©2017 Society of Economic Geologists, Inc.

Economic Geology, v. 112, pp. 919–940

Magnetite as an Indicator Mineral in the Exploration of Porphyry Deposits:

A Case Study in Till near the Mount Polley Cu-Au Deposit, British Columbia, Canada*
L. K. Pisiak,1 D. Canil,1,† T. Lacourse,2 A. Plouffe,3 and T. Ferbey4
1 School of Earth and Ocean Sciences, University of Victoria, Victoria, British Columbia V8W 2Y2, Canada
2 Department of Biology, University of Victoria, Victoria, British Columbia V8W 2Y2, Canada
3 Geological Survey of Canada, Ottawa, Ontario K1A 0E8, Canada
4 British Columbia Geological Survey, Victoria, British Columbia V8W 9N3, Canada

Abstract
The investigation of this paper focused on whether the composition of hydrothermal ore-related magnetite
in till could be used to locate porphyry deposits in terrane where glacial overburden overlies rocks that host
porphyry Cu-Au mineralization. This hypothesis was tested using 20 till samples collected in a ~900-km2 area
surrounding the Mount Polley porphyry Cu-Au deposit in south-central British Columbia, Canada. At least
100 magnetite grains were randomly selected from the magnetic fraction of each till sample. Nineteen trace
elements in ~50 magnetite grains in each sample were measured by laser ablation-inductively coupled plasma-
mass spectrometry (LA-ICP-MS). The large beam or raster size used for laser ablation (to 100 mm) homog-
enizes any heterogeneous trace element distributions in magnetite that result from oxy-exsolution and/or and
dissolution/reprecipitation, avoiding this issue with the few micron size of an electron beam. Linear discrimi-
nant analysis (LDA) performed on a compilation of magnetite compositions measured by LA-ICP-MS from
worldwide porphyry deposits and intrusive igneous rocks define the chemical signature (Mg, Al, Ti, V, Mn, Co,
Ni) of hydrothermal magnetite exclusive to porphyry systems. Application of our two LDA models to the 985
magnetite compositions we measured in the till samples surrounding Mount Polley showed anomalous amounts
of hydrothermal magnetite grains in till up to 2.5 km west-southwest and 4 km northwest of the deposit—a
pattern that is consistent with the ice-flow history of the region. Our LDA models for magnetite trace element
compositions have strong potential to be an effective tool in exploration for buried porphyry systems.

Introduction oxide Cu-Au (IOCG), and porphyry Cu(-Mo-Au) (McQueen

Indicator mineralogy is a powerful tool in exploration and has
had proven success in locating a variety of mineral deposits,
including diamondiferous kimberlite pipes and base metal
sulfides (Averill, 2001). For porphyry Cu(-Au) systems, indi-
cator minerals have included a large variety of accessory
and alteration minerals that occur in these deposits such as
apatite, epidote, garnet, jarosite, rutile, and tourmaline, and
ore-related minerals such as pyrite, chalcopyrite, and gold
(Averill, 2011; Kelley et al., 2011; Celis et al., 2014; Hashmi
et al., 2015). Magnetite chemistry has been applied in sedi-
mentary provenance studies (Grigsby, 1990; Razjigaeva and
Naumova, 1992; Yang et al., 2009), and as a resistate mineral
ubiquitous in porphyry systems, shows potential as an indica-
tor mineral in exploration (Dupuis and Beaudoin, 2011; Nad-
oll et al., 2012; Dare et al., 2014).
Magnetite (Fe3O4) is a member of the spinel group of min-
erals that are recognized for their systematic variations in
composition as a function of temperature, cooling rate, fO2, or
silica activity (Buddington and Lindsley, 1964; Irvine, 1965;
Sack and Ghiorso, 1991; Roeder, 1994; Barnes and Roeder,
2001; Dare et al., 2014). Several studies suggest that varia-
tion in minor and trace element contents in magnetite is sys-
tematically related to the type of mineralizing environment,
including volcanogenic massive sulfide (VMS), skarn, Fe
† Corresponding
author: e-mail, dcanil@uvic.ca
*A digital supplement containing electronic Appendix Tables A1 and A2 is
available at http://economicgeology.org/ and at http://econgeol.geoscience-
world.org/. Appendix text follows this paper.
0361-0128/17/4494/919-22 919
and Cross, 1998; Kamvong et al., 2007; Dupuis and Beau-
doin, 2011; Nadoll et al., 2012, 2014, 2015; Dare et al., 2012,
2014; Huang et al., 2014; Chen et al., 2015; Canil et al., 2016;
Makvandi et al., 2016).
Despite a dearth of experimental data for the partition of
trace elements in magnetite crystallized from fluid (Ilton and
Eugster, 1989; Simon et al., 2004), several studies have used
an empirical approach to fingerprint the chemical signa-
ture of magnetite from various hydrothermal environments.
Dupuis and Beaudoin (2011) developed discrimination
diagrams for a number of deposit types including IOCG,
Kiruna, banded iron formation, porphyry Cu, skarn, Fe-Ti-
V, Ni-Cu-PGE, and VMS. Nadoll et al. (2012, 2014) used
factor analysis to identify factors that are able to discrimi-
nate between igneous, low-temperature hydrothermal and
metamorphic magnetite, identifying Mg, Al, Ti, V, Mn, Co,
Zn, and Ga as key elements. They determined that hydro-
thermal magnetite has elevated values for the Mg-Mn factor,
whereas igneous magnetite can be identified by high values
of Co-Ni-V. Nadoll et al. (2015) calculated discriminant
measures on magnetite from porphyry Cu and skarn depos-
its, as well as their respective host rocks, in the southwestern
United States. Dare et al. (2014) used data from a variety of
hydrothermal ore deposits and igneous rocks to broadly dis-
tinguish hydrothermal from igneous magnetite using Ti, Ni,
and Cr. Canil et al. (2016) used principal components analy-
sis on hydrothermal magnetite from porphyry Cu(-Au-Mo)
and skarn deposits to recognize positive correlations of Al,
Ti, and V related to temperature, and negative correlations
Submitted: June 24, 2016
Accepted: January 10, 2017
920 PISIAK ET AL.

of Sn and Mo, with Mn and Co, governed by fluid chemistry.
Few studies have examined magnetite as an indicator in till
proximal to an ore deposit (McMartin et al., 2011; Sappin
et al., 2014). There have been no ground-truth tests of mag-
netite in tills around a known porphyry deposit.
In this paper, the potential of magnetite trace element
chemistry as an indicator for porphyry Cu(-Au) deposits is
tested. Linear discriminant analysis to rigorously define the
signature of hydrothermal magnetite from mineralized por-
phyry systems is first used. Subsequently the application of
discrimination methods to 987 magnetite compositions mea-
sured in till surrounding the Mount Polley porphyry Cu-Au
deposit, British Columbia, Canada (Fig. 1) shows the veracity
and plausibility of using these methods for magnetite as an
exploration tool.
Geologic Setting
The Mount Polley porphyry Cu-Au deposit occurs in the
Quesnel terrane, an assemblage of Late Triassic to Early Juras-
sic arc rocks in the North American Cordillera (Fig. 1). The
intrusive complex hosting the Mount Polley deposit is a high-
level, NW-trending alkalic stock, approximately 4 by 6 km in
dimension, emplaced in metasedimentary and metavolcanic
rocks of the Nicola Group that forms an open NW-plunging
syncline (Fig. 2; Logan and Mihalynuk, 2005; Rees, 2013).
The intrusive complex is composed of fine-grained porphy-
ritic diorite and monzonite with subordinate plagioclase por-
phyry and syenite dikes with crystallization ages of 205 ± 3 Ma
based on the U-Pb zircon method (Mortensen et al., 1995;
Logan and Mihalynuk, 2005; Logan et al., 2007). Various mag-
matic-hydrothermal breccias are the major host for Cu-Au
mineralization, which occurs as disseminated and stockwork
chalcopyrite, bornite, and pyrite (Fraser et al., 1992; Rees
et al., 2013). The age of porphyry Cu-Au mineralization was
determined by Logan et al. (2007) as 205.2 ± 1.2 Ma based
on 40Ar/39Ar ages of hydrothermal biotite intergrown with
chalcopyrite.
The Late Wisconsinan Fraser Glaciation deposited a large
amount of glacial sediment over the Quesnel terrane and
much of the Interior plateau. Ice build-up of the Cordille-
ran Ice Sheet began ~29 ka, with complete deglaciation by
~11.5 ka (Ryder et al., 1991). Ice-flow indicators as measured
from erosional features on outcrop indicate two directions of
ice-flow movement in the Mount Polley region: (1) early west-
ward ice-flow (oriented at 250°–275°) during advancement of
valley and piedmont glaciers from the Cariboo Mountains in

H IC
N
HV

C CM
M

Fig. 1. Location map of the Mount Polley porphyry Cu-Au deposit (star), superimposed on the geologic terrane map of Brit-
ish Columbia (modified from Colpron and Nelson, 2011). Triangles are locations of samples used in the discriminant analysis
database from mineralized (yellow) and barren (blue) intrusions (Mountjoy , 2011; Canil et al., 2016). C = Catface, CM =
Copper Mountain, E = Endako, H = Hushamu, HV = Highland Valley, IC = Island Copper, M = Mesachie, N = Nimpkish,
P = Pine.
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 921

