et al., 2010). The test products were weighed and sized into ten size fractions using sieves of 38, 75, 150,
300, 425, 600, 1190, 1680 and 3000 microns. The size fractions were analysed for their chemical composition
by X-ray uorescence and SATMAGAN magnetic balance (Lund et al., 2013). Polished resin mounts were
prepared for ve nest size fractions (excluding 0-38 microns fraction) and analysed with QEMSCAN E430
Pro at LKAB, Lule. The QEMSCAN data was exported into Excel les and read into HSC Chemistry for
data processing.
The mineral textures were comprehensively characterised by optical microscopy with an attempt to develop
a quantitative systematic besides the modal mineralogy, i.e. GEM-types. For that purposes the grain size of
magnetite and the associating minerals were quantied in each texture.
The purpose of the processing of the mineral liberation data was to quantitatively track how different
kind of multiphase particles deriving from different textural type behave in processing. For that purpose a
particle tracking technique by Lamberg and Vianna (2007) was used. The Particle Tracking was done with

HSC Chemistry software developed by Outotec (Lamberg et al., 1997).The algorithms of the Particle Tracking
were also used in textural classication of the progeny particles, as described below.
Particle tracking uses mass balanced mineral by size data. Instead of using mineral grades by QEMSCAN, the
modal analysis was done with element to mineral conversion technique and calculation procedure developed
by Lund et al. (2013). Using HSC the raw mineral grades with solids ow rates data was rst balanced on
unsized basis (1D) and then for size by mineral basis (2D) with additional constraint that the 1D mass balance
is conserved.
The QEMSCAN analysis gave a total of 16 minerals but to make this match with the mineralogy of element
to mineral conversion some of the trace minerals were combined to more abundant phases (Lund et al., 2013).
This gave a total of ve most important minerals to be used in textural classication and in the particle tracking.
After combining the minerals, the next step of the Particle Tracking is to adjust the mass proportions of the
particles to meet the mass balanced 2D modal composition. This stage adjusts the mass proportion of measured particles but does not change their composition. The adjustment for the liberation data is very small but
is needed to make the particle data fully consistent with the 2D chemical mass balance. This means that if the
modal and chemical composition of the size fractions and streams is back calculated from the adjusted particles
the result is exactly the same as received from the 2D mass balance.
To mass balance particles they must be grouped somehow and in the particle tracking this is done by their
composition. The basic binning stage groups particles for different binary and ternary combinations. In the
system of ve minerals plus others, i.e. six phases, and ve size fractions after this step, the total number of
particles in each stream was 2166. Some of the particle groups may have very few particles if any. Therefore,
the advanced binning is a stage which combines established groups to ensure that each group has enough
particles for sound mass balancing of particle groups. The liberation measurement is challenging in very ne
(<10 micron) particle sizes, and in coarse size ranges the number of measured particles will be very small.
Therefore, the Particle Tracking includes a step to extrapolate the liberation information (i.e. particles) for size
fractions which have not been measured. In this study liberation was measured in ve size fractions, and the
remaining two size fractions were extrapolated.
After the previously described stages each process stream will be composed of similar particles classes but they
are not in balance. The last step reconciles and mass balances the whole circuit for the given particle classes.
This is done by keeping the total solids ow rate of the stream xed and by minimizing the weighed sum of
squares between measured and reconciled mass proportion of the particle classes. The particle tracking has an
option for smoothing the data, but this step was not used in this study.The error analysis was done with Monte
Carlo simulation which means that based on given standard deviations HSC drew a new dataset and solved the
particle mass balance. This was done 100 times, and nally the standard deviation for the particle grades and
recoveries was calculated.

3. DESCRIPTION OF THE TEXTURES

Textures are critical mineralogical characters in governing the ore behaviour in mineral processes. This is
commonly accepted and has been widely known since decades (Butcher, 2010). But unanswered question is
remaining how to describe textures and how to use the textural information in utilising the ore deposit and
managing the production, e.g. in geometallurgy (Bonnici, 2008).

The geological denition for textures is: The relative size, shape, and spatial interrelationship between grains
and internal features of grains in a rock (Spry, 1969; Fettes and Desmons, 2007). Textures, refers thus to a
generally description on the small scale properties which mostly is in qualitative terms, e.g. pegmatitic texture,
ned grained granoblastic texture. Descriptive mineralogical information, such this is insufcient in geometallurgy as it only gives possibilities to use the texture information for classication of different ore types. In
geometallurgy this may lead to a great number of domains and for each of them separate sample sets need
to be collected for geometallurgical/metallurgical testing. If, on the other hand, texture could be described
numerically, and even with additive parameters, this information could be processed with geostatistical
methods and be used in a similar ways as the metal grades are currently used in resource estimation (Glacken
and Snowden, 2001).

3.1. The Malmberget iron ore

The Malmberget iron ore deposit consists of several ore bodies of massive magnetite and hematite
(Geijer, 1930) which are exposed for an extensive regional metamorphism and deformation, intruded by several
generations of felsic and mac dykes (Martinsson and Virkkunen, 2004). The magnetite and hematite ores are
fully metamorphically overprinted seen as recrystallized coarse grained ore minerals, metamorphic textures,
different oxidation textures and a chemical redistribution of elements between the magnetite and hematite.
Compared to the well-known Kiirunavaara ore body, the Malmberget deposit is lower in tonnages, slightly
lower in Fe grade and has a moderate but varying P content (Bergman et al., 2001). Mineralogically the iron
ores are rather simple. Magnetite and hematite are the main ore minerals, and typical gangue minerals are
apatite and amphibole-pyroxene.
In the massive ore a broad variation of mineral-textures can be identied.The magnetite has textural variations
mainly in the magnetite grain size, grain shape and association. The texture is mostly granoblastic with distinct
triple junctions at the grain boundaries. Coarse grained magnetite occurs as porphyroblasts or as veins in a ner
grained matrix of magnetite. Larger magnetite grains can have an elongated shape following the mineral lineation in the host rock. In the western part of the deposit, hematite is the main mineral together with magnetite.
Typical metamorphic textures are exsolutions patches of spinel in magnetite or oxidation surfaces of hematite.
The magnetite grains show also intergrowth of ilmenite as lamellas or rutile needles.
In contrast to the massive ore semi-massive ore surrounds the ore bodies, particularly seen in the eastern part
of the Malmberget deposit. The semi-massive ore has traditionally been called ore breccia (Geijer 1930). It is
lower in Fe and higher in SiO2 and can extend several tens of meters with a decreasing iron grade. The main
non-iron minerals are the silicates i.e amphibole, feldspar (albite and orthoclase), quartz and biotite in various
proportions that generate several different mineral assemblages with complicated textures.
The distinct chemical discrepancies of the massive ore and the semi massive ore are important to consider as
the ore resource estimation uses traditional 3D block models which solely are based on the chemical analyses
of the ore samples. The mass proportion amounts of semi-massive ore that is present in the process is consequently very signicant and will not only inuence the ore reserve estimation but also contribute to elemental
variations in the mineral process. Therefore, a classication between the massive ore and the semi-massive ore
can clarify the true distribution of the chemical elements from the valuable and the gangue minerals and give
more precise information for dening the ore boundaries.

3.2. Textural classication

Using drill core logging (i.e. macroscopical observations) and optical microscopy an attempt was done to
develop a classication based on the mineralogical and textural relation of the massive and the semi-massive
ore. Textures were described and classied according to magnetite grain size and shape, generic mineral grain
size and the main associating mineral with magnetite (Table 1). The following textural types, arranged in the
order of increasing magnetite grade, were identied: disseminated, veiny, patchy, banded, granules, speckled and
massive.
Table 1. A classication scheme outlined for the massive and semi-massive ore from the Malmberget deposit based on
the mineralogical and textural relation.

