Spectral characteristics of propylitic alteration minerals as a vectoring tool for porphyry copper deposits

Short-­‐wave	   infrared	  (SWIR)	  reflectance	  spectroscopy	   is	  a	  quick	  and	  effective	  method	  of	  detecting and	  characterising	  hydrothermal	  alteration	  associated	  with	  ore	  deposits,	  and	  can	  identify	  not	  only mineral	  species	  but	  also	  changes	  in	  the	  major	  element	  composition	  of	  minerals.	  Porphyry	  deposits represent	  large	  accumulations	  of	  valuable	  metal	  in	  the	  Earth’s	  crust	  and	  have	  extensive	  alteration signatures	  making	  them	  an	  attractive	  target	  for	  exploration,	  particularly	  by	  remote	  sensing	  which can	  cover	  large	  areas	  quickly.	  Reflectance	  spectroscopy	  has	  been	  widely	  applied	  in	  sericitic	  (phyllic), argillic	  and	  advanced	  argillic	  alteration	  domains	  because	  it	  is	  particularly	  effective	  in	  discriminating bright	  clay	  minerals.	  However,	   the	  propylitic	  domain	  has	   remained	  relatively	  unexplored	  because propylitic	  rocks	  are	  typically	  dark	  and	  produce	  relatively	  poorly-­‐defined	  spectra. This	  study	  utilised	  an	  ASD	  TerraSpec	  4	  handheld	  spectrometer	  to	  collect	  SWIR	  spectra	  from	  rocks surrounding	   the	   Batu	   Hijau	   Cu-­‐Au	   porphyry	   deposit	   in	   Indonesia,	   where	   previous	   work	   has identified	  systematic	  spatial	  variations	  in	  the	  chemistry	  of	  chlorite,	  a	  common	  propylitic	  alteration mineral.	  Spectra	  were	  collected	  from	  90	  samples	  and	  processed	  using	  The	  Spectral	  Geologist	  (TSG) software	  as	  well	  as	  the	  Halo	  mineral	  identifier	  to	  characterise	  mineralogy	  and	  extract	  the	  positions and	   depths	   of	   spectral	   absorption	   features,	   which	   were	   then	   correlated	   with	   major	   element geochemistry.	  Two	  diagnostic	  chlorite	  absorption	  features	  located	  at	  around	  2250	  nm	  and	  2340	  nm correlate	  with	  the	  Mg#	  (Mg/[Mg+Fe])	  of	  chlorite,	  both	  in	  terms	  of	  wavelength	  position	  and	  depth. As	   the	  Mg#	   increases,	   the	  wavelengths	  of	  both	   features	   increase	   from	  2249	  nm	  to	  2254	  nm	  and from	  2332	  nm	   to	   2343	  nm	   respectively,	   and	   absorption	  depths	   also	   increase	   significantly.	   In	   the spatial	   dimension,	   these	   feature	   variations	   act	   as	   reasonably	   strong	   vectors	   to	   the	   orebody, showing	  systematic	  increases	  over	  a	  transect	  away	  from	  the	  porphyry	  centre,	  peaking	  at	  distances of	   around	   1.6	   km,	   which	   matches	   the	   spatial	   trend	   displayed	   by	   Mg#,	   as	   well	   as	   various	   trace element	   indicators	   in	   chlorite.	   The	   hull	   slope	   in	   spectra	   between	   1400	   nm	   and	   1900	   nm	   is	   also shown	  to	  increase	  with	  Mg#,	  and	  the	  position	  of	  an	  absorption	  feature	  at	  1400	  nm	  increases	  with the	  Al:Si	  ratio,	  a	  parameter	  that	  also	  tends	  to	  increase	  with	  proximity	  to	  porphyry	  deposits. Feature	   depth	   variations	   in	   particular	   appear	   to	   represent	   a	   new	   finding	   in	   chlorite	   reflectance spectroscopy;	   however,	   the	   causes	   are	   not	   entirely	   clear	   and	   require	   further	   investigation. Nonetheless,	   the	   systematic	   behaviour	   provides	   a	   potentially	   useful	   new	   tool	   for	   exploration	   in propylitic	  alteration	  zones.