Fig. 2. Regional geology, mineral occurrences, and till sample locations (red dots) in the Mount Polley study area (modified
from Plouffe et al., 2013a; bedrock geology from Logan et al., 2010). Samples analyzed in this study are indicated as larger
symbols. Locations for Mount Polley bedrock samples are also shown (blue dot).
922 PISIAK ET AL.
the east, and (2) later north-westward ice-flow (oriented at
290°–330°) during the last glacial maximum related to an ice
divide to the south of Mount Polley at 52°N latitude (Plouffe
et al., 2013a). The NW-trending direction is the dominant
ice-flow movement recorded in the study area (Hashmi et al.,
2015).
Sampling and Analytical Methods
Till samples
A total of 74 till samples from an area of ~900 km2 surround-
ing the Mount Polley deposit gathered as part of a Natural
Resources of Canada Targeted Geoscience Initiative 4 pro-
gram (Plouffe et al., 2013b; Ferbey et al., 2014) was the
basis for our study (Fig. 2). For each sampling location,
~10 kg of till was collected at a depth of ~80 cm below the
soil weathering horizon. Bulk till samples were prepared at
Overburden Drilling Management Ltd. (Nepean, Canada)
by wet sieving to a <2-mm-size fraction and then concen-
trating by density on a shaking table. The recovered heavy
minerals were further concentrated in methylene iodide
(specific gravity 3.2) and sieved into three different sized
fractions: 1 to 2, 0.5 to 1, and 0.25 to 0.5 mm. Magnetic
separation was performed on the heavy mineral fraction of
the processed till samples. Magnetite grains (2–0.25 mm)
were picked from the ferromagnetic heavy mineral fraction
using a hand magnet and binocular microscope at the Uni-
versity of Victoria (Victoria, Canada). Approximately 100 to
125 grains/till sample were randomly selected, mounted in
epoxy, and polished for further study (Pisiak et al., 2015).
The composition of magnetite from intrusive rocks in the
Mount Polley deposit was previously investigated by Canil
et al. (2016).
Petrography
Reflected light microscopy was used to examine internal
textures of the magnetite. Magnetite grains were classified
according to their crystal form, state of preservation (e.g.,
intensity of pitting, alteration, or fracturing), exsolution tex-
tures, and abundance of mineral inclusions. A subset of mag-
netite grains in each sample was further examined with a
Hitachi S-4800 field emission scanning electron microscope
(SEM) at the University of Victoria using an accelerating volt-
age of 16 kV and an emission current of 20 nA. Backscattered
electron (BSE) imaging and semiquantitative energy disper-
sive X-ray spectrometry (EDS) were used to identify exsolu-
tion and alteration phases and inclusions.
Electron microprobe analysis
The concentrations of Mg, Al, Si, Ca, Ti, V, Cr, Mn, Fe, and
Ni in magnetite from a subset of the till samples were deter-
mined using a CAMECA SX 50 electron microprobe (EMP)
at the University of British Columbia (Vancouver, Canada).
The EMP was operated at an accelerating voltage of 15 kV,
beam current of 20 nA, and spot size of 5 μm. Counting times
of 20 s were used for all elements except V, Cr, and Ni (40 s).
Less than 5% of all analyses for most elements in magne-
tite were below the detection limits of the EMP (250 ppm),
except Cr and Ni for which ~30% of all analyses were below
detection.
Laser ablation-inductively coupled plasma-mass
spectrometry (LA-ICP-MS)
Fifty magnetite grains per till sample were randomly selected
for elemental analysis by LA-ICP-MS at the University of
Victoria. Grains with appreciable alteration, fracturing, or a
significant number of inclusions were excluded because they
would not be amenable to ablation without obvious contami-
nation of other phases. Analysis of Mg, Al, Si, Ca, Sc, Ti, V,
Mn, Cr, Co, Ni, Cu, Zn, Ga, Nb, Mo, Sn, Ta, and W was per-
formed using a 213-nm Nd: YAG UV laser operating at 50 to
63% energy and 10 Hz, and interfaced to a Thermo Scientific
XSERIES 2 Quadrupole ICP-MS.
Laser ablation using spots (25–55 μm diam) and rastered
lines (100–150 μm long) were employed depending on the
mineral size and texture (Fig. 3). Data collection for each
analysis included 30 s of background signal followed by 20 to
30 s of ablation, with a minimum of 60 s washout. External
calibration using standard reference material silicate glasses
NIST611, NIST613, and NIST615 was performed for every
10 analyses. Time-resolved spectra were processed off-line
using a custom spreadsheet. Obvious mineral inclusions at the
surface were avoided for ablation (Fig. 3), but heterogeneities
in time-resolved spectra, caused by small inclusions inter-
sected at depth by the laser, were edited to remove contami-
nated sections from the spectra. When such heterogeneities
were too large and/or frequent in the spectra, the analysis was
discarded.
Three different methods of standardization were employed:
(1) internal standardization using an assumed stoichiometric
amount of Fe in pure Fe3O4 magnetite (i.e., 72 wt % Fe), (2)
internal standardization using the Fe concentration as deter-
mined by EMP analysis, and (3) the ablation yield correction
factor (AYCF) method of Liu et al. (2008). The AYCF method
is an internal standard-independent calibration that corrects
for the absolute amount of material ablated and calibrates to
the total counts per second for each analysis assuming that
all elements in the sample have been analyzed (Liu et al.,
2008). For magnetite, the elements selected for LA-ICP-MS
analysis cover the range of known substitutions in magnetite
that occur in measurable concentrations and, therefore, safely
satisfies this assumption. A comparison of the three different
standardization methods is discussed further below.
Accuracy, precision, and detection limits using a beam with
a 25-μm diameter, and rastered over a 100-μm length were
monitored over time in each analytical session using the USGS
basaltic glass reference BCR2g (Table 1). Accuracy for Mg, Si,
Ca, Sc, Ti, V, Mn, Co, Ni, Nb, Mo, and W, using Fe as the
internal standard (in % RSD), is better than 5%. Elements Cr,
Fe and Ta have accuracies between 5 and 10%; Al, Cu, and
Sn between 10 and 20%; and Zn and Ga at 20 to 30%. Preci-
sion was better than 20% for all elements except Cr, Zn, Ga,
and W. Detection limits are reported using the 3σ criterion,
whereby the limit of detection for each element is calculated
using three times the standard deviation of the gas background
signal that contains zero analyte (Longerich et al., 1996).
Petrography
Magnetite grains in till exhibit a wide variety of morphologies,
textures, alteration, and inclusions (Fig. 3). In the till samples,
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 923

Fig. 3. Petrographic varieties of magnetite from till in reflected light at 20× magnification. A. Euhedral, homogeneous
magnetite grain with 100-μm-long laser-ablation raster, apatite (Ap) inclusion and minor hematite (Hem) alteration patches
(sample 12 PMA 081 A01-C3). B. Subhedral magnetite grain with ilmenite (Ilm) trellis lamellae and 110-μm-long laser-
ablation raster (12 PMA 077 A01-D1). C. Subhedral magnetite grain with very fine trellis lamellae, potassium feldspar (Kfs)
and apatite inclusions, hematite alteration, and two 30-μm laser-ablation spots (12 PMA 088 A01-D4). D. Subrounded,
heterogeneous magnetite grain with fine trellis lamellae, hematite alteration, as well as blebby titanite (Ttn) predominantly
concentrated in the core (12 PMA 087 A01-G6).

Table 1. Detection Limits, Accuracy, and Precision for Magnetite LA-ICP-MS Analysis

Element Isotope DL, min (ppm) DL, max (ppm) DL, med (ppm) Accuracy (%) Precision (%)

Mg 26 2.13 92 6.53 5 13
Al 27 0.53 121 6.23 13 11
Si 29 123 42,820 380 1 6
Ca 42 81 899 160 5 5
Sc 45 0.03 3.49 0.41 3 15
Ti 47 0.45 691 4.23 2 13
V 51 0.08 32 0.52 5 15
Cr 52 0.71 109 1.76 9 22
Mn 55 0.3 61 0.9 3 15
Fe 57 11 3,546 76 8 14
Co 59 0.02 1.17 0.13 4 15
Ni 60 0.1 6.08 1.63 4 17
Cu 63 0.08 3.85 0.54 14 15
Zn 66 0.12 28 0.4 22 21
Ga 69 0.01 1.5 0.07 29 30
Nb 93 0.003 0.91 0.05 2 12
Mo 95 0.02 1.17 0.12 2 14
Sn 118 0.04 0.88 0.08 15 14
Ta 181 0.001 0.94 0.01 10 17
W 182 0.004 1.49 0.04 1 26

Notes: Accuracy and precision calculated based on analysis of BCR2g; abbreviations: min = minimum, max = maximum, med = median
924 PISIAK ET AL.

magnetite grains vary in shape from subhedral-euhedral octa-
hedrons and cubo-octahedrons, to more commonly rounded
to subrounded forms, which likely result from sediment trans-
port. Some magnetite grains are highly fractured or strongly
pitted, whereas others are relatively pristine or homogeneous
(Fig. 3A). In reflected light, magnetite is typically gray with
a brownish tint, but is often characterized by exsolution of a
slightly lower reflective phase. Exsolved phases of ilmenite
(FeTiO3) or, very rarely, ulvöspinel (Fe2TiO4), and/or rutile
(TiO2) typically occur as trellis-type lamellae ranging from
submicroscopic to 10 μm in width (Figs. 3, 4). Parallel- or
sandwich-type lamellae are also observed, but are much less
common. Some magnetite grains contain more than one exso-
lution phase. For example, Figure 4A shows thick ulvöspi-
nel + rutile lamellae with thin interstitial trellis replaced by
an unknown Si-Al-Mg phase in magnetite. Ilmenite exsolu-
tion occurs by subsolidus oxidation during cooling, whereas
ulvöspinel exsolution occurs from cooling alone below 600°C
(Haggerty, 1991).
Two predominant types of alteration are present in mag-
netite from till: (1) martitization (hematite alteration) due to
oxidation, and (2) “sphenitization” by late-stage fluids (Hag-
gerty, 1976). Martitization is identified in reflected light,
and predominantly occurs along grain boundaries, crystallo-
graphic planes (lamellae), adjacent to fractures, or as localized
patches, with no preference for any specific textural or mor-
phological type of magnetite (Fig. 3). In comparison, altera-
tion of magnetite to titanite (sphenitization) is highly variable
and in some cases cryptic. In exsolved magnetite, titanite is
a replacement product of Ti-rich lamellae (Fig. 4C), form-
ing after oxy-exsolution, whereas in magnetite grains without
exsolution, titanite occurs as irregular stringers or fine dis-
seminations (Fig. 4D). In addition, titanite may form along
grain boundaries or as localized patches (Fig. 3D), similar to
hematite. Mineral inclusions may be present and/or abundant
in some but not all magnetite grains (Fig. 3A, C). There was
no relationship of inclusion occurrences to grain size, habit,
or chemistry. Apatite is the most common type of inclusion
followed by quartz, K-feldspar, and chalcopyrite in decreasing
abundance. Magnetite grains containing multiple generations
of coarser grained magnetite (>200 μm) recognized as disso-
lution/reprecipitation during hydrothermal processes in skarn
deposits (Hu et al., 2015) were rare in the total population,
and not recognized in any of the 985 magnetite grains ana-
lyzed. Nevertheless, as will be shown in what follows, only a
subset of the till magnetites are likely hydrothermal in origin.
Chemistry of Magnetite
Comparison of analytical methods
The complete compositional dataset for all magnetite grains
in till (n = 985) is given in an online Appendix Table A1. We
tested the accuracy of three calibration methods for the pro-
cessing of magnetite LA-ICP-MS spectra, in order to opti-
mize its application in mineral exploration. The “EMP Fe”
calibration method uses the Fe concentration in magnetite as