Ore
type

Semi
massive
ore

GEMtype

Femineral

Fe
grade
*

Texture
type

Sub
type

Grain
shape

Grain
size,
mean
(m)

Texture description

Main
associating
mineral to
magnetite

Fsp

Mag

<10 %
(L)

Disseminated

Fine

Euhedral to
anhedral

44

Fine grain magnetite occur

as disseminated in the
silicate matrix

Fsp-Qtz

59-74

Mag grains, irregular

distribution in the grain
boundaries of ne grain
silica matrix or as smaller
inclusions in fsp grain

Mag

Qtz
bearing
Fsp+Amph

Mag

Amph(Ap-Bt)

Massive
ore

Fine

Fsp,
Bt-(AmphAp),
Qtz
bearing
Fsp+Amph

Amph(Ap-Bt),
Ap(Amph)

Ap(Amph)

Mag

Mag

Mag

10-80
%
(L-H)

>80 %
(H)

Veiny,
patchy,
banded,
granules

Speckled

90-100
%
(L-H)

90-100
%
(L-H)

90-100
%
(L-H)

Euhedral to
anhedral

Coarse

Euhedral to
anhedral

90-129

Coarse

Sub- to
anhedral

133

Larger magnetite grain

giving a semi-massive
appearance

FspQtzAmph

Sub- to
anhedral

100-200

Amph

Euhedral

200-400

Mag grain in a massive

appearance as a compact
equigranular matrix with
coarser grains occurring as
vein like structure

Subhedral

400-600
6002400

Mag grain occurs as

granoblastic texture with
distinct triple junctions
at grain boundaries or as
porphyroblasts (outline in
linear direction)

Bt+Amph+Ap

Mag grain occurs as

granoblastic texture with
distinct triple junctions
at grain boundaries.
Elongated grains outline
in a linear direction or
as porphyroblasts gives a
brittle appearance

Ap

Fine

Euhedral
Massive

Fsp-Qtz-BtAmph

Mag grain occur as single

grains in grain boundaries or as aggregate often
outline in veins or patches
in the matrix

Medium

Coarse

Porphyroblasts

Euhedral

20003200

8001400

Clear dependence between the magnetite grain size and magnetite (i.e. Fe) grade was found (Table 1). In the
low grade and disseminated texture ore the magnetite grain size is an average about 40 microns, and it increases
through a semi-massive ore to 140 microns up to 1400 microns in the massive ore (Fig. 2, Table 1). Even this

can be regarded as a general rule the massive ore and some textural types of the semi-massive ore show also
variation in the magnetite grain size and can be further classied as ne, medium and coarse grained.
The feldspar (Fsp) GEM-type in the semi-massive ore shows the biggest variation in textural features. It basically
consists of two different materials identied by their colour. Melanocratic magnetite rich material brecciates the
leucocratic feldspar -rich magnetite-poor matrix.The melanocratic parts consist of magnetite, apatite and amphibole veins,
bands or patches that may develop into networks in the leucocratic rock. The mineralogy of the leucocratic brecciating
rocks consists of a varying proportion of albite, orthoclase, quartz and biotite with minor magnetite (Table 2).

Table 2. Eight representative and different identied macrotextures of the classied Fsp GEM-type for different
ore bodies in Fabian and Printzskld ore bodies.
Texture type

Sub-type

(#) An increasing Fe-grade in the

geochemistry analysis

Fine grain texture

(Mgt-Fsp)
(Micro scale) (500 m)

Picture
(Macro scale) (4 cm)

Coarse grain texture

(Mgt-Fsp)
(Micro scale) (500 m)

Non existing texture

Disseminated (1)

Banded (2)

a)Waving
(3)

veins

Veiny
b)Small veins (7)
typical in Pz ore

Patchy(4)

Granules (5)

Clustered (6)

Speckled(8)

Non existing texture

The magnetite grains in the leucocratic parts are smaller in individual samples in all textural variants than in
the melanocratic parts as shown in Fig. 2. The difference is about 40-60 microns larger. In both leucocratic
and melanocratic parts the magnetite grain size increases with magnetite grade, and in addition the overall
magnetite grain size increases from type 1 to 7 (Fig. 2).The texture type 2 (banded) is deviating from the linear
correlation and thus shows an uncertainty when the Fe-grade is low.
140

Grain size (m)

120

Texure type

Melanocratic parts

(1) Disseminted
(2) Banded coarse
(2) Banded fine
(3) Veiny coarse
(3) Veiny fine
(4) Patchy coarse
(4) Patchy fine
(5) Granules coarse
(5) Granules fine
(6) Clustered coarse
(6) Clustered fine
(8) Speckled

100
80
60
Leucocratic parts
40
20
0

10

20

30

40

50

60

Fe (wt.%)

Fig. 2 Magnetite grain size magnetite grade (Fe) shows a positive correlation in the Fsp GEM-type of Fabian ore.
Fine grained and coarse grained form a bimodal system with two end members. The absence of texture type 7 is identied for Printzskld ore body.

The Bt-(Amph-Ap) GEM-type is structurally controlled and is located in the footwall of the massive ore.
Texturally, the Bt-(Amph-Ap) GEM-type is unique showing a schistose appearance due to the orientation of
biotite. Tabular biotite grains, mainly (10-400 m), sometimes as large as 2000 m, occur disseminated or as a
cluster that follows the lineation in the host rock either interstitially to a magnetite- or a silicate granoblastic
matrix (Table 3).
The Py bearing Amph-(Ap-Bt) GEM-type contains a signicant amount of sulphides. Single euhedral grains of
pyrite and chalcopyrite are in general 5-200 m in size, occasionally up to 4800 m and occur as dissemination
and in veinlets, at the grain boundaries or interstitially to magnetite. This texture occurs both in the massive
ore but more commonly in the semi-massive ore. The composition of this GEM-type regarding the mineral
texture and the modal mineralogy is closest to the the Amph-(Ap-Bt) GEM-type.
Two different GEM-types exist in the massive ore: Amph-(Ap-Bt) and Ap-(Amph). The Amph-(Ap-Bt) type
shows wide variation in the amphibole (-pyroxene) grain sizes and grain shapes, and two end members can be
identied.The coarser grained variant consists of large recrystallised amphibole (-pyroxene) grains occurring as

a cluster (0.8-3.2 mm) enclosing smaller magnetite grains (Table 3). In the ner grain variant, amphibole grains
(0.2-1 mm) occur interstitially to magnetite grains, either with or without the inclusion of small magnetite
grains. In both types the magnetite grains are similar in size; the biggest difference between them is that in the
coarse grained variant the magnetite grains also occur as inclusions in amphibole, whereas in the ner grained
variant, the magnetite occurs in the grain boundaries. These two textures can easily be identied during the
drill core logging due to the signicant difference in amphibole grain size (Table 3). The Ap-(Amph) type is
texturally homogeneous and have an equally sized apatite and magnetite which are closely associated existing
in a granoblastic or porphyroblastic texture (Table 3).
Table 3. Different ore textures that are identied and dened in massive and semi-massive ore.
Fine grain clustered (9)

Amph-Px

Mag

500 m

Subhedral to anhedral magnetite grains

(56-260 m) occur in a granoblastic matrix with amphibole-pyroxene
Anhedral ne grain amphibole (25180 m) occurs interstitially to magnetite
grains, either with or without the inclusion of magnetite grains (22-70 m)
The grain size of the matrix is medium
to coarse

Coarse grain clustered (10)

Single subhedral to anhedral magnetite
grains (10-200 m) occur interstitially in
grain boundaries and as several inclusions
in coarse or ne grained amphibole-pyroxene
The coarse amphibole grains exist as a
cluster (800-3200 m) or as ne single
grains (80-100 m) interstitially to magnetite
The grain size of the matrix is medium
to coarse

Mag

Amph-Px

500 m

Schistose (11)
500 m

Tabular single biotite grains (10-400 m)

occur as larger grains (up to 2000 m),
even a cluster or aligned in lineation interstitially to magnetite
The matrix is either a granoblastic magnetite or a silicate matrix
The grain size of the matrix is medium
to coarse

Bt
Fsp

Mag

Granoblastic texture (12)

Mag

0.5 mm

10

Anhedral to subhedral magnetite grains

occur either as a ne grained matrix
(100-400 m) or as a medium grained
(400-800 m) matrix
The magnetite matrix forms a dense texture sometimes giving a complex appearance of complicated grain boundaries
Other mineral associations are identied
to have a porphyroblastic texture, such as
magnetite-apatite, magnetite-ilmenite