Introduction
Porphyry deposits represent some of the largest accumulations of metal in the Earth's crust and are the primary source of the world's Cu and Mo, and an important source of Au and other metals (Sillitoe 2010). Deposits are formed as a result of hydrothermal fluids that exsolve from intrusive magmatic bodies, precipitating metals into the surrounding rocks. The outward movement of these hydrothermal fluids creates distinct alteration zones recognized by the occurrence of specific mineral assemblages (Cooke et al. 2014a;Sillitoe 2010). The propylitic alteration zone represents the most distal signature of mineralisation, detectable kilometres away from the main orebody (Cooke et al. 2014a) and, as such, is an important target for exploration. Alteration zoning is not only expressed in mineralogy, but in whole--rock and mineral geochemical trends which can act as vectors towards orebodies (e.g. Emmons 1927;John 1978;Norman et al. 1991;Cooke et al. 2014b;Wilkinson et al. 2015).
Remote sensing is a valuable tool in mineral exploration, providing a quick way to identify and map hydrothermal alteration products over large areas. Much work has been done on characterising the hydrothermal alteration associated with ore deposits in terms of spectral signatures (Yang & Huntingdon 1996;Herrmann et al. 2001;Sun et al. 2001;Jones et al. 2005) including porphyry deposits specifically (Cudahy et al. 2001;Chang et al. 2011;Dilles 2012;Zadeh et al. 2014;Halley et al. 2015). However the focus of these studies is on the clay--dominated, argillic and sericitic alteration assemblages. In relation to a particular exploration application, Chang et al. (2011) showed that it is possible to use spectral features of alunite in the advanced--argillic alteration zone as pointers towards porphyry ore deposits.
Investigations into the SWIR spectroscopy of chlorite and epidote, two of the most prevalent propylitic alteration minerals are few. However, sensitivity of SWIR spectra to mineral chemical changes that are recognisable both in the laboratory (King & Clark 1989;Liebscher 2004;Bishop et al. 2008) and in hyperspectral imaging (Cudahy et al. 2001;Roache et al. 2011) have been recognised. This indicates that spectral characteristics might be able to act as a vector to orebodies in the propylitic domain as well as the more proximal alteration zones.
The large footprint of propylitic alteration should present an attractive target for remotely sensed mineral exploration. However, the rocks that typify propylitic alteration zones are often dark (low reflectance) making their characterisation exceedingly difficult. Fortunately, the availability of increasingly sensitive portable field spectrometers has opened up the possibility of tackling this problem. These spectrometers are an ideal tool for detailed investigations into the spectral features of rocks; they can produce high--resolution spectra, free from the effects of atmospheric scattering and absorption, and have absorption features clearly detectable even at reflectance values as low as 1%.
This study utilised an ASD TerraSpec 4 spectrometer to collect infrared spectra of rocks from the Batu Hijau porphyry copper system in Indonesia. It builds on previous research done as part of the AMIRA P765A research project (AMIRA International Ltd) which included studies of spatial variation in the geochemistry of chlorite (Wilkinson et al. 2015) and epidote (Cooke at al. 2014b). Most importantly, the work utilised the same samples analysed by Wilkinson et al. (2015) so that extensive prior knowledge of whole--rock geochemistry and mineral chemistry was available. The primary aims were to characterise the infrared spectra of propylitic rocks throughout the Batu Hijau alteration footprint, and to identify any spectral features that might reflect systematic geochemical variation in the spatial domain which could therefore act as vectors towards economic mineralisation. Longer term, it is hoped that such features could be targeted by airborne hyperspectral or even satellite remote sensors, such as ASTER or WorldView--3, allowing a relatively quick way to search large areas of propylitic alteration for contained porphyry deposits.