Fig. 4. Backscattered electron (BSE) images of magnetite. A. Coarse ulvöspinel (Usp) and rutile (Rt) lamellae with an inter-
stitial, fine, dark trellis lamellae of an Si-Al-Mg-rich phase (sample 12 PMA 095 A01-C11). B. Submicroscopic ilmenite (Ilm)
lamellae in magnetite (12 PMA 095 A01-A15). C. Preferential titanite (Ttn) alteration of ilmenite lamellae in magnetite (12
PMA 585 A01-F4). D. Stringers and disseminations of titanite in magnetite (12 PMA 088 A01-B15).
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 925
determined by EMP as the internal standard for LA ICPMS
spectra. To obviate the step of obtaining EMP data, the “stoi-
chiometric Fe” calibration method assumes the maximum
stoichiometric amount of Fe that can occur in magnetite
(72%) as the internal standard and, therefore, is an approxima-
tion that produces the maximum possible trace element con-
centrations. The third type of calibration, the AYCF method,
calibrates to the total counts per second in each analysis and,
therefore, requires no internal standard, and thus no prior
step of EMP analysis for Fe (Liu et al., 2008).
For all major and minor elements in magnetite, the stoichio-
metric Fe calibration method produced consistently higher
concentrations (by 13–35%) than the EMP Fe method (Fig.
5A). In contrast, the AYCF method produces element con-
centrations within only a few percent of that calculated using
the EMP Fe calibration method (Fig. 5B) with a maximum
difference observed of 20 to 25%. The low-mass elements
Mg, Al, Si, and Ca could not be compared between standard-
ization methods because spurious concentrations arise when
calibrated using Fe as an internal standard, possibly due to
the low concentration of Fe in the NIST reference materi-
als used for standards. On the other hand, the calibration
curves for Mg, Al, Si, and Ca, using the internal standardless
AYCF method, are well correlated and considered reliable.
Thus, the AYCF method is preferred over either internal
25,000
R² = 0.99
(A)
Cr (ppm) - 'Stoichiometric Fe'
20,000
15,000
10,000
5,000
0 0.1
25,000
(B)
R² = 1
20,000
Cr (ppm) - AYCF
15,000
10,000
5,000
0
0 5,000 10,000 15,000
Cr (ppm) - 'EMP Fe'
Fig. 5. A comparison of LA-ICP-MS calibration methods using Cr concen-
trations in magnetite as a representative example of a trace element, showing
1:1 reference line (solid black) and fitted trend (dotted). A. Good correla-
tion of data (R2 >0.99) is observed between calibration methods, however,
the “stoichiometric Fe” method of standardization consistently produces the
highest element concetrations. B. Element concentrations determined by
AYCF calibration that closely matches those of the “EMP Fe” method
standardization methods by providing accurate calibration of
all elements in magnetite while eliminating the time for an
additional analytical step in determining the concentration of
the internal standard.
Elements such as Mg, Al, Ti, V, Cr, and Mn occur in sig-
nificant concentrations in magnetite and are often detectable
by both EMP and LA-ICP-MS. Other elements that occur in
much lower concentrations (<0.1 wt %) such as Si, Ca, Cr, and
Ni, however, are often near or below detection for the EMP
(Fig. 6A). Significant scatter observed for certain elements
may be an artifact of the difference in beam size between the
EMP and LA-ICP-MS (Fig. 6B). For example, extensive exso-
lution is observed in a large proportion of the magnetite grains
in this study (Figs. 3, 4) and the lamellae are often at scales
coarser than the electron beam size (5 μm). The smaller beam
size of EMP is less able to homogenize entire bulk composi-
tion during analysis in this case (Dare et al., 2012). This differ-
ence in sampling would contribute significantly to differences
in concentrations for any element that is heterogeneously dis-
tributed in magnetite, such as Ti, which can be concentrated
in exsolution lamellae (Fig. 6B). A larger beam size is possible
for EMP, but the area rastered by the laser (150 × 80 μm) is
more efficient for homogenizing material during analysis, as
can be seen in images of Figure 3. In what follows, we use
100000
(A)
10000
Cr (ppm) – LA-ICP-MS
1000
100
10 R² = 0.98
1
0.1
1 10 100 1000 10000 100000
Cr (ppm) - EMP
10
Ti (wt.%) – LA-ICP- MS
1
R² = 0.33
0.1
20,000 25,000 0.01
0.01 0.1 1 10
Ti (wt.%) - EMP
Fig. 6. A comparison of LA-ICP-MS and EMP analytical methods for mag-
netite. A. Good correlation is observed for Cr between the two techniques,
however, the EMP performs poorly when close to the detection limit as indi-
cated by data scatter below ~1,000 ppm. B. Significant scatter is present in Ti
between the two techniques, likely the result of a 5-um EMP beam size being
unable to homogenize exsolution lamellae with host magnetite.
 926 PISIAK ET AL.
element concentrations determined by LA-ICP-MS analysis
rather than EMP because of the greater sensitivity and lower
detection limits of the former instrument, and the fact a laser
spot or raster samples more of the crystal during analysis and
is, therefore, a better representation of the total preexsolution
magnetite composition before exsolution occurred.
“Sphenitization:” High Si and Ca in magnetite
About 90% of the magnetite grains sampled from tills in the
Mount Polley region are characterized by elevated concentra-
tions of Si and Ca, up to 7.3 and 6.0 wt %, respectively. Both
Si and Ca do not have extensive substitutions in spinel group
minerals from the Earth’s crust (O’Neill and Navrotsky, 1984).
A number of studies have documented several weight percent
Si in magnetite from various environments with Ca typically
<1 wt % (Vincent and Phillips, 1954; Shcheka et al., 1977;
Newberry et al., 1982; Shiga, 1988; Westendorp et al., 1991;
Dare et al., 2014). The high Si contents could be attributed to
an Fe2SiO4 equiv spinel component in solid solution with mag-
netite rather than contamination by minute silicate impurities
(e.g., Newberry et al., 1982; Westendorp et al., 1991). Alter-
natively, Shcheka et al. (1977) suggested variable amounts of
a Ca2SiO4 spinel component in magnetite based on similarly
high Ca contents. The majority of Si-bearing magnetite in the
aforementioned studies is described as zoned or containing
discrete domains in which Si is concentrated, features that are
not observed in magnetite from our till samples.
Most magnetite grains in till that are enriched in Si and Ca
are characterized by approximately equal atomic proportions
of these two elements (Fig. 7). This is inconsistent with sub-
stitution as a Ca2SiO4 spinel component, but is the expected
1:1 stoichiometry of Si and Ca in titanite (CaTiSiO5), which
is observed petrographically in many magnetite grains on a
variety of scales (Figs. 3, 4). Using the Si and Ca contents, the
maximum amount of titanite alteration in magnetite is 33%
(Fig. 8).
2,500
Till
Mt. Polley
2,000
Ca (ppm)/m.w.
1,500
1,000
500
0 magmatic magnetite (Nadoll et al., 2014); however, magnetite
0 500 1,000 1,500 2,000 2,500
Si (ppm)/m.w.
Fig. 7. Ca and Si contents in magnetite normalized to their molecular
weight to show approximate atomic proportions. The 1:1 line (dashed) is the
expected stoichiometry for Ca and Si atoms in titanite. Several magnetite
grains from till with high Si and low Ca may indicate potential interference
by minute quartz inclusions in these grains, or possible Si in solid solution.
Mount Polley magnetite data from Canil et al. (2016) and recalibrated using
the AYCF method.
Sphenitization of magnetite can occur by metasomatic
introduction of SiO2 and CaO in late-stage H2O-rich fluids or
by the breakdown of plagioclase, followed by the removal of
FeO from magnetite to form other minerals in the rock such
as sulfide, chlorite, or amphibole (Haggerty, 1976). This phe-
nomenon has been documented in metasomatized intrusive
rocks (Weisse et al., 1985), I- and S-type granites (Whalen and
Chappell, 1988), and hydrothermally altered oceanic crust
(Shau et al., 2000). Weisse et al. (1985) reported preferential
titanite alteration of ilmenite lamellae in magnetite from the
Nahant gabbro, United States, similar to that observed in till
near Mount Polley (Fig. 4C). Haggerty (1976) documented
examples of titanite replacement for both exsolved lamellae
and host magnetite.
In five porphyry and skarn deposits of British Colum-
bia, sphenitization was observed petrographically in very
few samples. The chemistry of magnetite from Mount Pol-
ley (<0.3 wt % Si, <0.05 wt % Ca) confirms that this is not
a characteristic of the deposit (Canil et al., 2016). In fact,
greater degrees of sphenitization in magnetite in till samples
were observed at greater distances downice or away from
the Mount Polley deposit. As sphenitization is relatively
common during regional, low-grade metamorphism (Hag-
gerty, 1976), grains in the till with any appreciable amount
of this effect are likely sourced from the metamorphosed
mafic volcanic and sedimentary rocks of the Late Triassic
Nicola Group that dominate the study area (Fig. 2) rather
than being hydrothermal in origin and specific to the Mount
Polley porphyry deposit.
Compositional variability in magnetite from till
The large compositional variability observed in the detrital
magnetite grains reflects the diverse bedrock sources for till
in the Mount Polley region. For brevity, we describe only the
salient relationships between a few of the trace elements. Tita-
nium is the most variable of all elements measured in magne-
tite, ranging from 50 ppm to 13.5 wt %, and shows a strong
positive correlation with Al, which ranges up to 7.6 wt % (Fig.
8A). The Ti and Al contents of magnetite are related to tem-
perature (Turnock and Eugster, 1962; Buddington and Lind-
sley, 1964). Therefore, elevated Ti and Al concentrations in
most magnetite grains in till is consistent with a large pro-
portion of material derived from igneous source rock, likely
the volcanic rocks of the Nicola Group that cover most of the
study area (Fig. 2). In contrast, hydrothermal magnetite typi-
cally contains substantially lower Ti and Al than magnetite in
igneous rocks (Fig. 8A; Dupuis and Beaudoin, 2011; Dare et
al., 2014; Nadoll et al., 2014; Canil et al., 2016).
Trivalent V and Cr have maximum concentrations of 0.7
and 8.1 wt %, respectively. High V contents are typical of
at the Mount Polley deposit also contains significantly high V
(Fig. 8B). Magnetite with elevated Cr concentrations occurs
in serpentinized mafic-ultramafic rocks as “ferritchromit,” a
product of hydrothermally altered Cr-Al spinels (e.g., Mellini
et al., 2005). Chromium-rich magnetite could be derived from
serpentinites of the Slide Mountain terrane that crop out in
the far northeast corner of the study area (Fig. 2). Elevated
concentrations of Si and Ca related to secondary sphenitiza-
tion are also observed in some grains (Fig. 7).
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 927

100,000 Till
1000
SW USA Porphyry
BC Porphyry
10,000 Mount Polley 100

Al (ppm)

Mo (ppm)
1,000 10

100 1

(A) (C)
10 0.1
10,000 100

1,000 10

Sc(ppm)
V (ppm)

100 1

(B) (D)
10 0.1
1 100 10,000 0.01 1 100

Ti (ppm) Sn(ppm)
Fig. 8. Select binary plots showing the covariation in magnetite chemistry from till for (A) Ti and Al, (B) Ti and V, (C) Sn and
Mo, and (D) Sn and Sc. Magnetite analyses from Mount Polley and other porphyry deposits in British Columbia (Canil et
al. (2016) were recalibrated using the AYCF method in this study. Magnetite data from porphyry deposits in the southwest
United States is from Nadoll et al. (2015).