Table 14 cont.
Porphyroblastic texture (13)
0.5 mm

Mag

Euhedral to subhedral magnetite grains

(800-1400 m) can have an elongated shape outlining a mineral lineation
parallel to the general fabric of the rocks
or occurring either as porphyroblasts
(up to 4600 m) or outlined in veinlike
structures in the ner anhedral magnetite
matrix
The grain boundaries are simple and
straight with clear triple junctions causing
the rock to be brittle

Mag

Euhedral to anhedral (200-2000 m)

single hematite grains or as larger porphyroblasts (up to 20 mm) interstitial to
magnetite grains
Other mineral associations are identied as
having a porphyroblastic texture such as
magnetite-apatite, magnetite-ilmenite

Hem

0.5 mm

Exsolution texture (14)

Mag

Hem

Magnetite grains show oriented exsolved

spinel patches (50-200 m)
Oxidation surfaces of hematite in magnetite grains or along the grain boundaries in
the magnetite matrix

0.5 mm

Lamella texture (15)

Mag

Magnetite grains show intergrowth of

ilmenite as oriented lamellas (trellis and
sandwich) approx. (2 m)
Homogenous hematite grains have inclusions of (20-100 m) Fe-Ti lamellas or as
(10-50 m) rutile needles at the rim of
the grain

Ilm

25 mm

The combinations of the different mineral textures are unique for almost each GEM-type, but what is common
that each of them has both a ne and a coarse grained variant (Table 4).
The textural study shows that when the mineral textures are considered the modal GEM-types divide into
numerous types, and the closer one looks the more complicated and numerous the classes become (Table 4).
In a geometallurgical context the use of classes is problematic since the treating of non-numeric data in block
modelling is challenging. Therefore, to change the geological model from descriptive to practical mineral

11

Table 4. The iron ore is classied into several classes and subtypes until unique categories of the mineral information
is distinguished to be used in a geological model. The Fsp GEM-type is used for detailed mineral texture analysis.
Classied iron ore

(2) Fabian

Ore type

Sub-ore type

GEM-type

Textures

Semi-massive ore

Feldspar rich

Fsp

1-8, 11

Massive ore
Semi-massive ore

(3) Printzskld
Massive ore

Amphibole rich

Amph-(Ap-Bt)

9-10, 12-15

Apatite rich

Ap-(Amph)

12-15

Feldspar rich

Fsp

1-8, 11

Biotite rich*

Bt-(Amph-Ap)

11-13

Apatite rich

Ap-(Amph)

12-15

*This sub-ore type not sampled

textures must be changed from qualitative to quantitative. Therefore, the mineral textures are now considered
from the mineral processing viewpoint.
In mineral processing, ore is comminuted to liberate the minerals and to make the particle size suitable for
downstream processes. Comminution is an energy intensive stage, and therefore a good balance is aimed
between mineral liberation and throughput (i.e. used energy/feed tons). Full liberation is not a feasible target
since besides high energy required the separation efciency of downstream processes tends to decrease toward
very ne particle sizes (<20 microns). Therefore, after the comminution stage the targeted degree of liberation
for the ore minerals is typically 90-95% and different kinds of composite particles are present. A term liberation
distribution is used here to summarise the information on mineral deportment; thus, it describes the distribution of particles by their composition (cf. particle size distribution).
In comminution, the ore is broken into particles using multiple stages of size reduction, such as blasting,
crushing and grinding, and classication, like screening and hydrocycloning. The behaviour of an ore in the
comminution is dependent on machine parameters, such as nature and magnitude of the applied comminution
energy (unit operation properties and operational parameters), and on the materials physical properties, such as
elasticity, hardness and strength. These together will dene in a given process the relationship between specic
energy (energy/mass) and overall size distribution of the material.The term grindability describes this, and it is
the measure of the specic energy consumption required to reduce a certain mass of material from a given fresh
and initial size down to a dened product size by grinding. Similarly, the term crushability is used for crushing.
Comminution circuits are designed and operated to provide targeted product neness, and almost without exception this gure is xed or changed very seldom. However, it is common that the grain size and association
of the ore minerals vary within the ore body and therefore in plant feed on a daily basis. A good geometallurgical model should therefore not only forecast the metallurgical response but also give the best operational
parameters, i.e. the target liberation degree and accordingly the target grinding neness for any given rock unit
or plant feed (blend).
The liberation distribution of a comminuted ore is dependent not only on (1) nature and magnitude of applied
comminution specic energy and (2) crushability and grindability, but also on (3) modal mineralogy and (4)
mineral textures. The rst factor is independent of material. The three other factors cannot be separated fully
from each other since the physical properties controlling the comminution behaviour are highly dependent
on modal mineralogy and mineral texture. Since there are established techniques to experimentally determine

12

the crushability and grindability and to use them in process simulations, the third factor is taken as an isolated
parameter, and it gives together with the rst parameter the overall particle size distribution of the material
after comminution.
How does one decouple the effect of the modal mineralogy and texture into the liberation distribution?
Firstly, the mineral grain size is basically a pure textural property, but as shown above the magnetite grain size
has a positive correlation with the magnetite grade (Fig. 2). The associating mineral is controlled both by the
modal mineralogy and the texture. In the Fsp GEM-type the association of magnetite in the melanocratic and
leucocratic parts is different (Table 2). These examples show that the modal mineralogy and the mineral
textures are intimately mingled with each other, and their separation using a traditional textural denition is
practically impossible. Therefore, a new denition for mineral textures is given:
Two samples are texturally different if the liberation distribution by size (compensated against modal
mineralogy) is different after comminuted in similar conditions.
This separates the comminution properties and the modal mineralogy from the textural properties as well
as states that for the textural classication, the liberation distribution must be compensated against modal
mineralogy and studied by size.

3.3. Textural archetypes

By using the liberation distribution of the different GEM-types an attempt to quantify the mineral textures was
done using samples including different textural types (Table 4).
All of these samples show a variation in the modal mineralogy by size fraction as illustrated by the
magnetite grade by size in Fig. 3. This shows that the breakage is selective, and an attempt to forecast the
liberation distribution from uncrushed samples by applying the random breakage model will fail. Also, the grade
distribution varies by ore type. The Fsp GEM-types show an increase in the magnetite content by particle size.
A similar pattern is found in the Ap GEM-types with an exception that the nest size fraction has the highest
magnetite grade. Differing from the previous the Amph GEM-type shows the highest magnetite content in the
middle size fractions, i.e. 75-150 and 150-300 microns.The variation in the mineral grade by size gives the rst
challenge when comparing the liberation distribution (Fig 3).

100

Magnetite grade (wt.%)

Magnetite liberation degree (%)

Fa_Fsp

90

Pz_Fsp
Fa_Amph

80
70
60

Fa_Ap

50

Pz_Ap

40
30
20
10
0
38-75 m

Fa_Fsp
Pz_Fsp

100
90
80
70
60
50
40
30
20
10
0
Bulk

75-150 m 150-300 m 300-425 m 425-600 m

Size fraction

38-75 um

75-150 um 150-300 um 300-425 um 425-600 um

Size fraction

Fig. 3 The magnetite grade plotted against the particle size in ve different GEM samples from Fabian (Fa) and
Printzskld (Pz) (left). The degree of liberation for magnetite by size fraction in the Fa_Fsp and Pz_Fsp samples.

13

The degree of liberation of magnetite decreases in all samples by particle size (Fig. 3), which is logical and
found almost always in ore samples. However, the overall liberation degree (of magnetite) can be different in
the bulk sample even if the liberation degree by size is similar, if the overall particle size distribution if different.
Therefore, the degree of liberation from the textural point of view must be studied by size.The Fsp GEM-types
from Fabian and Printzskld show that clearly (Fig. 3): they have identical liberation degrees for magnetite in
four size fractions, but in the size fraction 425-600 microns, the Fabian sample shows much better liberation.
This causes the overall liberation degree to be different (Fig. 3).
In order to compare the liberation distribution of ve different samples, a characteristic size fraction,
150-300 microns, was selected. The magnetite grade in the 150-300 microns size fraction is close to the bulk
sample; the mass proportion in crushed sample was high enough and the number of particles measured gives
sound statistics. In the size fraction the degree of liberation of magnetite has a positive correlation with the

Liberation degree og magnetite (%)

magnetite grade, but the Fsp-type differs signicantly from the Amph- and Ap-types. In the Fsp-type the
magnetite liberation is better than the grade would suggest (Fig. 4A).