Background Geology
Batu Hijau is a giant Cu--Au porphyry deposit located in the south--western part of Sumbawa island, Indonesia (Fig. 1). It is situated in the Sunda--Banda Arc which is host to numerous magmatic-hydrothermal deposits (Garwin et al. 2005), related to calc--alkaline magmatism resulting from subduction of the Indo--Australian Plate beneath the Eurasian plate. Prior to the start of open pit mining, Batu Hijau contained an estimated 1644 Mt of ore with copper and gold grades of 0.44% and 0.35 g/t respectively (Cooke et al. 2005).
The Batu Hijau district is largely made up of an Early to Middle Miocene volcaniclastic sequence which comprises volcanic lithic breccias and volcanic sandstones with local limestone layers. This is cut by several intrusive phases dating from the Early to Middle Miocene onwards which include andesites and andesite porphyries, quartz diorites, and porphyritic tonalites (Garwin 2002). Mineralisation is strongly associated with tonalite porphyries in four main centres: Batu Hijau, Sekongkang, Arung Ara and Katala (Fig.  2). Batu Hijau is significantly larger and higher grade than the rest (Garwin 2002). At least three intrusive episodes occurred in the Batu Hijau tonalite porphyry complex at the core of the system, with mineralisation and alteration intensity decreasing with each subsequent intrusion (Clode et al. 1999).

Methodology
Short--wave infrared (SWIR) spectra were collected for 90 samples from the Batu Hijau deposit (Table 2). Samples were scanned in dark laboratory conditions using an ASD TerraSpec 4 reflectance spectrometer with the Hi--Brite Contact Probe accessory that acts as an illumination source and collects reflected light. Technical specifications are outlined in Table  1. Samples consisted of resin-mounted, polished rock slices, providing an ideal flat surface for the contact probe. Spectral files were created as an average of 3 scans at different points on each sample. Double the recommended scan time was used (10 seconds, following the recommendation of Chang and Yang (2012) for dark, low reflectance rocks) to increase signal--to--noise ratios. Two sets of spectral files were generated via this process for each sample, acquired on different days, to assess the impact of differences in ambient lighting. In general, there is excellent agreement between the replicates (Table 2).
Spectra were primarily analysed using The Spectral Geologist (TSG) software, and selected for further investigation based on classification mineral identification by The Spectral Assistant (TSA), a matching algorithm which identifies a linear mixture of two minerals in the library that best match the spectrum (CSIRO 2010). Only spectra determined to contain either chlorite or epidote as the principal SWIR--active mineral phase (Table 2) were used in the subsequent analysis. Wavelength positions and depths of features were calculated from the two average spectra for each individual sample. The average of these results are given in Table 3. These values were extracted from the profiles of hull--quotient corrected spectra; this correction accentuates absorption features by removing broad background variation (Fig. 3).
Analysis of the spectra was also undertaken by ASD Inc. (PANalytical) utilising the Halo mineral identifier in order to compare results obtained using a different matching algorithm. The considerably larger library of mineral spectra that Halo incorporates also enabled the identification of additional minerals that TSG could not.
Geochemical data were previously collected as part of the AMIRA P765A project. These data include whole--rock geochemistry, electron microprobe analyses (EMP) of chlorite and epidote grains, and laser ablation inductively--coupled--plasma mass spectrometry (LA--ICP--MS) analyses of chlorite grains (Table  3). EMP data were primarily used to define the major element compositions of minerals; this was acquired on multiple spots on multiple grains in each sample, with an average of 10 measurements per sample. These were validated based on stoichiometry prior to inclusion in the database. Chlorite LA--ICP--MS data were used to define major element chemistry in four samples where EMP data were not available. Outliers in the EMP data were identified as any measurements where at least one major element (Fe, Al, Si, O, Mg, Ca) fell outside the bounds of the mean of the dataset ± 2σ for each mineral. Multiple spot analyses for each sample were averaged and ± 1σ standard error used as a practical measure of combined analytical uncertainty and natural, within--sample variability.