Scandium and Sn are measurable in nearly all magnetite
grains but are <77 and 12 ppm, respectively. Tin concentra-
tions are positively correlated with Mo and Sc in hydrother-
mal magnetite, but not in the variety of magnetite in the tills
(Fig. 8 C, D). Other minor elements that are detectable in
most magnetite grains include Ni (max ,3021 ppm), Cu (max
1,807 ppm), Co (max 759 ppm), and Ga (max. 213 ppm). Gal-
lium is weakly correlated with Mn (100–30,000 ppm) and Zn
(30–10,000 ppm), suggesting minor components of MnGa2O4
and ZnGa2O4. More than 10% of all analyses were below
detection for Nb, Mo, Ta, and W for which the maximum con-
tents of each element is less than ~150 ppm.
Identification of Hydrothermal Magnetite
Binary plots
Dupuis and Beadoin (2011) developed two diagrams to distin-
guish magnetite compositions of porphyry, iron oxide Cu-Au
(IOCG), skarn, Kiruna, banded iron formation (BIF), and Fe-
Ti-V deposits. Their Ni/(Cr + Mn) versus Ti + V plot is unable
to accurately classify magnetite from Mt. Polley (Fig. 9). Tests
of a second discrimination diagram of Ca + Al + Mn versus
Ti + V found that porphyry magnetite from till and bedrock
at Mount Polley partly coincide with the porphyry field (Fig.
10), but also extend into the fields for IOCG, skarn, Kiruna,
and Fe-Ti-V deposits (Nadoll et al., 2014; Canil et al., 2016).
The latter diagram is very problematic because Ca + Al + Mn
values will be grossly modified by sphenitization, which we
documented in several magnetite grains in till sourced from
low-grade metamorphosed bedrock (Fig. 7). The Al + Mn
versus Ti + V diagram of Nadoll et al. (2014, 2015) defines
fields for distinguishing skarn and porphyry deposits from the
southwest United States. With the exception of one sample
from Mount Polley, magnetite from the deposit is accurately
classified as porphyry (Fig. 11); however, magnetite from sev-
eral unmineralized igneous intrusions (Mountjoy, 2011) and
all magnetite from till are similarly classified as porphyry.
Finally, a plot of Ti versus Ni/Cr in magnetite proposed by
Dare et al. (2014) broadly distinguishes hydrothermal from
magmatic grains. Hydrothermal porphyry magnetite from
Mount Polley (Canil et al., 2016) falls within the hydrothermal
field in this plot (Fig. 12). Chromium is typically in very low
concentrations in hydrothermal magnetite (i.e., <150 ppm at
Mount Polley). Therefore, for a large number of analyses of
magnetite from Mount Polley and a small number of magne-
tite grains in till, the detection limit for Cr was used, produc-
ing a minimum Ni/Cr value. Magnetite from the till samples
primarily plots in the igneous field, with a small proportion
of grains overlapping into the hydrothermal magnetite field
(Fig. 12).
Because the discrimination diagrams shown in Figures 9,
10, and 11 are unable to properly identify hydrothermal mag-
netite from Mount Polley, these diagrams would be too gen-
eral or ineffective in practice for an exploration program. The
Ti versus Ni/Cr diagram is straightforward and shows poten-
tial for application of magnetite in exploration for porphyry
928 PISIAK ET AL.

Fig. 9. Ni/(Cr + Mn) vs. Ti + V plot for magnetite compositions with discrimination fields from Dupuis and Beaudoin (2011).
Magnetite from Mount Polley plots at much lower Ni/(Cr + Mn) values than that proposed for porphyry deposits due to
relatively high Mn concentrations at the deposit (data from Canil et al., 2016). Most magnetite grains plot in the Fe-Ti-V field
with very few grains overlapping the porphyry field.

Fig. 10. Ca + Al + Mn vs. Ti + V plot of magnetite compositions with discrimination fields from Dupuis and Beaudoin (2011).
Magnetite from Mount Polley plots slightly above the porphyry field based on relatively high Mn concentrations (data from
Canil et al., 2016). Magnetite in till shows minor overlap with the porphyry field.
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 929

Fig. 11. Al + Mn vs. Ti + V plot of magnetite compositions with skarn and porphyry discrimination fields from Nadoll et
al. (2015). The skarn + porphyry field is a transitional zone between the two deposit types. Magnetite from Mount Polley is
successfully classified as porphyry magnetite, with the exception of one analysis (data from Canil et al., in press, recalculated
using the AYCF calibration method). There is no distinction for igneous magnetite, therefore, magnetite data from barren
igneous intrusions in the Cordillera (Mountjoy, 2011) and till are erroneously classified as porphyry magnetite.

Fig. 12. Ti vs. Ni/Cr plot of magnetite compositions with igneous and hydrothermal discrimination fields from Dare et al.
(2014). Analyses with Cr values below detection were assigned the value of the detection limit (1.76 ppm), which gives a mini-
mum Ni/Cr value. Magnetite from Mount Polley is successfully classified as porphyry magnetite (data from Canil et al., 2016).
Magnetite from till is predominantly classified as igneous, but a small proportion of grains plot in the hydrothermal field.
930 PISIAK ET AL.

deposits (Fig. 12), but hydrothermal magnetite is poor in Cr
and detection limits will introduce practical uncertainty. Hu
et al. (2015) also contested the validity of these diagrams in
Figures 9 and 10 because EMP analysis (on which these dia-
grams are based) is prone to the problem of a small beam size
in sampling the larger heterogeneities in hydrothermal mag-
netite. We show that laser ablation, particularly with rastering,
can sample at a scale beyond most heterogeneities from exso-
lution, and better integrate the total composition of magnetite
(irrespective of the origin) while still avoiding obvious mineral
inclusions at the surface (Fig. 3). Nevertheless, there may be
too wide a compositional range in hydrothermal magnetite to
be accurately classified using simple binary diagrams.
Discriminant analysis
A multivariate statistical method that applies the collective
significance of all elements, rather than just a few, may be
a more robust approach to characterize the composition of
magnetite as an indicator mineral (Nadoll et al., 2014, 2015).
We used discriminant analysis (Davis, 2002) to test for differ-
ences between magnetite from various sources. We compiled
a database of trace element concentrations in magnetite (n
= 481) from three petrologically distinct groups to create a
training dataset for the discriminant analysis (Table 2). Hydro-
thermal porphyry magnetite was designated as that occurring
in veins and hydrothermal breccias, thus showing direct
petrographic evidence of precipitation from fluids. Ore igne-
ous magnetite is eu-subhedral and occurs in fresh or altered
intrusive rocks located at a porphyry deposit. Barren igneous
magnetite is eu-subhedral, but from unmineralized intrusions
that are remote from or show no relationship to any kind of
ore deposit. The database was then divided into two subsets
based on the whether the analytical data were measured all in
one lab, or in different labs.
For brevity here, the extended procedures and details of
the discriminant analyses are given in the Appendix 2 and
elsewhere (Pisiak, 2015). We determined the elements that
have the most discriminating power among magnetite com-
positions from these three groups (hydrothermal porphyry
magnetite, ore igneous magnetite, and barren igneous mag-
netite) to define discriminant functions that uniquely iden-
tify the provenance of magnetite grains. A linear method of
discriminant analysis (LDA) was chosen due to the simplic-
ity of linear functions. Table 3 summarizes the results of the
two discriminant analyses, subsequently referred to as LDA1
and LDA2 for the two different subsets of data, described
in detail in Appendix 2. The discriminant scores for the two
analyses show a clear separation of the magnetite composi-
tional groups in a territorial plot of the linear functions (Fig.
13A, B). Factor 1 (F1) effectively separates hydrothermal

Table 2. Source of Magnetite Data for Discriminant Analyses

Source Elements analyzed Locality Sample type n

Canil et al. (i016) Mg, Al, Si, Sc, Ti, V, Cr, Mn, Co, Copper Mountain Cu-Au deposit, BC, CAN HTP 19
  Ni, Cu, Nb, Mo, Sn, W, Re Endako Mo deposit, BC, CAN HTP 14
Island Copper Cu-Mo-Au deposit, BC, CAN HTP 27
OIG 12
Mount Polley Cu-Au deposit, BC, CAN HTP 45
OIG 5
Pine Cu-Au deposit, BC, CAN HTP 28

Mountjoy (2011) Mg, Al, Sc, Ti, V, Mn, Co, Ni, Highland Valley Cu deposit, BC, CAN OIG 32
  Cu, Nb, Mo, Sn, W Hushamu intrusion, BC, CAN BIG 9
Nimpkish batholith, BC, CAN BIG 7
Mesachie Lake pluton, BC, CAN BIG 16
Argonaut intrusion, BC, CAN BIG 8

Nadoll et al. (2015) Na, Mg, Al, Si, S, Ca, Ti, V, Cr, Chino Cu deposit, NM, USA HTP 13
  Mn, Co, Ni, Cu, Zn, Ga, Ge, As, OIG 32
  Se, Sr, Y, Nb, Mo, Ag, In, Sn, Ba, Cobre Cu deposit, NM, USA HTP 8
  Ce, Au, Tl, Pb, Th OIG 8
Henderson Mo deposit, CO, USA OIG 41
Morenci Cu deposit, AZ, USA HTP 8
OIG 5
Morenci Garfield Cu deposit, AZ, USA OIG 7
Morenci Producer Cu deposit, AZ, USA HTP 7
OIG 1
Morenci Sun Ridge Cu deposit, AZ, USA HTP 15
OIG 4
Safford Dos Pobres Cu deposit, AZ, USA HTP 24
OIG 7
Safford Lone Star Cu deposit, AZ, USA HTP 25
OIG 20
Safford San Juan Cu deposit, AZ, USA HTP 18
OIG 16
    Inner Zone batholith, Japan BIG 12

Notes: Abbreviations: n = sample size; BIG = barren igneous; HTP = hydrothermal porphyry, OIG = ore-related igneous, only elements that were analyzed
in magnetite from all localities by the respective author are listed
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 931

Table 3. Results of Linear Discriminant Analysis for Two Magnetite Databases

A. LDA1                    Dataset: MDB1

      Discriminant functions Standardized functions Scores of group means

F1 F2 F1 F2 F1 F2

Mg 0.703 0.384 Mg 0.956 0.521             BIG –1.791  0.995

Al –1.387 0.091 Al –1.401 0.092             HTP 1.03  0.049
V –0.075 1.169 V –0.666 1.039             OIG –1.399 –1.005
Mn 0.412 –0.75 Mn 0.549 –1
Co –0.827 –0.525 Co –0.509 –0.323
Ni –0.306 –1.149 Ni –0.268 –1.005 Statistics
Constant –12.15 –11.256
F1 F2

Eigenvalue 1.667 0.402

Wilks lambda: 0.268 Canonical correlation 0.791 0.535
p-value: <0.001 Cumulative dispersion 0.806 1
N = 222