Pz_Ap

3.5

Association index (AI)

95
94

Fa_Ap

Fa_Fsp

93
92

Fa_Amp

91
90

Pz_Fsp
50

60

70

80

90

2
1.5
1
0.5
0

100

Fsp

38-75
150-300 m
425-600 m

8
7
6
5
4
3
2
1
0

AI Mgt-Fsp

AI Mgt-Bt

AI Mgt-Amp

Bt

Amph

10.0

1.0

0.1

AI Mgt-Ap

Ap
38-75 m
75-150 m
150-300 m
300-425 m
425-600 m

100.0

75-150
x
m
300-425 m

Magnetite deportment

Pz_Fsp
Fa_Ap

2.5

Magnetite grade (wt%)

Associatin index

Fa_Fsp
Fa_Amp
Pz_Ap

96

Liberated

Binary with Fsp

Binary with Bt Binary with Amp

Fig. 4 A) The degree of liberation for magnetite versus the magnetite grade of three different GEM-type textures (Fsp,
Amph and Ap) representing the size fraction, 150-300 microns (top left). B) The association index (AI) for magnetite
in size fraction 150-300 microns in the same GEM-type texture (top right). C) The association index of magnetite in
sample by size from Fa_Fsp fractions. D) The deportment of magnetite in sample by size fractions from the Fabian Fsp
GEM-type.

The comparison of mineral association for non-liberated particles is challenging. If just comparing the mass
proportion of magnetite with for example albite, the gure is affected by the magnetite liberation and the albite
grade. If the liberation degree is high, the association of magnetite with albite must be low, and similarly if the
albite grade is low the mass proportion of the binary magnetite-albite grains will also be low. Therefore, a new
Association Index (AI) was developed. It aims to describe how common it is to nd the target mineral locked
with the other minerals regardless of the liberation degree and the modal composition. The association index
is calculated for each mineral pair using the following formula:

AIAB =

Mineral A deportment with mineral B (when fully liberated grains are excluded) [wt%]
Mineral B grade (in a phase excluding mineral A) [wt%]

14

If the association index of a mineral pair A-B is 1, then the association of mineral A with B is as common as
the modal composition would suggest. If the association index is greater than 1, then the association is more
common than expected, and if it is lower than 1 then it is rarer than expected. Index value zero shows that there
is no association between minerals A and B.
Examples of the calculations of the associate index are given in Table 5. The three rst factious samples show
a similar association index with different modal compositions. Sample 4 shows that even if the deportment
percentages are equally high for the pair A-B and A-C, the association index shows that A can be found
relatively more often with phase C than with B. Sample 5 shows that the association index cannot be calculated
for binary systems (C=0).
Table 5. Calculation examples for the association index. For the calculation modal [(1)-(3)] and deportment [(4)-(7)],
the analysis result is needed.
Sample

(Row)

Formula

Modal composition (wt. %)

A

50

40

50

50

50

(1)

25

30

40

40

50

(2)

25

30

10

10

(3)

Deportment of the target mineral A, mass proportion (wt. %) of A in different particle classes
Liberated A

90

80

90

90

90

(4)

In binary A-B

10

10

(5)

In binary A-C

10

(6)

In ternary A-B-C

(7)

Mass proportion of minerals in a fraction (phase) excluding the target mineral A (wt. %)
B

50

50

80

80

100

(8)

100*(2)/[(2)+(3)]

50

50

20

20

(9)

100*(3)/[(2)+(3)]

Deportment of A excluding the liberated class (wt. %)

with B

50

50

80

50

100

(10)

100*[(5)+(7)]/[100-(4)]

with C

50

50

20

50

(11)

100*[(6)+(7)]/[100-(4)]

Association index for A

with B

1.000

1.000

1.000

0.625

1.000

(12)

(10)/(8)

with C

1.000

1.000

1.000

2.500

(13)

(11)/(9)

The association indexes of magnetite with feldspars (albite+orthoclase), biotite, amphibole and apatite are
shown for the selected size fractions of ve different textural types from the Malmberget iron ore in Fig. 4b.
The Fsp GEM-types of Fa and Pz show quite similar association indexes. The association index of magnetite
with feldspar (AIMgt-Fsp) is smaller than 1, which means that magnetite is locked with feldspar less often than
the modal mineralogy would suggest. On the other hand the association index of magnetite with amphibole
(AIMgt-Amph) is higher than 1 showing that the magnetite is met more often locked with amphibole than the
modal mineralogy would suggest.The association index of magnetite with apatite shows a signicant difference
between the Fsp and Amph (and Ap) GEM-types. In the Fsp GEM-types the magnetite is rarely occurring
with apatite, but in the Amph and Ap GEM-types association is more common than the modal composition
would suggest.
Because the association index is calculated for particulate material, it carries information on both the mineral textures (grain size, shape, associating mineral) and the ore breakage. This is clearly illustrated in Fig. 4c
which shows the variation of the association index by size. If the texture is fully homogenous and the breakage

15

fully random, the association indexes should be 1 for all minerals in all sizes. Deviation from 1 can be due to
heterogeneous textural or non-random breakage. The association indexes approach 1 as the particle size gets
coarser, as shown in Fig. 4d for the Fsp GEM-type. This is because in coarse particle sizes the particles start
to be identical in their modal composition and liberation distribution. This point is not reached yet at the
425-600 microns particle size for Fsp GEM-type sample of Fabian because the liberation degree of magnetite
is as high as 74% (Fig. 3).
The decoupling of the texture and breakage is not possible with the association index. Some more
information is needed,and therefore the association index values are compared to the textural description of original
uncrushed ore samples. The Fsp GEM-type consists of leucocratic and melanocratic parts and the magnetite
and amphibole content is higher in the latter and feldspar grade is higher in the former. Association index
values for magnetite with feldspar (AIMgt-Fsp) lower than 1 and higher values with amphibole (AIMgt-Amph) can
therefore be explained based on the texture. Additionally, the variation in the association index of magnetite
with apatite between the GEM-types can be explained from textural differences. In the Amph GEM-type,
apatite preferentially occurs in contact with magnetite, but this is not the case in the Fsp GEM-type.
The association index of magnetite with biotite shows strong variation by particle size: in ne particle sizes the
value is very high and decreases by size. Since biotite does not show a preferential association with magnetite
in the original uncrushed ore samples, this must be a breakage feature. Thus, it is signicantly more common
to nd magnetite locked with biotite in ne particle sizes than in coarse sizes, and this is not because biotite is
more common in ne size fractions. Therefore, it indicates rather that magnetite-biotite particles dont break
preferentially through grain boundaries rather the opposite.
Coming to the denition of texture, the Fsp GEM-types of Fabian and Printzskld can represent similar
textures since the liberation degree of magnetite and association indexes do not differ from each other strongly.
The Fabian Amph and Ap GEM-types are also potentially similar in texture, and the Printzskld Ap GEMtype is a textural type of its own. The question remains how small or big the difference between the liberation
distributions and the related key gures used here (degree of liberation and association index) should be to
justify saying that two samples are similar or different in their textures? In addition how this information will
be gathered and used in a geometallurgical context?
To extend the modal compensation also into liberation degree, the calculation algorithms of the particle
tracking technique were applied (Lamberg and Vianna, 2007). The particle tracking technique is for mass
balancing the liberation data, but for doing that it includes steps where the particles are classied in a
systematic way (basic binning and advanced binning), and the modal composition of the liberation measurement
is rened to match with the modal composition calculated by the element to mineral conversion. Therefore,
the liberation data of ve samples was classied (basic binning) producing an identical particle population in
each sample. After this, the liberation distribution of each sample was forecasted using the textural information
from another sample. For this an algorithm given in Table 6 was used (Lamberg and Lund, 2012). For example,
to forecast the liberation distribution of sample A from sample B (called archetype), the liberation spectrum of
the archetype (sample B) was taken, and it was rened by using the modal composition of sample A.