Spectrally-determined mineralogy
The Spectral Assistant identified chlorite and epidote as the most abundant SWIR--active mineral groups in the samples, with white--mica and carbonates also making up significant contributions (Figs. 4A, 5A, 5B). Chlorite, where identified, was almost always the primary mineral. Epidote was common as the second most abundant SWIR--active mineral in chlorite--dominated samples, and in some as the primary mineral. The results from Halo (Figs. 4B, 5C) were generally similar, but identified clays (primarily smectites and vermiculites) as dominant in the biotite and actinolite zones, with chlorite commonly attributed as the secondary mineral. Although the identification of these phases should be treated with caution, their presence would be consistent with the later argillic overprint that has been well mapped in the proximal parts of the alteration system (Fig.  2). Halo also classified many spectra that TSA could not, with most being zeolite--dominant, particularly in the very distal samples. Minerals contained in each group are shown in Table 4.
Of the 90 samples, 46 produced spectra with the primary mineral identified by TSA as either chlorite or epidote (Fig.  5D). Of this subset, 30 were attributed to containing primarily chlorite, 10 to containing primarily epidote, and 6 where it was unclear (likely containing roughly equal amounts of each). Separate spectra collected on different areas of the same sample were very consistent ( Fig. 6) indicating that the SWIR response is reproducible and varies more between samples (as a function of spatial position and mineral abundance) than within samples.

Chlorite
Chlorite chemistry from throughout the propylitic zones at Batu Hijau shows significant variation in terms of Fe and Mg which have a strong inverse correlation (Fig. 7A) related to the well--established solid solution between the iron and magnesium end--members (Deer et al. 2009). The Mg/(Mg+Fe) mass ratio (Mg#) ranges from 0.30 to 0.62 which translates to Mg/Fe molar ratios between 0.78 and 0.51 with a mean of 0.68±0.01 (1σ). Comparison of chlorite Mg# to the whole--rock Mg# shows no correlation (Fig. 7B) suggesting a lack of protolith control of chlorite composition. Si content shows an inverse correlation with Al ( Fig.  7C), probably as a result of the Tschermak substitution: Deer et al. 2009). Ca is also strongly correlated with Si and inversely correlated with Al (Fig. 7D).
Chlorite--dominated spectra are characterised by two key absorption features centred around 2250 nm and 2340 nm (Fig.  8), caused by Fe--OH and Mg--OH bond stretching respectively (Herrmann et al. 2001). A third feature occurs at around 2000 nm but this is masked by a large feature at 1910 nm, present in all samples, that is attributable to the presence of molecular water. A feature at around 1400 nm is also consistently present, caused by OH -- (Clark et al. 1990;Bishop et al. 2008), but is not diagnostic of chlorite. In many samples, an absorption feature at around 2195 nm is probably a result of the presence of clay minerals containing Al--OH groups (Clark et al. 1990;Herrmann et al. 2001). There is also a notable slope (hull) that descends from ~1900 nm to ~1400 nm with variable gradient.
Within the spectra classified as chlorite--dominant, the most noticeable variation occurs in the position and depth of absorption features centred around 2250 nm and 2340 nm, which are strongly coupled (Fig. 9). For the absorption centred around 2250 nm, the exact wavelength position of the feature minimum varies between 2255 nm and 2248 nm. For the absorption centred around 2340 nm this varies between 2343 nm and 2328 nm. The maximum depth of both features varies between 0.03 and 0.55 (fraction of reflectance range).
The positions of these two features show a relatively strong inverse correlation with the Mg# of chlorite in the samples (Fig.  10). Samples containing more Mg--rich chlorite correspond to spectra where both the 2250 nm and 2340 nm features are shifted to lower wavelengths. Spectra with absorption feature minima at the lowest wavelengths appear to be anomalous, with wavelength values more in fitting with the samples not classified as chlorite--dominated.
The depth of both features also shows a negative correlation with the Mg#, which is considerably stronger at depths below ~0.34 (Fig.  11). Data points with greater depth values appear to be more in the range of depths shown by epidote--dominated samples.
Changes in the slope of the hull between 1900 nm and 1400 nm are also observed, with more Fe-rich samples having a steeper slope (Fig. 12). The aforementioned features are illustrated in selected spectra (Fig. 13).
The position of the OH -absorption feature at around 1400 nm also shows systematic variation which correlates fairly strongly with the Al:Si ratio of chlorite in the samples, with Al--rich chlorite producing spectra with the feature shifted to higher wavelengths (Fig. 14).