B. LDA2                    Dataset: MDB2

      Discriminant functions Standardized functions Scores of group means

F1 F2 F1 F2 F1 F2

Mg 0.627 0.477 Mg 0.814 0.619 BIG –1.553   0.831

Al –1.5 0.281 Al –1.493 0.279 HTP 1.03   0.068
Ti 0.259 –0.192 Ti 0.529 –0.392 OIG –1.064 –0.784
V –0.404 0.79 V –0.354 0.691
Mn 0.212 –0.439 Mn 0.275 –0.571
Co –0.541 0.215 Co –0.354 0.141 Statistics
Ni –0.567 –1.043 Ni –0.501 –0.92
Constant –14.699 –2.355 F1 F2

Eigenvalue 1.356 0.295

Canonical correlation 0.759 0.477
Wilks lambda: 0.328 Cumulative dispersion 0.821 1
p-value: <0.001
N = 253

Notes: Abbreviations: F1 = factor 1, F2 = factor 2; all elements in ppm; BIG = barren igneous, HTP = hydrothermal porphyry, OIG = ore-related igneous
LDA1
F1 = ln[0.703Mg/Fe - 1.287Al/Fe - 0.075V/Fe + 0.412Mn/Fe - 0.827Co/Fe - 0.306Ni/Fe] - 12.15
F2 = ln[0.384Mg/Fe + 0.091Al/Fe + 1.169V/Fe - 0.75Mn/Fe - 0.525Co/Fe - 1.149Ni/Fe] - 11.256
LDA2
F1 = ln[0.627Mg/Fe - 1.5Al/Fe + 0.259Ti/Fe - 0.404V/Fe + 0.212Mn/Fe - 0.541Co/Fe - 0.567Ni/Fe] - 14.699
F2 = ln[0.477Mg/Fe + 0.281Al/Fe - 0.192Ti/Fe + 0.79V/Fe - 0.439Mn/Fe + 0.215Co/Fe - 1.043Ni/Fe] - 2.355
The coefficients in the standardized functions are normalized to the variance of each element; Eigenvalues indicate the amount of variance explained by
each function; the canonical correlation is a measure of the correlation between the discriminant scores and the variables; cumulative dispersion is the
percentage of variance that is added by each function

magnetite from the ore-related and barren igneous groups,
with hydrothermal magnetite generally characterized by posi-
tive F1 values. Coefficients of the standardized discriminant
functions show the relative importance of each variable on the
respective function (Table 3). The greatest influence on F1
are Mg and Al, suggesting these two elements have the stron-
gest discriminating power for separating hydrothermal from
igneous magnetite. Factor 2 (F2) effectively separates ore and
barren igneous magnetite, with the latter tending toward posi-
tive F2 values.
Differences in the LDA1 and LDA2 models are mostly
affected by Mn, Ni, and V in F2, but with a smaller influence
of Mn and V in LDA2. Slight variation is expected between
the analyses because of the difference in variables selected.
Minor overlap is observed between compositional fields and
is slightly more pronounced in LDA2 (Fig. 13B). Neverthe-
less, the derived compositional fields from both models show
strong potential for identifying magnetite provenance in min-
eral indicator studies. Our approach shows a classification
accuracy of 91 and 78% for the LDA1 and LDA2 models,
respectively (Table 4), which suggests that this classification
method can be confidently applied to magnetite compositions
of unknown provenance as an exploration tool.
Application to the Mount Polley Porphyry
Cu-Au Deposit
Hydrothermal magnetite in till
The classification methods for magnetite composition derived
from discriminant analysis and binary plots discussed above
932 PISIAK ET AL.

5 5
(A) LDA1 BIG
3
(C) Tills - LDA1
OIG
3
HTP 1
BIG
1 -1
Factor 2

Factor 2
HTP
-3
-1 -5
-7
-3
-9 OIG
-5 -11
5 5
(D) Tills - LDA2
(B) LDA2
3
3
1 BIG
1
Factor 2

-1

Factor 2
-1
HTP
-3

-3 -5
OIG
-5 -7
-5 -3 -1 1 3 5 -7 -2 3
Factor 1 Factor 1
Fig. 13. Territorial plots of the discriminant scores for the full MDB1 and MDB2 training datasets for linear discriminant
analyses LDA1 (A) and LDA2 (B). Equations for Factors 1 and 2 are given in Table 3. Scores of group means indicated as
black diamonds in each respective field. Abbreviations: BIG = barren igneous, HTP = hydrothermal porphyry, OIG = ore-
related igneous. Panels (C) and (D) show territorial plots of the discriminant scores for magnetite of unknown origin in the
till samples using the linear functions of LDA1 and LDA2, respectively.

Table 4. Linear Discriminant Analysis Classification Summary for were directly tested using till samples surrounding the Mount
Individual Magnetite Groups Polley porphyry deposit as a ground-truth evaluation. For a
A. MDB1      Elements: Mg, Al, V, Mn, Co, and Ni
total of 20 till samples distributed upice, downice, and remote
from the deposit, 50 magnetite grains were analyzed in each
        Predicted group sample to determine if hydrothermal magnetite from the
deposit could be successfully identified in till (Fig. 2).
HTP OIG BIG Total % For each till sample, magnetite grains were assigned a dis-
Actual group HTP 30 3  1 34  88 criminant score using each of the two sets of linear functions
OIG  0 6  1  7  86 derived for LDA1 and LDA2. A territorial plot of F1 versus
BIG   0 0 15 15 100 F2 allowed for classification of magnetite grains into one of
56   91 the three predefined groups (Fig. 13C, D) and for the relative
percentage of hydrothermal grains in each till sample from
B. MDB2   Elements: Mg, Al, Ti, V, Mn, Co, and Ni the area (Table 5). The discriminant functions of LDA1 iden-
tified 71 hydrothermal porphyry magnetite grains in the area,
        Predicted group whereas those of LDA2 identified 93 grains. Most magnetite
grains from each classification plot within the ore igneous
HTP OIG BIG Total %
compositional field using both LDA1 and LDA2 functions,
Actual group HTP 50  8  2  60 83 suggesting that the discrimination of ore from barren igneous
OIG  4 19  2  25 76 magnetite is not robust.
BIG  1  5 10  16 63 Magnetite compositions from till were also plotted in the
101 78
discriminant diagrams proposed in the literature as described
above. The diagrams of Dupuis and Beaudoin (2011) were
Notes: Discriminant functions calculated on 65% of the MDB1 and MDB2
datasets and applied to the remaining 35% of the data; abbreviations: unable to correctly classify magnetite from Mount Polley and
BIG = barren igneous, HTP = hydrothermal porphyry, OIG = ore-related could not identify a sufficient number of porphyry magnetite
igneous grains in till (Figs. 9, 10). The Al + Mn versus Ti + V diagram
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 933

Table 5. Classification Results for Magnetite Grains in Till

A. LDA1 B. LDA2 C) Ti vs Ni/Cr diagram

Till sample n BIG OIG HTP % HTP BIG OIG HTP % HTP n IG HT % HT

12 PMA 586 A01 68 12 54 2 3 23 40 5 7 12 12 0 0

12 PMA 095 A01 30 1 27 2 7 9 19 2 7 30 27 3 10
12 PMA 083 A01 50 1 48 1 2 8 40 2 4 50 49 1 2
12 PMA 585 A01 50 9 41 0 0 16 33 1 2 50 49 1 2
12 PMA 081 A01 49 10 32 7 14 20 20 9 18 49 38 11 22
12 PMA 088 A01 50 2 47 1 2 15 31 4 8 50 44 6 12
12 PMA 077 A01 50 3 46 1 2 11 39 0 0 50 47 3 6
12 PMA 087 A01 50 3 47 0 0 14 36 0 0 50 47 3 6
12 PMA 080 A01 42 3 26 13 31 12 18 12 29 43 30 13 30
12 PMA 578 A01 50 11 37 2 4 16 30 4 8 50 44 6 12
12 PMA 575 A01 50 1 48 1 2 20 28 2 4 50 49 1 2
12 TFE 089 A01 50 6 42 2 4 16 32 2 4 50 45 5 10
12 PMA 101 A02 50 4 38 8 16 13 24 13 26 50 42 8 16
12 TFE 078 A01 50 13 36 1 2 28 22 0 0 50 48 2 4
12 PMA 094 A01 49 3 35 11 22 10 27 12 24 49 43 6 12
12 TFE 102 A01 49 8 35 6 12 21 24 4 8 49 43 6 12
12 PMA 098 A01 50 5 36 9 18 18 20 12 24 50 46 4 8
12 PMA 566 A01 44 4 39 1 2 13 29 2 5 47 43 4 9
12 PMA 071 A01 49 1 46 2 4 12 34 3 6 50 48 2 4
12 PMA 574 A01 49 3 45 1 2 21 24 4 8 50 42 8 16
Total 979 103 805 71 7 316 570 93 9 929 892 93 10

Notes: The number of grains is significantly reduced for Ti vs. Ni/Cr classification of sample 12 PMA 586 A01 since Cr was not selected for measurement
during one analytical run; abbreviations: n = total number of grains, BIG = barren igneous, HT = hydrothermal, HTP = hydrothermal porphyry, IG =
igneous, OIG = ore-related igneous

of Nadoll et al. (2015) was useful in distinguishing porphyry
from skarn magnetite, but with no distinction provided for
igneous magmatic magnetite (Fig. 11). The Ti versus Ni/Cr
diagram of Dare et al. (2014) correctly identified all magne-
tite samples from Mount Polley as hydrothermal. When the
discriminant diagram was applied to magnetite from till, 93
hydrothermal magnetite grains were identified, which is simi-
lar to the classification by the LDA2 model (Fig. 12, Table 5).
The proportion of hydrothermal grains estimated in each
till sample are generally consistent between the LDA1 and
LDA2 models, with the latter yielding slightly higher per-
centages in most samples (Fig. 14A). In contrast, there is less
agreement between the LDA models and the Ti-Ni/Cr dis-
crimination diagram (Fig. 14B, C). A significant number of
grains that were classified as hydrothermal by the Dare et al.
(2014) plot were not similarly identified as hydrothermal in
our discriminant analyses.
The textures of hydrothermal magnetite grains were exam-
ined to determine if they were potentially useful as a petro-
graphic discrimination tool. The proportion of hydrothermal

35
LDA1 - %HTP Grains
LDA2 - %HTP Grains

30 30
25
20 20
15
10 10
5 (A) (B)
0 0
0 5 10 15 20 25 30 35 35
LDA2 - %HTP Grains

LDA1 - %HTP Grains 30

25
20
15

Fig. 14. The percentage of hydrothermal porphyry magnetite grains iden- 10

tified in each till sample using the three different methods of classification. 5 (C)
The proportion of hydrothermal porphyry magnetite grains in till are gen-
erally consistent between the LDA1 and LDA2 models (A), whereas there 0
is less agreement between the LDA models and the Ti-Ni/Cr discrimina- 0 5 10 15 20 25 30 35
tion diagram (B, C). Ti-Ni/Cr diagram - %HTP Grains
934 PISIAK ET AL.