16

Table 6. Adjusting the modal composition in generating the particle population based on archetype for
given ore sample or block (Lamberg and Lund, 2012).
The particle population for a given sample is calculated by taking the particle population of the corresponding
archetype. This is now iteratively adjusted, and for the adjustment, a correction factor, k, is calculated for each
mineral (i) in a size fraction before each iteration round:
L

pj, fraction = pj, fraction x Y( r(i)p x ki, fraction )

j=1

Basically the formula above is a ratio of mineral grade from the geological model (M(i)) and the mineral grade
back calculated from the liberation data of the archetype (denominator). p refers to mass proportion of particle in
a size class and x(i) is the mass proportion of mineral in a particle.
The mass proportion of particle j (pj) is recalculated on each iteration round using the correction factor and an
equation:

ki, fraction =

M(i)fraction

n
(p(j)
j=1

x r(i)p )

Using equations given above, the particle mass proportion is iteratively adjusted until the difference between mineral grades of the geological model and archetype has reached the required tolerance.

The results of the magnetite deportment of one pair Fa_Fsp and Pz_Fsp is given in Table 7. The algorithm
forecasts the liberation distribution well in all other size fractions except in the coarsest 425-650 microns. This
is most probably due to the small number of particles measured (about 400).
Table 7. The magnetite deportment measured and estimated using the Pz_Fsp and the difference between the estimated and measured.
Measured

Estimated using
Pz_Fsp

Difference

Bulk

38-75 um

75-150 um

150-300 um

300-425 um

425-600 um

Mgt Lib

89.4

96.4

95.2

Mgt-Fsp

6.6

1.4

2.9

92.5

85.6

74.3

4.7

10.8

14.2

Mgt-Bt

1.1

1.4

Mgt-Amph

2.0

0.6

1.2

0.8

0.8

1.7

0.5

1.5

1.8

5.9

Mgt-Ap

0.4

Mgt Lib

90.7

0.0

0.0

0.1

0.3

2.1

96.2

95.9

91.5

87.3

86.7

Mgt-Fsp

7.0

2.1

2.4

5.7

10.8

10.6

Mgt-Bt

0.9

0.6

1.0

1.7

0.1

0.5

Mgt-Amph

0.8

0.8

0.4

0.6

1.1

1.2

Mgt-Ap

0.1

0.1

0.2

0.1

0.0

0.0

Liberation

1.3

-0.2

0.8

-1.0

1.7

12.4

Average

0.4

0.2

0.2

0.4

0.4

2.8

Fig. 5 shows the error in forecasting the liberation distribution with all of the combinations of the ve samples.
Putting a limit that the average error must be lower than 1%, and then the ve samples can be simplied into
two textural archetypes: Fsp and Amph/Ap. One sample from both groups is selected to represent the group,
and this sample is called a textural archetype.

17

Fa_Fsp
Fa_Amp
Pz_Ap

Average error in magnetite

depormtent, %

100

Pz_Fsp
Fa_Ap

10

0.1
Fa_Fsp

Pz Fsp

Fa_Amp

Fa_Ap

Pz_Ap

Fig. 5 The average error showing the magnetite deportment in the samples.

4. FRAMEWORK FOR A METALLURGICAL MODEL

The process model takes the information of the geological model and transfers it to forecast on the metallurgical performance. When developing a metallurgical model, one needs to answer questions, such as: (1) what is
the purpose of the model? (2) What is the level of the complexity of how material (ore) is described? (3) And
how detailed is the information that the model can provide on a metallurgical performance?
In a geometallurgical context the purpose of the model is to return important metallurgical
parameters in the resource (block) model.Typically in the processing plants this would include information like
throughput, concentrate quality, recovery of the commodities and the tailing quality. In the case of Malmberget, the
metallurgical parameters considered by the process model are (1): throughput, recovery of iron and concentrate
grade in terms of iron, phosphorous and silica. The tailing quality is not a limiting factor in Malmberget, and
there is no need to include that in the model.
In this study the geological model describes the resource on a mineralogical basis, and the process model
logically uses the same approach. Three different levels can be used (2). The unsized, 1D, level uses minerals on
an unsized basis, which means that there is no information on the particle sizes available. The sized, 2D, level
uses the mineral by size for the material description and for the process models. This enables including particle
size and the metallurgical functions can be size dependent.The liberation, 3D, level uses liberation information.
Therefore, the process must describe the behaviour of particles by their composition and size.
(3) The level of processing details means how detailed the process is described in terms of unit operations and
operational parameters. In the lowest level, one black box represents the whole process and no operational
parameters are included. In the other end of the complexity lies a model where all unit operations are described
and all effective operational parameters are included.
In this study the feasibility of the mineralogical approach is tested using a simple process: the one stage dry
magnetic separation (cobbing) test for the Fsp GEM-type of Fa and Pz ore (Fa_Fsp and Pz_Fsp). The process
model is developed for all three levels, namely 1D, 2D and 3D.

18

To reach the 2D level of the modal mineralogical, a mass balancing needs to be done on a size fraction level.
The readers are referred to (Lund et al., 2013) for the calculation rules and detailed results. The 3D level was
reach by the liberation analysis of the concentrate and tail and was then mass balanced using the Particle Tracking technique.

4.1. Comminution - particle breakage model

The particle brakeage model gives the liberation distribution of a sample when the information on corresponding textural archetype and modal composition is given (Fig. 6 and 7).The model converts 1D, i.e. unsized
modal composition, to 3D (liberation distribution) using a textural archetype as a basis and a simple algorithm
in adjusting the liberation data of the archetype to match with the given modal mineralogy. The texturalarchetype also includes the information on how the modal composition varies by size. The overall size distribution
model developed in a M.Sc. thesis by Koch (2013) was used.

100

Mgt=80

90

Mgt=70
Mgt=60

Liberation (%)

80

Mgt=50
Mgt=40
Mgt=30

70
60
50
40
30
20
10
0

Bulk

0-38

38-7

50

75-1

300 00-425 25-600 0-1190 0-1680 0-3000

3
4
60
119
168

150-

Size fraction (m)

Fig. 6 The mass proportion of fully liberated magnetite in different size fractions as the magnetite grade in the ore
varies between 30 and 80 wt%. The bulk refers to combined size fractions.

Magnetite association (.%)

50
40

Mag-(Amph-Px) binaries

Fsp=90
Fsp=75
Fsp=50
Fsp=25
Fsp=10

60

Magnetite association (%)

Mag-Fsp binaries
60

30
20
10
0

Bulk

0-38

38-7

50

75-1

50
40
30
20
10
0

0
0
0
-300 00-425 25-600 00-119 90-168 80-300
4
3
6
11
16

150

Size fraction (m)

Fsp=90
Fsp=75
Fsp=50
Fsp=25
Fsp=10

Bulk

0-38

38-7

50

75-1

0
0
0
-300 00-425 25-600 00-119 90-168 80-300
3
4
6
11
16

150

Size fraction (m)

Fig. 7 The mass proportion of magnetite is shown in a magnetite-feldspar binary (left) and a magnetite-(amphibole+
pyroxene) binary (right), as the ratio of feldspar to (feldspar+amphibole+pyroxene) varies between 10% and 90%. The
bulk refers to combined size fractions. The magnetite grade is 50 wt%.