Epidote
Epidote geochemistry shows most variation with respect to Fe and Al (Table 3) representing the solid solution between epidote and clinozoisite. Molar Fe:Al ratios range from 0.50 to 0.24, meaning all samples (except one) fall into the classification of epidote (Franz & Liebscher 2004).
Other major elements, including Ca, show remarkably little variation.
In epidote--dominated spectra ( Fig. 15) two absorptions are once again observed at around 2250 nm and 2340 nm, caused by Fe--OH bonds (Clark et al. 1990). However, the most diagnostic feature occurs at 1550 nm and the presence of epidote in a sample is most readily identified by this. Once again, absorptions are seen at 1400 nm and 1910 nm.
The range in wavelength positions of the 2250 nm and 2340 nm absorption features (2251--2255 nm and 2336--2342 nm respectively) is significantly less than in chlorite. The depth of the 2340 nm feature is generally greater than in chlorite (mean = 0.3 compared with 0.2) but that of the 2250 nm absorption is about the same (Table 3). No significant variations in absorption features were observed that correlate with chemical composition (e.g. the Fe:Al ratio), however this may be due, in part, to a paucity of data.

Spatial patterns in chlorite spectral response and chemistry
A number of spectral and chemical features of chlorite vary spatially, relating primarily to distance away from the Batu Hijau porphyry centre and from the Sekongkang prospect.

2250 and 2340 absorption positions
The wavelength position of the features at 2250 nm and 2340 nm are lowest near the centre of the Batu Hijau system, and show a systematic shift to longer wavelengths away from the centre, providing a strong vector to the orebody (Fig.  16). Plotting the data as a function of radial distance from the centre reveals a general increase of the wavelength positions from the centre to 1.2 km, beyond which the values generally plateau (Fig. 17), approximately coincident with the edge of the actinolite subzone. The 2340 nm feature also follows this trend in the northwest with respect to the Sekongkang porphyry centre (Fig. 16B).

2250 and 2340 absorption depths
The depths of the 2250 nm and 2340 nm absorptions show similar patterns, deepening systematically away from the porphyry centre (Fig. 18). The vector to mineralisation remains strong, albeit less clear than in the case of wavelength positions. The Sekongkang prospect is not picked out by feature depth variation which is interesting as a potential discriminator between well-mineralised and poorly mineralised systems. Unlike the wavelength positions, absorption depths continually deepen away from the centre (Fig. 19), although values are unusually elevated at around 1 km to 1.5 km.

Major element chemistry
The chlorite Mg/Fe substitution, as reflected by the Mg#, is likely to be the principal control on absorption position and possibly depth. This also shows systematic, but slightly more complicated, changes from the centre outwards (Fig.  20). Chlorite, within approximately 1 km of the centre, is enriched in Mg and then shows a rapid decrease in Mg:Fe ratio to around 1.5 km; beyond this, the relative content of Mg increases progressively to the limit of sampling at about 4.5 km. Samples at distances less than 500 m from the centre (which correspond to tonalite and carbonate--hosted chlorite) are anomalous, as previously identified in chlorite trace element chemistry (Wilkinson et al. 2015).
The Al:Si ratio in chlorite shows a fairly strong inverse correlation with distance, particularly at distances beyond 3 km from the centre (Fig. 21). However, this is not reflected in any obvious spatial variation in the position of the 1400 nm feature (which is thought to be correlated with Al/[Al+Si]). This may be due, in part, to that fact that no samples beyond 3 km displayed chlorite-dominated spectra.