magnetite grains showing no exsolution varied with the classi-
fication method: 58% for LDA1 and 66% for LDA2. Although
hydrothermal magnetite typically contains relatively low Ti
contents, ilmenite exsolution lamellae in magnetite from
porphyry systems have been documented at several localities
(e.g., Celis et al., 2014; Nadoll et al., 2015; Canil et al., 2016).
Additionally, magmatic magnetite from rapidly cooled volca-
nic rocks is commonly free of exsolution (e.g., Grigsby, 1990;
Frost and Lindsley, 1991; Dare et al., 2014). The absence
of exsolution in magnetite is not a reliable indication of its
provenance from hydrothermal porphyry systems. No other
textural features such as the presence or type of inclusions,
degree if sphenitization, or martitization were linked to mag-
netite classification by discriminant model.
Mineral indicator maps for till surrounding Mount Polley
The percentage of hydrothermal magnetite grains in each till
sample surrounding the Mount Polley porphyry deposit as
determined by LDA1, LDA2, and the Ti-Ni/Cr discriminant
diagram (Table 5) is plotted in Figures 15 to 17. Samples hav-
ing populations of >10% hydrothermal magnetite grains were
deemed anomalies, whereas samples below this amount were
considered to be the regional background signal.
Classification of magnetite by the LDA1 discriminant func-
tions identified six till samples with anomalous concentrations
of hydrothermal magnetite grains (Table 5). Five of these
samples were similarly identified with the LDA2 classification
and occur along a trend that begins immediately southwest
of the deposit with the farthest anomalous sample approxi-
mately 4 km northwest of Mount Polley (Figs. 15, 16). To the
west, the anomaly is detected ~2.5 km from the deposit. Till
samples at 5 km and farther afar from the deposit contain
hydrothermal magnetite counts within the background range.
Northwest transport of hydrothermal magnetite from the
Mount Polley deposit is consistent with the dominant ice-
flow direction in the study area and documented using several

Fig. 15. The percentage of hydrothermal porphyry magnetite grains in the till samples as determined by the LDA1 model
plotted at each sample location. Symbol size is proportional to percentage of hydrothermal porphyry magnetite grains in the
sample. The Mount Polley deposit is underlain by the unit in pink and highlighted by the yellow star. Bedrock geology as in
Figure 2.
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 935
Fig. 16. The percentage of hydrothermal porphyry magnetite grains in the till samples as determined by the LDA2 model at
each sample location. Symbols as in Figure 15.
ore-related and pathfinder elements, and other ore and alter-
ation minerals (Hashmi et al., 2015). The transport of magne-
tite grains several kilometers west-southwest of the deposit is
also consistent with the observations of Hashmi et al. (2015),
who found chalcopyrite, gold, and epidote grains in till sam-
ples west-southwest of the deposit. This bidirectional trans-
port of material is characteristic of glacial dispersal, whereby
material was initially transported by the earlier ice-flow move-
ment to the west-southwest during glacial advancement from
the Cariboo Mountains, followed by reworking and transport
of debris by ice flow to the northwest during the glacial maxi-
mum (Parent et al., 1996).
The Ti versus Ni/Cr method of magnetite classification
(Fig. 12) identified the same quantity of hydrothermal mag-
netite grains as the LDA2 discriminant functions, but spread
over a larger number of till samples (Table 5). Using this dis-
crimination method, an apparent bidirectional detrital glacial
transport from the deposit to the west and northwest remains,
but the dispersal of hydrothermal magnetite grains over a
larger number of samples produces a more diffuse anomaly
compared to those from the LDA classifications (Fig. 17). The
four till samples with the greatest proportion of hydrothermal
magnetite grains identified by the Ti-Ni/Cr diagram occur on
a trend west from the deposit to a distance of 5 km, much
farther than that determined by the LDA classifications, and
by Hashmi et al. (2015) for chalcopyrite, gold, and epidote
in till. Hydrothermal magnetite grains would be primarily
transported to the northwest based on the dominant ice-flow
trajectory at Mount Polley. In contrast, classification using the
Ti-Ni/Cr diagram shows that hydrothermal magnetite from
the deposit has been transported greater distances to the west
than to the northwest, conflicting other glacial evidence from
the region (Hashmi et al., 2015).
All till samples with anomalous concentrations of hydro-
thermal magnetite grains can be directly traced to the Mount
Polley porphyry deposit based on ice-flow directions in the
study area, with the exception of one till sample that occurs
5 km north of the deposit (Figs. 15, 16). This sample (12 TFE
102 A01) contains 12% hydrothermal magnetite grains iden-
tified by both LDA1 and the Ti versus Ni/Cr discrimination
936 PISIAK ET AL.

Fig. 17. The percentage of hydrothermal porphyry magnetite grains in the till samples as determined by the Ti-Ni/Cr dis-
crimination diagram (Dare et al., 2014) plotted at each sample location. Symbols as in Figure 15.