19

4.2. Concentration model dry magnetic separation

The process model describes quantitatively the particle behaviour in each unit process stage and therefore
returns the metallurgical response (grades for Fe, Si and P and recovery for Fe) for any given geological
unit (sample, block, domain). In the minerals processing the behaviour of particles is dictated by the particle
properties; therefore, the unit models have to include the particle properties, such as size, composition and
density. The structure of the simulator binding up the unit models must be based on particles (Lamberg, 2011).
The unit operations models used in minerals processing can be divided into four types: 1) Comminution
models where particle size distribution changes. 2) Separation models where particles are distributed between
two or more output streams based on their physical properties. 3) Leaching and precipitation models where
the liquid phase is an active component and minerals dissolve and new phases are formed through chemical
reactions. 4) Simple mixers and mass distributors where material is mixed and distributed between the outputs.
In the comminution unit models (grinding mills, crushers), it is possible to use the particle breakage
model described above. Therefore, in the model the forecasting of the liberation distribution and the total size
distribution can be decoupled. In the latter the traditional populations balance breakage models can be used
(Weller et al., 1996: Alruiz et al., 2009;Vogel and Peukert, 2003).
Most of the separation and leaching models used in industry are semi-empirical. In minerals processing the
fundamental process model based entirely on physics, chemistry and particle properties are still quite far away
from being practical and accurate enough for everyday use. The development of property based models in
minerals processing requires that particle properties can be measured in different parts of the process. For this
the liberation analysis is a state-of-the-art technique (Sutherland and Gottlieb, 1991; Gu, 2003; Fandrich et al.,
2007), and X-ray tomography is an emerging method (Miller et al., 2003).
The liberation measurement gives the quantitative information on the particles in a process stream, but for
modelling purposes liberation data must be mass balanced. This is done by the Particle Tracking Technique
(Lamberg and Vianna, 2007), a quantitative description of how different particle types (liberation classes) behave
in single process units and in a full process.
For the Malmberget concentration process, the rst unit model was developed for the dry magnetic
separation stage, i.e. cobbing. Fig. 8 shows the distribution of minerals on an unsized basis between the
concentrate and tailing for the Fsp GEM-type from Fabian. In a particle size of P80 = 1 mm, about 30% of the
material is rejected into the tail (MagTail) with about 6% magnetite losses, i.e. the recovery of magnetite into
the magnetic concentrate is 94% (Fig. 8).

20

Fig. 8 The processed mineral grade and recoveries in a Mrtsell dry magnetic separation (cobbing) test from for the
Fsp GEM-type of the Fabian ore type.

Studying the behaviour of the minerals by particle size it can be seen that for magnetite the recovery is quite
constantly between 92 to 96% in particle size range from <38 microns to >1.68 mm (Fig. 9). All of the
gangue minerals show a similar pattern having the recovery minimum between 38 and 106 microns (Fig. 9).
Noteworthy is that the biotite shows higher recoveries than the other gangue minerals, especially in the nest
particle size fraction <38 microns. Also, minerals show a signicant difference in particle sizes coarser than 500
microns; albite shows clearly the lowest recovery, whereas apatite and biotite the highest. Excluding the nest
size fraction, the mineral recoveries can be explained by their association with magnetite (Fig. 9).Therefore, the
different patterns found for each mineral in their recovery by size are due to their liberation distribution and
the association with magnetite.

Mgt
Ab
Bt
Amph
Ap

90

Recovery (%)

80
70
60
50
40
30
20

100

Mineral Recovery %

100

Mineral
Ab
Amp
80
Ap
Bt
60
Mgt
40
20

10
0

10

100

1000

10000

Size fraction (m)

10

20

30

40

50

60

70

80

90 100

Mineral association with magnetite %

Fig. 9 The cobbing test with a sample from the Fsp GEM-type of Fabian shows the recovery of minerals by size (left).
The mass proportion of minerals associated with magnetite vs. mineral recovery where ve size fractions between 38
and 600 microns are shown. The fully liberated grains are shown for magnetite (right).

21

Entering into the liberation level, it is obvious that magnetite rich particles enter into the magnetic concentrate,
whereas particles rich in gangue minerals are found in the tailing (Fig. 10). The particle tracking technique,
however, gives quantitative information on the behaviour of the particles.
Table 8. QEMSCAN pictures of typical particles from the magnetic concentrate and tail, Fabian Fsp GEM-type samples. Brown is magnetite, orange is feldspar.
Ore type

Size fraction
(microns)

Concentrate

Tail

425-600

300-425

Fsp GEM-type,
Fabian ore body
150-300

75-150

38-75

Fig. 10 shows the recovery of magnetite-feldspar binaries into the magnetic concentrate by size and as a
function of the magnetite grade. Interestingly, the recovery curve is upward convex in ne particle sizes and
straightens toward the coarse particles.

Recovery into concentrate (%)

100
90
80
70

0-38 um

60

38-75 m

50

75-150 m

40

150-300 m

30

300-425 m

20

425-600 m

10
0

10

20

30

40

50

60

70

80

90

100

The mass proportion of magnetite inparticle (wt%)

Fig. 10 The recovery of the magnetite-feldspar binary particles into the magnetic concentrate by size fraction as a
function of magnetite grade in the particle.

22

The recovery curves are deviating only little from the linear; therefore, for simplicity a linear model was
used. For the simulation of the magnetic separation the split values for each liberated minerals were given
(Table 9). The basic assumption behind the model is that particles with a similar composition and size will
behave in the process in a similar way regardless of from which part of the ore they are derived.The simulation
was done using HSC Sim, and for multiphase particles the software uses the weighted average of the split values
given for the liberated minerals.
Table 9. A model for the dry magnetic separation. Split values for minerals by size when occurring fully liberated.
Fraction

Magnetite

Feldspar

Biotite

Amphibole

Apatite

Others

38-75 um

0.961

0.030

0.109

0.035

0.031

0.331

75-150 um

0.965

0.033

0.048

0.064

0.012

0.651

150-300 um

0.971

0.054

0.088

0.090

0.104

0.554

300-425 um

0.971

0.042

0.000

0.148

0.000

0.410

425-600 um

0.987

0.112

0.000

0.371

0.047

0.856

Three cases were simulated (Table 10). In the rst one Fa-Fa, the feed material was from Fabian, and the Fabian
process model (split values in Table 9) was used. In the second one called Pz-Fa, the feed material is from the
Printzskld, and the Fabian model is used. In the third simulation, the Fabian sample was used as a textural
archetype, and the Printzskld modal composition was used to generate modelled Printzskld feed. Also in
this simulation the Fabian model was used (Pz(Fa)-Fa). The linear model (Fa-Fa) introduces a small error, the
standard deviation of the difference for grade is only 0.29% and for recovery 1.3% (Fa-Fa). The Fa model
succeeds to forecast the metallurgical performance for Pz when the liberation information is readily available:
the standard deviation of the difference for grades is 0.93% and for recovery 2.7% (Pz-Fa).When the liberation
information is not available and the feed sample is generated from the textural archetype, the forecast is still
reasonably good; the standard deviation of the difference in grades is 0.92% and the recovery 5.5%.These error
levels clearly full the need that the relative error should be below 5% for technical sampling and estimations.
Therefore, the conclusion based on the limited testing is that the approach could be used for producing metallurgical parameters into the block model.
Table 10. The observed grades and recoveries in the magnetic concentrate in the cobbing test. The results are the
sum of the size fractions from 38 to 600 size fractions. Diff: Difference (Sim-Meas), R.Diff: relative difference =
100*(Sim-Meas)/Meas.
SampleModel
Case

Meas

Sim

Diff

R.Diff%

Meas

Sim

Diff

R.Diff%

Meas

Sim

Diff

R.Diff%

Mgt wt%

89.7

90.2

0.6

0.6

87.6

88.9

1.3

1.4

87.6

87.2

-0.4

-0.5

Fsp wt%

5.63

5.28

-0.35

-6.24

7.11

5.33

-1.78

-24.99

7.11

5.59

-1.52

-21.34

Bt wt%

0.85

0.76

-0.08

-9.57

1.24

1.71

0.47

38.14

1.24

2.95

1.71

138.21

Amp wt%

2.14

2.12

-0.02

-0.95

1.83

2.06

0.23

12.70

1.83

2.54

0.71

38.71

Fa-Fa

Pz-Fa

Pz from Fa archetype - Fa

Ap wt%

0.09

0.10

0.01

5.60

0.58

0.85

0.28

47.73

0.58

0.79

0.21

36.33

Fe wt%

65.4

65.8

0.4

0.6

63.9

64.8

1.0

1.5

63.9

63.7

-0.2

-0.3

Si wt%

2.58

2.44

-0.14

-5.33

3.06

2.64

-0.43

-13.90

3.06

3.10

0.04

1.27

P%

0.01

0.02

0.00

7.66

0.09

0.14

0.04

48.08

0.09

0.13

0.03

36.69

Mgt Rec%

94.9

93.0

-1.8

-1.9

91.6

92.6

1.0

1.1

91.6

90.9

-0.8

-0.8

Fsp Rec%

13.0

11.9

-1.1

-8.6

17.0

12.7

-4.3

-25.2

17.0

13.3

-3.7

-21.6

Bt Rec%

30.5

26.9

-3.6

-11.8

8.6

11.9

3.2

37.7

8.6

20.5

11.9

137.5

Amp Rec%

20.0

19.3

-0.7

-3.5

16.9

19.0

2.1

12.4

16.9

23.4

6.5

38.3

Ap Rec%

13.4

13.8

0.4

3.2

4.9

7.2

2.3

47.2

4.9

6.6

1.8

35.9

Std grade

0.29

0.93

0.92

Std Rec

1.3

2.7

5.5

23

As the process model is based on the particle properties, the model can be used more widely than traditional
unit models which need to be calibrated if there is a change in size distribution, modal mineralogy or liberation.