Discussion
The results from this study support the effectiveness of reflectance spectroscopy in discriminating alteration zones (e.g. Sun et al. 2001;Jones et al. 2005;Zadeh et al. 2014). In detail, the minerals identified in spectra are generally consistent with the documented alteration mineralogy around Batu Hijau (Clode et al. 1999;Garwin 2002). The dominance of clay minerals in the central biotite zone likely represents a sensitive response to the late--stage sericite/paragonite and intermediate argillic alteration overprints ( Fig. 2; Garwin 2002). It is unsurprising that biotite was not identified in the spectra given its remarkably low reflectance (Cudahy et al. 2001) especially when occurring alongside highly reflective smectites. However, vermiculite, which can form from hydrothermal alteration of biotite, was identified. The presence of chlorite in the biotite zone, and its dominance (alongside epidote) in the propylitic zone are consistent with expected propylitic alteration assemblages (e.g. Cooke et al. 2014a). Actinolite, where present, generally occurs in proportions too low to be detected. Zeolite minerals identified in distal samples have been previously noted and attributed to a final, low temperature, alteration stage (Clode et al. 1999).
Geochemical variation in chlorite has previously been recognised as an effective vector to mineralisation in the Batu Hijau system (Wilkinson et al. 2015). The major element geochemistry derived from electron microprobe analysis unsurprisingly matches that previously reported by LA--ICP--MS (Wilkinson et al. 2015) and shows that Mg# and Al:Si ratios in chlorite vary spatially with respect to the orebody. The lack of correlation between chlorite and whole--rock Mg# suggests that bulk rock composition is not a primary control of chlorite chemistry. An increase in chlorite Fe content away from the porphyry centre to a distance of ~1.5 km has been proposed to occur as a result of the outward advection and cooling of hypersaline brines enriched in Fe (Wilkinson et al. 2015). Al:Si ratios in chlorite vary as a result of substitution in the tetrahedral site (Deer et al. 2009) and are linked to fluid temperature (Cathelineau 1988). Ca concentrations may be controlled by this reaction, and other major 2 + ions are also likely to be involved including Mg 2+ and Fe 2+ , as well as trace elements such as Sr 2+ .
These mineral chemical patterns are reflected in SWIR spectral features, especially those centred around 2250 nm and 2340 nm. It is well established that the chlorite Mg# influences the wavelength positions of these features, with more Fe--rich chlorites causing shifts to higher wavelengths (Yang & Huntington 1996;Herrmann et al. 2001;Jones et al. 2005;Bishop et al. 2008), and this study confirms this relationship. Features in this region are all attributed to overtones of metal--OH bond stretching and bending (Hunt 1977) which will be affected by the mass and/or ionic radius of the cations involved.
The apparent increase in absorption depths with decreasing Mg# in chlorite is not documented in any other study and may represent a new finding, but caution should be exercised with this interpretation. It is important to consider whether or not the Mg# is linked to increases in absorption depths, or if both factors are independently controlled by another variable that shows the same spatial pattern. If Mg# does directly control the absorption depth, this is postulated to be a result of the higher mass Fe cations having a stronger effect on the metal--OH bond stretches and vibrations.
An alternative explanation is that an increase in the abundance of chlorite in a sample causes an increase in the depth of its unique absorption features. This is because in any mixed (polymineralic) spectrum the prominence of a feature attributable to one mineral directly depends on its proportion in the sample (Thompson et al. 1999;Herrmann et al. 2001). However, this explanation would contradict studies that have found chlorite abundance to decrease away from porphyry centres (Norman et al. 1991) rather than increase. In the Batu Hijau samples, the proportion of chlorite has not been accurately determined so that this possibly cannot currently be tested.
The effect of mixed mineral assemblages on the SWIR spectra of rocks is perhaps the biggest problem encountered in this study. Although both spectral analysis techniques employ spectral unmixing at their core, the effectiveness is limited -especially when the number of contributing minerals is high and when minerals have overlapping features. For example, the chlorite--epidote-calcite assemblage, which is common in porphyry systems, will have overlapping features in the region of 2340 nm (Dalton et al. 2004). At Batu Hijau, calcite is relatively rare and so presumably has little effect, however epidote is abundant, especially as the secondary mineral in chlorite-dominated spectra. This is likely to affect both of the important chlorite features at 2250 nm and 2340 nm.
In terms of wavelength positions, the narrower range of wavelength positions that epidote--related absorptions occupy (Table  3) could create a bias in chlorite spectra towards more central positions in such samples. However, the fact that chlorite Mg# correlates with feature positions (Fig. 10) with the same trend as observed in previous studies on epidote--free samples (Yang & Huntington 1996;Herrmann et al. 2001;Jones et al. 2005;Bishop et al. 2008) indicates that any epidote interference is limited. In terms of depth, the presence of epidote should deepen the 2340 nm absorption feature, which might explain the anomalously high absorption depth values in the plots demonstrating correlations between absorption depth and chemistry (Fig. 11). With this considered, the absorption feature at 2250 nm is likely to be a better indicator of chlorite composition than that at 2340 nm. This point is also made in Herrmann et al. (2001) because the secondary Al--OH feature in white mica also overlaps the chlorite feature at 2340 nm. However, use of the 2250 nm feature may be limited in some deposits because it has been shown to completely disappear where there is pervasive weathering (Suryantini 2003).
Two other spectral features were shown to vary with chlorite chemistry -the slope of the hull between 1400 nm and 1900 nm, and the position of the 1400 nm absorption feature. The increase in hull slope in more Fe--rich chlorite has been recognised previously (Thompson et al. 1999) and is attributable to the strong, broad absorption caused by electronic (charge transfer) effects in Fe 2+ in the VNIR (Hunt 1977;Clark et al. 1990). However, this feature occurs in many Fe--bearing minerals and so may just reflect an abundance of these in the rock, rather than a specific response to chlorite, thus explaining the lack of any strong systematic spatial variation. The 1400 nm wavelength position appears to relate to the Al:Si ratio in chlorite. Shifts in this absorption have been previously observed in chlorite (King and Clark 1989) and are likely due to modification of the O--H vibration frequencies as a result of the Tschermak substitution (Duke 1994). Unlike the hull slope, this feature appears to offer considerable promise for use in exploration using field--portable or core--shed--based SWIR spectrometers. However, both features are inapplicable to satellite remote sensing due to strong atmospheric scattering effects at these wavelengths (Duke 1994).
This study has been unable to conclude anything about potential mineral chemical controls on spectral variations of epidote due to a distinct lack of overlapping data: of the 16 samples that could be classified as epidote--dominated, only five had accompanying geochemical data. This is unfortunate because epidote has shown promise as a vectoring tool in porphyry deposits (Cooke et al. 2014b) and spectral variation across the epidote--clinozoisite solid--series is certainly recognisable (Roache et al. 2011). This is an area where future study would be useful.