methods (Table 5). Hydrothermal magnetite grains at this
location are not likely to be derived from the Mount Polley
deposit based on known ice-flow movements and the remote-
ness of the sample. However, the sample is located 1 km
south-east (upice) of an alkalic porphyry Cu-Au occurrence.
We suspect that other similar covered and unknown occur-
rences that host hydrothermal magnetite might be present in
the area and, therefore, may be prospective targets for further
exploration.
Summary and Further Work
Previous studies have shown potential for magnetite as an
indicator mineral in exploration of porphyry deposits but
there have been few if any published tests in the field. We
carried out such a test on magnetite in till sampled around
a known porphyry Cu-Au deposit at Mount Polley, where
the glacial history has been well studied (Plouffe et al., 2013;
Hashmi et al., 2015) and the composition of magnetite from
mineralized and intrusive lithologies at the deposit are known
(Canil et al., 2016). Laser ablation-ICP-MS was performed
on detrital magnetite grains (n = 985) from 20 till samples
collected over an area of ~900 km2 surrounding the Mount
Polley deposit. The LA-ICP-MS method, which employs a
large beam diameter and raster, homogenizes and integrates
magnetite compositions in grains showing exsolution, and
when calibrated using the “standardless” AYCF method is
an efficient protocol for analyzing large numbers of grains in
an indicator mineral exploration study, ensuring that all trace
elements (Mg, Ti, Al, V, Mn, Co and Ni) imperative for use
in the discriminant models for magnetite can be consistently
measured to low detection limits.
Previously proposed binary discriminant diagrams are not
optimal for distinguishing hydrothermal magnetite in an
exploration study. Caution must be taken when interpret-
ing magnetite chemistry using certain elements affected by
secondary processes such as sphenitization (for Si and Ca),
which was relatively common in detrital magnetite from till.
We instead employed linear discriminant analysis (LDA) to
successfully classify magnetite from various porphyry deposits
and barren igneous rocks using Mg, Ti, Al, V, Mn, Co, and Ni.
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 937
Anomalous values of hydrothermal magnetite were detected
in samples up to 4 km from the deposit using the LDA models
developed in this study, which is consistent with the ice-flow
history and known glacial transport direction of the region.
The success of the LDA models in the Mount Polley area
indicates that our approach has a proven potential in mineral
exploration in glacial drift for porphyry systems in glaciated
terrane. Further refinement of the LDA models will be pos-
sible with a larger compilation of new magnetite compositions
from a greater number of known deposits in other regions.
The compositional differences between ore-related and bar-
ren igneous magnetite needs further examination because
ore-related grains in till can provide an additional exploration
tool for porphyry systems where mineralization has not yet
been exposed at the surface.
In this study, the number of analyses per sample was arbi-
trarily set at 50. Analyzing fewer grains per till sample will
increase classification error, especially for distal samples
downice of a deposit where the number of hydrothermal mag-
netite grains in till becomes scarce. The optimum number of
grains per till sample to ensure that a target of viable size can
be generated requires further investigation.
Acknowledgments
We are grateful to J. Spence for assistance with LA ICPMS
analyses. We thank S. Dare and P. Nadoll for reviews and J.
Hanley and L. Meinert for editorial handling. Till collection
and sample processing was supported by Natural Resources of
Canada Targeted Geoscience Initiative 4 grants to AP and TF.
LP was supported by student scholarships from Geoscience
BC, the University of Victoria, and the Society for Economic
Geologists. Analytical work was supported by British Colum-
bia Ministry of Energy and Mines and NSERC of Canada
Discovery grants to DC.
REFERENCES
Aitchison, J., 1986, The statistical analysis of compositional data: New York,
Chapman and Hall, 416 p.
Averill, S.A., 2001, The application of heavy indicator mineralogy in mineral
exploration with emphasis on base metal indicators in glaciated metamor-
phic and plutonic terrains: Geological Society, London, Special Publication
185, p. 69–81.
Barnes, S.J., and Roeder, P.L., 2001, The range of spinel compositions in ter-
restrial mafic and ultramafic rocks: Journal of Petrology, v. 42, p. 2279–2302.
Boutroy, E., Dare, S.A.S., Beaudoin, G., Barnes, S.-J., and Lightfoot, P.C.,
2014, Magnetite composition in Ni-Cu-PGE deposits worldwide: Applica-
tion to mineral exploration: Journal of Geochemical Exploration, v. 145,
p. 64–81.
Buddington, A.F., and Lindsley, D.H., 1964, Iron-titanium oxide minerals
and synthetic equivalents: Journal of Petrology, v. 5, p. 310–357.
Canil, D., Grondahl, C., Lacourse, T., and Pisiak, L.K., 2016, Trace elements
in magnetite from porphyry Cu-Mo-Au deposits in British Columbia, Can-
ada: Ore Geology Reviews, v. 72, p. 1116–1128.
Celis, M.A., Bouzari, F., Bissig, T., Hart, C.J.R., and Ferbey, T., 2014, Petro-
graphic characteristics of porphyry indicator minerals from alkalic porphyry
copper-gold deposits in south-central British Columbia (NTS 092, 093):
Geoscience BC, Summary of Activities 2013, Report 2014-1, p. 53–62.
Chayes, F., 1960, On correlation between variables of constant sum: Journal
of Geophysical Research, v. 65, p. 4185–4193.
Chen, W.T., Zhou, M.-F., and Gao, J.-F., 2015, Geochemistry of magnetite
from Proterozoic Fe-Cu deposits in the Kangdian metallogenic prov-
ince, SW China: Mineralium Deposita, 15 p., http://dx.doi.org/10.1007/
s00126-014-0575-7.
Chou, I., and Eugster, H.P., 1977, Solubility of magnetite in supercritical
chloride solutions: American Journal of Science, v. 277, p. 1296–1314.
Colpron, M., and Nelson, J.L., 2011, A digital atlas of terranes for the north-
ern Cordillera: British Columbia Ministry of Energy and Mines, British
Columbia Geological Survey GeoFile 2011-11.
Dare, S.A.S., Barnes, S.J., and Beaudoin, G., 2012, Variation in trace element
content of magnetite crystallized from a fractionating sulfide liquid, Sud-
bury, Canada: Implications for provenance discrimination: Geochimica et
Cosmochimica Acta, v. 88, p. 27–50.
Dare, S.A.S., Barnes, S.J., Beaudoin, G., Méric, J., Boutroy, E., and Potvin-
Doucet, C., 2014, Trace elements in magnetite as petrogenetic indicators:
Mineralium Deposita, v. 49, p. 785–796.
Davis, J.C., 2002, Statistics and data analysis in geology, 3rd ed.: New York,
John Wiley & Sons, 638 p.
Dupuis, C., and Beaudoin, G., 2011, Discriminant diagrams for iron oxide
trace element fingerprinting of mineral deposit types: Mineralium Depos-
ita, v. 46, p. 319–335.
Ferbey, T., Plouffe, A., and Anderson, R.G., 2014, An integrated approach to
search for buried porphyry-style mineralization in central British Columbia
using geochemistry and mineralogy: Geological Survey of Canada, TGI-4
project, Current Research 2014-2, 12 p.
Fraser, T.M., Godwin, C.I., Thompson, J.F.H., and Stanley, C.R., 1992, Geol-
ogy and alteration of the Mount Polley alkalic porphyry copper-gold deposit,
British Columbia: BC Ministry of Energy and Mines, Geological Fieldwork
1992, British Columbia Geological Survey Paper 1993-1, p. 295–300.
Frost, B.R., and Lindsley, D.H., 1991, Occurrence of iron-titanium oxides in
igneous rocks, in Lindsley, D.H., ed., Oxide minerals: Petrologic and mag-
netic significance: Reviews in Mineralogy, v. 25, p. 433–468.
Grigsby, J.D., 1990, Detrital magnetite as a provenance indicator: Journal of
Sedimentary Petrology, v. 60, p. 940–951.
Haggerty, S.E., 1976, Opaque mineral oxides in terrestrial igneous rocks:
Reviews in Mineralogy, v. 3, p. HG101–HG300.
——1991, Oxide textures—a mini-atlas: Reviews in Mineralogy, v. 25,
p. 129–137.
Hashmi, S., Ward, B.C., Plouffe, A., Leybourne, M.I., and Ferbey, T., 2015,
Geochemical and mineralogical dispersal in till from the Mount Polley
Cu-Au porphyry deposit, central British Columbia, Canada: Geochemistry:
Exploration, Environment, Analysis, v. 15, p. 234–249.
Hu, H., Lentz, D., Li, J.-W., McCarron, T., Zhao, X.-F., and Hall, D., 2015,
Re-equilibration processes in magnetite from iron skarn deposits: Eco-
nomic Geology, v. 110, p. 1–8.
Huang, X., Qi, L., and Meng, Y., 2014, Trace element geochemistry of mag-
netite from the Fe(-Cu) deposits in the Hami region, eastern Tianshan oro-
genic belt, NW China: Acta Geologica Sinica, v. 88, p. 176–195.
Ilton, E.S., and Eugster, H.P., 1989, Base metal exchange between magnetite
and a chloride rich hydrothermal fluid: Geochimica et Cosmochimica Acta,
v. 53, p. 291–301.
Irvine, T.N., 1965, Chromian spinel as a petrogenetic indicator. Pt. 1: Theory:
Canadian Journal of Earth Sciences, v. 2, p. 648–672.
Kamvong, T., Zaw, K., and Siegele, R., 2007, PIXE/PIGE microanalysis of
trace elements in hydrothermal magnetite and exploration significance: A
pilot study: Australian Conference on Nuclear and Complementary Tech-
niques of Analysis, 15th, and Vacuum Society of Australia Congress, 9th,
Melbourne, Australia, University of Melbourne, 4 p.
Kelley, K.D., Eppinger, R.G., Lang, J., Smith, S.M., and Fey, D.L., 2011,
Porphyry Cu indicator minerals in till as an exploration tool: Example from
the giant Pebble porphyry Cu-Au-Mo deposit, Alaska, USA: Geochemistry:
Exploration, Environment, Analysis, v. 11, p. 321–334.
Liu, Y., Hu, Z., Gao, S., Günther, D., Xu, J., Gao, C., and Chen, H., 2008,
In situ analysis of major and trace elements of anhydrous minerals by LA-
ICP-MS without applying an internal standard: Chemical Geology, v. 257,
p. 34–43.
Logan, J.M., and Mihalynuk, M.G., 2005, Regional geology and setting of
the Cariboo, Bell, Springer and Northeast porphyry Cu-Au zones at Mount
Polley, south-central British Columbia: BC Ministry of Energy and Mines,
Geological Fieldwork 2004, British Columbia Geological Survey Paper
2005-1, p. 249–270.
Logan, J.M., Mihalynuk, M.G., Ullrich, T., and Friedman, R.M., 2007, U-Pb
ages of intrusive rocks and 40Ar/39Ar plateau ages of copper-gold-silver min-
eralization associated with alkaline intrusive centres at Mount Polley and
the Iron Mask batholith, southern and central British Columbia: BC Min-
istry of Energy and Mines, Geological Fieldwork 2006, British Columbia
Geological Survey Paper 2007-1, p. 93–116.
Longerich, H., Jackson, S.E., and Günther, D., 1996, Laser ablation induc-
tively coupled plasma mass spectrometric transient signal data acquisition
938 PISIAK ET AL.
and analyte concentration calculation: Journal of Analytical Atomic Spec-
trometry, v. 11, p. 899–904.
Makvandi, S., Ghasemzadeh-Barvarz, M., Beaudoin, G., Grunsky, E.C.,
McClenaghan, M.B., and Duchesne, C., 2016, Principal component analy-
sis of magnetite composition from volcanogenic massive sulfide deposits:
Case studies from the Izok Lake (Nunavut, Canada) and Halfmile Lake
(New Brunswick, Canada) deposits: Ore Geology Reviews, v. 72, p. 60–85.
McMartin, I., Corriveau, L., and Beaudoin, G., 2011, An orientation study
of the heavy mineral signature of the NICO Co-Au-Bi deposit, Great Bear
magmatic zone, Northwest Territories, Canada: Geochemistry: Explora-
tion, Environment, Analysis, v. 11, p. 293–307.
McQueen, K.G., and Cross, A.J., 1998, Magnetite as a geochemical sampling
medium: Application to skarn deposits: The state of the regolith: Geological
Society of Australia Special Publication, v. 20, p. 194–199.
Mellini, M., Rumori, C., and Viti, C., 2005, Hydrothermally reset magmatic
spinels in retrograde serpentinites: Formation of “ferritchromit” rims
and chlorite aureoles: Contributions to Mineralogy and Petrology, v. 149,
p. 266–275.
Mortensen, J.K., Ghosh, D.K., and Ferri, F., 1995, U-Pb geochronology of
intrusive rocks associated with copper-gold porphyry deposits in the Cana-
dian Cordillera: CIM Special Volume 46, p. 142–158.
Mountjoy, D., 2011, Granitoid-hosted magnetite as an indicator mineral for
porphyry Cu ore deposits: B.Sc. thesis, Victoria, Canada, University of Vic-
toria, 41 p.
Nadoll, P., and Koenig, A.E., 2011, LA-ICP-MS of magnetite: Methods
and reference materials: Journal of Analytical Atomic Spectrometry, v. 26,
p. 1872–1877.
Nadoll, P., Mauk, J.L., Hayes, T.S., Koenig, A.E., and Box, S.E., 2012, Geo-
chemistry of magnetite from hydrothermal ore deposits and host rocks of
the Mesoproterozoic Belt Supergroup, United States: Economic Geology,
v. 107, p. 1275–1292.
Nadoll, P., Angerer, T., Mauk, J.L., French, D., and Walshe, J., 2014, The
chemistry of hydrothermal magnetite: A review: Ore Geology Reviews,
v. 61, p. 1–32.
Nadoll, P., Mauk, J.L., Leveille, R.A., and Koenig, A.E., 2015, Geochemis-
try of magnetite from porphyry Cu and skarn deposits in the southwestern
United States: Mineralium Deposita, v. 50, p. 493–515.
Newberry, N.G., Peacor, D.R., Essene, E.J., and Geissman, J.W., 1982,
Silicon in magnetite: High resolution microanalysis of magnetite-ilmenite