5. DISCUSSION
This is the rst geometallurgical model of the Malmberget iron ore deposit based on a mineralogical approach
which is developed in three separate steps.
In contrast to the developed geometallurgical concept by Niiranen and Bhm (2012) which is based on
geometallurgical and metallurgical testing using the Kiirunavaara deposit as a case study, the geometallurgical
model developed here aims to be a complementary and partly an alternative way. As this study uses mineralogical approach and generic tools the methodology can be applied directly in other ore types and commodities.
It demonstrates that when the geological model relies on a proper ore characterisation and provides highlevel quantitative mineralogical data the information is adequate (without elemental grades) to be used in the
resource estimation and in the block model.
The benet of a described geometallurgical model is that it enables to optimise and forecast the production on
a long term basis (Lamberg, 2011).To get the full benet the model should be established already in a feasibility
stage. For deposits in production, like Malmberget, the geometallurgical model can instead be used to recognise the process limitations and re-evaluate the mining plan. It facilitates also the possibility to make realistic
process predictions which potentially improve the performance by giving on daily basis useful and achievable
production targets.
To use modal mineralogy in geometallurgical classication for some weathered iron ores has shown to be useful
before (Paine et al., 2011; Neumann and Avelar, 2012) but the quantication and usage of mineral textures is
still a challenge and so far quite undeveloped. There are some new methods to produce the mineralogical and
textural data, rapidly and inexpensively (Hunt et al., 2009, 2011; Bonnici et al., 2008, 2009; Perez-Barnuevo et
al., 2012) and even use the textural information in modelling (Hunt et al., 2012). These methods are based on
optical image system which also addresses a more inexpensive alternative than the use of automated mineralogy
systems. The benet of using the developed textural archetypes to distinguish different mineral textures would
be an even more cost-effective analysis as it aims to identify the textural variants during drill core logging.
This was outside the scope of this study and must be developed pointing that there is urgent need to develop
better characterisation tools and analytical techniques to semi-automatically handle large amount of
samples and routinely identify different textures directly on drill cores (Pirard et al., 2008; da Costa et al., 2009;
Haavisto and Kaartinen, 2009).
The developed approach using textural archetypes needs still further development. How to identify if two
samples are texturally different and how different they must be to justify calling them different textural
archetypes? More testing is needed whether the particle breakage model can reliably forecast the liberation
distribution. Also it is still largely unknown how good the estimate on metallurgical performance must be in
order to be useful.
Using the modal mineralogy, a geometallurgical ore type (GEM-type) classication system was developed
which capture the most important features like the iron mineral grade, iron mineralogy and the gangue
mineralogy. This was to overcome the problem that the accuracy of the modal mineralogy by element to

24

mineral conversion was not as good as required (Lund et al., 2013). As the geometallurgical model uses mineral
grades, the error in the modal mineralogy is currently too high to be used in resource estimation (Pitard, 1989)
and this need to be further developed before to it can be applied as a process tool.
The developed metallurgical framework gave the opportunity to follow the particles and identify what types
of texture that matters in the process. Even if this geometallurgical model is based on a limited mineral process
(one separation step), an important questions to answer is if it still is possible to see some textural impact in
the nal concentrates? Parameters that is not included in the model and needs to be considered and further
developed is, in addition to a concentration model, also a comminution model that handle the variation in
hardness and throughput. More textural archetypes including more particle size fractions need to be tested.

6. CONCLUSION
In this paper a geometallurgical model using the Malmberget iron ore deposit as a case study was developed.
The main focus was the development of a method to quantify mineral textures which is one of the basic
requirements to dene the geological model. Firstly, this was solved by a descriptive classication scheme for
identication of the mineral textures which also was used to verify the accurateness of the developed textural
archetypes. Secondly, to quantify the mineral textures, a method was developed to describe and distinguish
quantitatively different mineral textures. This was done by processing the liberation data and decoupling the
mineral liberation and associations from the modal composition. In the recognition of unique textures two
key gures were used, namely the degree of liberation and the association index (AI). Each unique texture is
collected in a library and an individual member of it is called textural archetype.
Finally, the geometallurgical framework and the geological model were tested comparing the forecasted
metallurgical key gures against the ones received by metallurgical test. The difference in terms of product
quality (Fe, Si and P grade) and in iron recovery were reasonable small validating the developed methodology

7. ACKNOWLEDGEMENTS
This project is nancial supported by the Hjalmar Lundbohm Research Centre (HLRC), which we are
thankful for. The authors want to express their gratitude to people at the LKAB for their contribution of
data and knowledge. ke Sundvall, Magnus Stafstedt, LKAB are acknowledged for comments and support.

25

8. REFERENCES
Alruiz, O.M., Morrell, S., Suazo, C.J. and Naranjo, A., 2009, A novel approach to the geometallurgical modelling
of the Collahuasi grinding circuit: Minerals Engineering, v. 22, p. 1060-1067.
Andrews, J.R.G. and Mika, T.S., 1975, Comminution of a heterogeneous material: development of a model
for liberation phenomena: In Proceedings of 11th International Mineral Processing Congress, Cagliari, Italy,
20-26 April, Proceedings, p. 59-88.
Bergman, S., Kbler, L. and Martinsson, O., 2001, Description of regional geological and geophysical maps of
northern Norrbotten County (east of Caledonian orogen): SGU Geological Survey of Sweden, v. Ba 56, p.
110.
Bonnici, N., Hunt, J., Berry, R., Walters, S. and McMahon, C., 2009, Quantied mineralogy and texture:
Informed sample selection for communition and metallurgical testing: The Tenth Biennial SGA Meeting,
Townsville, Australia, 17th- 20th August, Proceedings, p. 679-681.
Bonnici, N., Hunt, J., Walters, S., Berry, R. and Collett, D., 2008, Relating Textural Attributes to Mineral
Processing - Developing a More Effective Approach for the Cadia East Cu-Au Porphyry Deposit: Ninth
International Congress for Applied Mineralogy, Brisbane, Australia, 8-10 September, Proceedings, p. 415418.
Bulled, D. and McInnes, C., 2005, Flotation plant design and production planning through geometallurgical
modeling Publication Series, pp. 809-814.: Australasian Institute of Mining and Metallurgy, v. Publication
Series, p. 809-814.
Butcher, A., 2010, Chapter 4 - A practical Guide to Some Aspects of Mineralogy that Affect Flotation, Flotation
Plant Optimisation, Spectrum Series 16, p. 83-93.
da Costa, G.M., Barrn,V., Mendona Ferreira, C. and Torrent, J., 2009, The use of diffuse reectance spectroscopy for the characterization of iron ores: Minerals Engineering, v. 22, p. 1245-1250.
David, D., 2007, The importance of geometallurgical analysis in plant study, design and operational phases.,
v. Australasian Institute of Mining and Metallurgy Publication Series, p. 241-247.
Dobby, G., Bennett, C., Bulled, D. and Kosick, X., 2004, Geometallurgial modeling The new approach
to plant design and production forecasting/planning, and Mine/Mill Optimization: Proceedings of 36th
Annual Meeting of the Canadian Mineral Processors, Ottawa, Canada, 20-22 January, Proceedings,
p. paper 15.
Fandrich, R., Gu, Y., Burrows, D. and Moeller, K., 2007, Modern SEM-based mineral liberation analysis:
International Journal of Mineral Processing, v. 84, p. 310-320.
Fandrich, R.G., Bearman, R.A., Boland, J. and Lim, W., 1997, Mineral liberation by particle bed breakage:
Minerals Engineering, v. 10, p. 175-187.
Farrell, J., Miller, A. and Gaze, R., 2011, Geometallurgical Sampling and Resource Estimation for Magnetite Deposits: First AusIMM International Geometallurgy Conference (GeoMet), Brisbane, Australia, 5-7
September, Proceedings, p. 311-319.
Fettes, D. and Desmons, J., 2007, Metamorphic Rocks A Classication and Glossary of Terms: Cambridge, U.K,
Cambridge University Press, 244 p.
Gaudin, A.M., 1939, Principles of Mineral Dressing: New York, USA, McGraw-Hill Companies, 554 p.
Gay, S.L., 2004, A liberation model for comminution based on probability theory: Minerals Engineering, v. 17,
p. 525-534.
Geijer, P., 1930, Geology of the Gllivare Iron Ore eld: Geological Survey of Sweden, v. Ca 22, p. 1-115.
Glacken, IM., Snowden, DV., 2001, Mineral Resource Estimation, in Mineral Resource and Ore Reserve
Estimation The AusIMM Guide to Good Practice, p. 189-198. The Australasian Institute of Mining and
Metallurgy, Melbourne, Australia.
Gu,Y., 2003, Automated Scanning Electron Microscope Based Mineral Liberation analysis: Journal of Minerals
& Materials Characterization & Engineering, v. 2, p. 33-41.
Haavisto, O. and Kaartinen, J., 2009, Multichannel reectance spectral assaying of zinc and copper otation
slurries: International Journal of Mineral Processing, v. 93, p. 187-193.
Hunt, J., Berry, R., Bradshaw, D.,Triffett, B. and Walters, S., 2012, Development of liberation/recovery domains:
examples from the Prominent Hill IOCG deposit, Australia: Process Mineralogy12, Cape Town, South
Africa, 7-9 November, Proceedings, p. 1-16.
26

Hunt, J., Berry, R. and Bradshaw, D., 2011a, Characterising Liberation and Flotation Potential Using Image
Analysis, Simulated Fragmentation and Small-Scale Flotation:The rst AUSIMM international geometallurgy conference, Brisbane, Australia, 5-7 September, Proceedings, p. 331-333.
Hunt, J., Berry, R. and Bradshaw, D., 2011b, Characterising chalcopyrite liberation and otation potential:
Examples from an IOCG deposit: Minerals Engineering, v. 24, p. 1271-1276.
Hunt, J., Berry, R., Bonnici, N., Walters, S., Kamenetsky, M. and McMahon, C., 2009, From Drill Core to
Processing - A Geometallurgical Approach to Mineralogy and Texure From Meso- to Micro-Scale: The
Tenth Biennial SGA Meeting, Townsville, Australia, 17th-20th August, Proceedings, p. 685-687.
King, R.P. and Schneider, C.L., 1998, Mineral liberation and the batch communition equation: Minerals
Engineering, v. 11, p. 1143-1160.
King, R.P., 1979, A model for the quantitative estimation of mineral liberation by grinding: International
Journal of Mineral Processing, v. 6, p. 207-220.
Koch, P., 2013, Textural variants of iron ore from Malmberget: characterization, comminution and mineral
liberation: Master thesis, Lule, Sweden, Sustainable Process Engineering, Lule University of Technology,
101p.
Lamberg, P. and Lund, C., 2012,Taking liberation information into a geometallurgical model-case study, Malmberget, Northern Sweden: Process Mineralogy12 , Cape Town, South Africa, 7-9 November, Proceedings,
p. 1-13.
Lamberg, P., 2011, Particles the bridge between geology and metallurgy: Conference in mineral engineering,
Lule, Sweden, 8-9 February, Proceedings, p. 1-16.
Lamberg, P. and Vianna, S., 2007, A Technique for Tracking Multiphase Mineral Particles in Flotation Circuits:
7th Meeting of the Southern Hemisphere on the Mineral Technology, Ouro Preto, Brazil, Proceedings, p.
195-241.
Lamberg, P., Hautala, P., Sotka, S. and Saavalainen, S., 1997, Mineralogical balances by dissolution methodology: Short Course on Crystal Growth in Earth Sciences, S. Mamede de Infesta, Portugal, September 8-10,
Proceedings, p. 1-29.
LKAB, 2011, Annual Report: Lule, Sweden, Lule Graska, 122 p.
Lund, C., Lamberg, P. and Lindberg, T., 2013, Practical way to quantify minerals from chemical assays at Malmberget iron ore operations An important tool for the geometallurgical program: Minerals Engineering,
v. 49, p. 7-16.
Lund, C., Lindberg, T. and Martinsson, O., 2010, Mineralogical-textural characterisation of different
apatite-iron ore bodies, Malmberget deposit, Sweden, treated in a sorting process in laboratory scale: Process
Mineralogy10, Cape Town, South Africa, Proceedings, p. 1-9
Martinsson, O. and Virkkunen, R., 2004, Apatite Iron Ore in the Gllivare, Svappavaara, and Jukkasjrvi Areas:
Society of Economics Geologists, Guidebooks Series, v. 33, p. 167-172.
Miller, J.D., Lin, C.L., Garcia, C. and Arias, H., 2003, Ultimate recovery in heap leaching operations as
established from mineral exposure analysis by X-ray microtomography: International Journal of Mineral
Processing, v. 72, p. 331-340.
Neumann, R. and Avelar, A.N., 2012, Renement of the isomorphic substitutions in goethite and hematite
by the Rietveld method, and relevance to bauxite characterisation and processing: Process Mineralogy 12,
Vineyard Hotel, Cape Town, South Africa, 7-9 November, Proceedings, p. 1-10.
Niiranen, K. and Bhm, A., 2012, A systematic characterization of the ore body for mineral processing at
Kiirunavaara iron ore mine operated by LKAB, Northern Sweden.: XXVI International Mineral Processing
Congress (IMPC), New Delhi, India, Proceedings, p. 1039.
Paine, M., Knig, U. and Staples, E., 2011, Application of rapid X-ray diffraction (XRD) and cluster analysis
to grade control of iron ores: Proceedings, 10th International Congress for Applied Mineralogy (ICAM),
Trondheim, Norway, 1-5 August, Proceedings, p. 495-501.
Prez-Barnuevo, L., Pirard, E. and Castroviejo, R., 2012, Automated textural analysis of the Kansanshi Copper
ore: Process Mineralogy12, Cape Town, South Africa, 7-9 November, Proceedings, p. 1-16.
Pirard, E., Bernhardt, HJ., Catalina, JC., Brea, C., Segundo, F., Castroviejo, R., 2008, From spectrophotometry
to multispectral imaging of ore minerals in visible and near infrared (VNIR) microscopy. Proceedings, 9th
International Congress for Applied Mineralogy (ICAM 2008), 8-10 september, Brisbane, Australia.

27

Pitard, F.F., 1989, Pierre Gys sampling theory and sampling practice Vol. 2, Sampling theory and sampling
practice: Boca raton , US, CRC Press, 247 p.
Spray, A., 1969, Metamorphic textures, Pergamon Press, 350 p
Stamboliadis, E.T., 2008, The evolution of a mineral liberation model by the repetition of a simple random
breakage pattern: Minerals Engineering, v. 21, p. 213-223.
Suazo, C.J., Kracht, W. and Alruiz, O.M., 2010, Geometallurgical modelling of the Collahuasi otation circuit:
Minerals Engineering, v. 23, p. 137-142.
Sutherland, D.N. and Gottlieb, P., 1991, Application of automated quantitative mineralogy in mineral
processing: Minerals Engineering, v. 4, p. 753-762.
Vogel, L. and Peukert, W., 2003, Breakage behaviour of different materialsconstruction of a mastercurve for
the breakage probability: Powder Technology, v. 129, p. 101-110.
Weller, K.R., Morrell, S. and Gottlieb, P., 1996, Use of grinding and liberation models to simulate tower mill
circuit performance in a lead/zinc concentrator to increase otation recovery: International Journal of
Mineral Processing, v. 4445, p. 683-702.

28