intergrowths: Contributions to Mineralogy and Petrology, v. 80, p. 334–340.
O’Neill, H.St.C., and Navrotsky, A., 1984, Cation distributions and thermo-
dynamic properties of binary spinel solid solutions: American Mineralogist,
v. 69, p. 733–753.
Parent, M., Paradis, S.J., and Doiron, A., 1996, Palimpsest glacial dispersal
trains and their significance for drift prospecting: Journal of Geochemical
Exploration, v. 56, p. 123–140.
Pisiak, L.K., 2015, Magnetite as a porphyry Cu indicator mineral: A test using
the Mount Polley porphyry Cu-Au deposit south-central British Columbia:
M.Sc. thesis, University of Victoria, Victoria, Canada, 149 p.
Pisiak, L.K., Canil, D., Grondahl, C., Plouffe, A., Ferbey, T., and Anderson,
R.G., 2015, Magnetite as a porphyry copper indicator mineral in till: A test
using the Mount Polley porphyry copper-gold deposit, south-central Brit-
ish Columbia (NTS 093A): Geoscience BC, Summary of Activities 2014,
Report 2015-1, p. 141–150.
Plouffe, A., Ferbey, T., Anderson, R.G., Hashmi, S., and Ward, B.C., 2013a,
New TGI-4 till geochemistry and mineralogy results near the Highland Val-
ley, Gibraltar, and Mount Polley mines, and Woodjam district: An aid to
search for buried porphyry deposits: Geological Survey of Canada Open
File 7473, 58 p.
Plouffe, A., Ferbey, T., Anderson, R.G., Hashmi, S., Ward, B.C., and Sacco,
D.A., 2013b, The use of till geochemistry and mineralogy to explore for
buried porphyry deposits in the Cordillera—preliminary results from a
TGI-4 intrusion-related ore systems project: Geological Survey of Canada
Open File 7367, poster.
Razjigaeva, N.G., and Naumova, V.V., 1992, Trace element composition of
detrital magnetite from coastal sediments of northwestern Japan Sea for
provenance study: Journal of Sedimentary Petrology, v. 62, p. 802–809.
Rees, C., 2013, The Mount Polley Cu-Au porphyry deposit, south-central
British Columbia, Canada: Society of Economic Geologists Guidebook
Series, v. 43, p. 67–98.
Reimann, C., and Filzmoser, P., 2000, Normal and lognormal data distribu-
tion in geochemistry: Death of a myth. Consequences for the statistical
treatment of geochemical and environmental data: Environmental Geol-
ogy, v. 39, p. 1001–1014.
Roeder, P.L., 1994, Chromite: From the fiery rain of chondrules to the
Kilauea Iki lava lake: Canadian Mineralogist, v. 32, p. 729–746.
Ryder, J.M., Fulton, R.J., and Clague, J.J., 1991, The Cordilleran Ice Sheet
and the glacial geomorphology of southern and central British Columbia:
Géographie Physique et Quaternaire, v. 45, p. 365–377.
Sack, R.O., and Ghiorso, M.S., 1991, Chromian spinels as petrogenetic indi-
cators: Thermodynamics and petrological applications: American Mineralo-
gist, v. 76, p. 827–847.
Sappin, A.-A., Dupuis, C., Beaudoin, G., Pozza, M., McMartin, I., and
McClenaghan, M.B., 2014, Optimal ferromagnetic fraction in till samples
along ice-flow paths: Case studies from the Sue-Dianne and Thompson
deposits, Canada: Geochemistry: Exploration, Environment, Analysis, v. 14,
p. 315–329.
Shau, Y.-H., Torii, M., Horng, C.-S., and Peacor, D.R., 2000, Subsolidus evo-
lution and alteration of titanomagnetite in ocean ridge basalt from Deep
Sea Drilling Project/Ocean Drilling Program Hole 504B, Leg 83: Impli-
cations for the timing of magnetization: Journal of Geophysical Research,
v. 105, p. B23,635–B23,649.
Shcheka, S.A., Romanenko, I.M., and Chubarov, V.M., 1977, Silica-bearing
magnetites: Contributions to Mineralogy and Petrology, v. 63, p. 103–111.
Shiga, Y., 1988, Silician magnetite from the Kamaishe mine, Japan: Mining
Geology, v. 38, p. 437–440.
Simon, A.C., Pettke, T., Candela, P.A., Piccoli, P.M., and Heinrich, C.A.,
2004, Magnetite solubility and iron transport in magmatic-hydrothermal
environments: Geochimica et Cosmochimica Acta, v. 68, p. 4905–4914.
Thompson, B., 1995, Stepwise regression and stepwise discriminant analysis
need not apply here: A guidelines editorial: Educational and Psychological
Measurement, v. 55, p. 525–534.
Turnock, A.C., and Eugster, H.P., 1962, Fe-Al oxides: Phase relationships
below 1000°C: Journal of Petrology, v. 3, p. 533–565.
Vermeesch, P., 2006, Tectonic discrimination diagrams revisited: Geo-
chemistry, Geophysics, Geosystems, v. 7, no. 6, 55 p., http://dx.doi.
org/10.1029/2005GC001092.
Vincent, E.A., and Phillips, R., 1954, Iron-titanium oxide minerals in layered
gabbros of the Skaergaard intrusion, East Greenland: Part I. Chemistry and
ore-microscopy: Geochimica et Cosmochimica Acta, v. 6, p. 1–26.
Weisse, P.A., Haggerty, S.E., and Brown, L.L., 1985, Paleomagnetism and
magnetic mineralogy of the Nahant gabbro and tonalite, eastern Massachu-
setts: Canadian Journal of Earth Sciences, v. 22, p. 1425–1435.
Westendorp, R.W., Watkinson, D.H., and Jonasson, I.R., 1991, Silicon-bear-
ing zoned magnetite crystals and the evolution of hydrothermal fluids at
the Ansil Cu-Zn mine, Rouyn-Noranda, Quebec: Economic Geology, v. 86,
p. 1110–1114.
Whalen, J.B., and Chappell, B.W., 1988, Opaque mineralogy and mafic min-
eral chemistry of I- and S-type granites of the Lachlan fold belt, southeast
Australia: American Mineralogist, v. 73, p. 281–296.
Yang, S., Wang, Z., Guo, Y., Li, C., and Cai, J., 2009, Heavy mineral composi-
tions of the Changjiang (Yangtze River) sediments and their provenance-
tracing implication: Journal of Asian Earth Sciences, v. 35, p. 56–65.
 MAGNETITE AS AN INDICATOR MINERAL, MT. POLLEY Cu-Au DEPOSIT, BC 939
APPENDIX1
Detailed Procedures for Discriminant Analyses of Magnetite
We compiled a database (n = 493) of trace element concen-
trations in magnetite to use as a training dataset for linear
discriminant analyses (Table 2). The first dataset (MDB1, n
= 222) consists of magnetite compositions from Canil et al.
(2016) and Mountjoy (2011) for six porphyry Cu(-Au-Mo)
deposits and four barren igneous intrusions in the Canadian
Cordillera. All data for MDB1 were obtained in the same LA-
ICP-MS laboratory, thus providing an internally consistent
dataset. The second dataset (MDB2, n = 271) combines the
aforementioned data with those of Nadoll et al. (2015) from
four porphyry Cu deposits, one Climax-Mo deposit (south-
west United States), and one barren igneous intrusion (Japan),
all measured at different LA-ICP-MS laboratories. The larger
MDB2 dataset is less internally consistent, but is a more com-
prehensive representation of porphyry magnetite composi-
tion from a large number of deposits, rocks and locations.
Magnetite in both datasets was classified as either (1)
hydrothermal porphyry magnetite by occurring in veins and
hydrothermal breccias, showing direct petrographic evidence
of precipitation from fluids; (2) ore-related igneous magne-
tite as an igneous mineral in fresh or altered intrusive rocks
in a porphyry deposit; or (3) barren igneous magnetite as an
igneous mineral from unmineralized intrusions. Summary sta-
tistics for the compositional groupings for magnetite in the
MDB1 and MDB2 datasets are given in Table A2-1 and A2-2,
respectively.
Discriminant analysis requires values for all selected ele-
ments of interest, thus only those elements that are common
between all the magnetite data sources were included in the
analyses. Magnetite from skarn deposits was available for
compilation of MDB1 and MDB2, but was excluded because
it often contains trace element abundances that are below
detection/reporting limits (Nadoll et al., 2012, 2014, 2015;
Canil et al., 2016). For the internally consistent MDB1 data-
set, only elements Mg, Al, Ti, V, Mn, Co, Ni, and Sn were
selected as variables, because Cu, Nb, Mo, and Sn were only
measured in 60 to 80% of the analyses. For the MDB2 data-
set, a larger number of analyses have elements that are below
detection, due to the higher limit of reporting (10σ of the
background) criterion used by Nadoll et al. (2015), compared
to the detection limit (3σ of the background) criterion we
apply. To maximize the number of samples from the Nadoll
et al. (2015) dataset, only elements Mg, Al, Ti, V, Mn, Co,
and Ni could be included in the discriminant analyses (Sn was
excluded). Censored values were not used to substitute for
any elements that are below detection.
Discriminant analysis assumes that each variable is normally
distributed, but geochemical data are commonly skewed and/
or contain outliers (Reimann and Filzmoser, 2000). Addi-
tionally, compositional data are “closed” to a constant sum of
100%, which conflicts with the assumption of variable inde-
pendence (Chayes, 1960; Vermeesch, 2006). The problem of
closure is minimal for trace elements in magnetite because
they comprise only a small portion of the bulk composition
1 Appendix
Tables A1 and A2 are available online.
(which is mostly Fe). Nonetheless, to minimize deviations
from normality and prevent the effects of data closure, each
element (xi) was transformed using the additive log-ratio
method of Aitchison (1986). This method transforms D-part
compositional data into (D – 1)-dimensional Euclidean space
to which multivariate statistical methods can be applied. Iron
by EMP analysis was chosen as the component divisor due
to its consistently high concentration and minimal variation
(<10% relative) in magnetite. Transformed data for magnetite
was imported into SYSTAT 13, a statistical software package,
to perform the discriminant analysis.
The linear method of discriminant analysis (LDA) was cho-
sen due to the simplicity of linear functions as compared to
quadratic discriminant analysis (QDA), which is more com-
plex and contains a larger number of parameters. In LDA,
equal covariance between groups is required to produce reli-
able discriminant functions. Although QDA is the preferred
method for datasets with unequal covariance, LDA is a very
close approximation and can perform better on independent
test data (Vermeesch, 2006). As a comparison, LDA and QDA
were carried out on both datasets and the discriminant models
were tested based on their predictive power for independent
test data. There was no significant improvement using QDA.
To assess the capability of LDA for distinguishing magne-
tite composition, the MDB1 and MDB2 datasets were each
subdivided into two parts: (1) a random 65% of the data used
as “training” for the discriminant analysis, and (2) the remain-
ing 35% left as independent “test” data. First, the optimum
set of elements was determined because some may not sig-
nificantly contribute to the discriminating power of the func-
tions and can, therefore, be eliminated from the analysis. The
stepwise method of discriminant analysis was used to incre-
mentally add elements into the discriminant analysis based on
their contribution to the model as indicated by the associated
change in the Wilks l parameter, which approaches zero with
increasing power for the chosen element(s). Decreasing the
number of variables in the discriminant analysis will naturally
increase Wilks l. Table A2-3 lists Wilks l values associated
with each element ranked in order of discriminating power
by forward-stepping LDA using the training datasets. At 0.05
level of significance, all eight elements (Mg, Al, Ti, V, Mn, Co,
Ni, and Sn) were statistically significant in the MDB1 training
data, whereas, only Al, Mg, Ni, and Co were significant in the
MDB2 training data.
The stepwise method is commonly used to optimize variable
selection, but it is not always the best method for selecting the
most useful subset of variables (Thompson, 1995). Alterna-
tively, multiple iterations of discriminant analyses can be per-
formed where each element is manually removed in turn from
the analysis and each discriminant model is evaluated on the
ability to correctly classify samples from an independent test
dataset. Classification results on the “test” data using various
manually chosen element subsets for LDA compared rankings
of the percent classification accuracy and the associated Wilks
l. For MDB1, a classification accuracy of 91% was achieved
when Ti and Sn were excluded from the analysis (Table A2-4).
940 PISIAK ET AL.
This is contrary to stepwise LDA, which determined that all
eight elements (Mg, Al, Ti, V, Mn, Co, Ni, and Sn) were signif-
icant in the model. Similarly, stepwise LDA determined that
Mg, Al, Co, and Ni were significant for the MDB2 training
data, but the derived functions using only these four elements
performed relatively poorly on the test data, with 66% accu-
racy. In contrast, 78% of magnetite samples were classified
accurately when using all seven elements (Mg, Al, Ti, V, Mn,
Co, and Ni; Table A2-4). The stepwise method is not ideal for
variable selection in this case. Therefore, only the elements
with the greatest predictive power for the independent test
data were selected for further analysis. Table 3 gives the equa-
tions for the discriminant functions (F1, F2) and Table 4 is a
detailed summary of the classification for the magnetite com-
positional groups using the preferred set of elements for each
dataset. The equations for the field boundaries separating
the magnetite groups (hydrothermal porphyry magnetite, ore
igneous magnetite, and barren igneous magnetite) are given
in Table A2-5.