Potential application to satellite remote sensing
The results from this study indicate that porphyry deposits could be targeted by using the absorption features at 2250 nm and 2340 nm which differ between Mg--rich chlorite that can be dominant in the inner propylitic zone (Wilkinson et al. 2015) and Fe--rich chlorite that can predominate further out, or which may be developed above buried porphyry systems (Halley et al. 2015). The precise wavelength positions can only be exploited, at present, using field spectrometers or hyperspectral sensors that are capable of collecting high--resolution spectra. It should therefore be possible to target porphyry deposits using hyperspectral imaging, and previous work has demonstrated the discrimination of Fe--chlorite from Mg--chlorite in hyperspectral data (Cudahy et al. 2001). However, the small wavelength shifts documented here are currently impossible to target using the broad bandwidths (0.02--0.2 µm) typical of satellite remote sensing.
Fortunately, the relative change in absorption feature depth is potentially recognisable using broad-band multispectral imagery collected by the ASTER (Advanced Spaceborne Thermal Emission Radiometer) satellite sensor. ASTER bands 7 and 8 measure surface reflectance in the ranges of 2235--2285 nm and 2295--2365 nm which individually span both absorption features of interest.
Deeper absorption features should produce a darker (lower reflectance) response and ASTER band ratio images could perhaps be used to enhance these features.

Conclusions
SWIR reflectance spectroscopy is shown to be a powerful tool for characterising propylitic alteration associated with porphyry deposits. The collection of high resolution spectra using handheld spectrometers such as the TerraSpec 4, combined with matching algorithms such as those used by TSG and Halo, allows quick identification of spectrally--active minerals, even in the case of low--reflectance propylitic "green--rocks". From the spectra, chlorite and epidote were found to be the most abundant SWIR--active minerals in rocks from Batu Hijau, and are in greatest abundance in the more proximal propylitic zone. Spectral signatures from the biotite zone commonly indicate the presence of smectites and other phyllosilicates attributable to later stage, intermediate argillic and sericitic overprinting.
Chlorite chemistry can be utilised as a powerful vector to ore (Wilkinson et al. 2015) and this appears to translate to features in the SWIR spectra of chlorite--dominated samples. Chlorite from Batu Hijau varies significantly in terms of Mg#, which is likely the main cause of shifts in the wavelength position of absorption features at around 2250 nm and 2340 nm, and possibly the cause of depth variation in these features. The positions and depths of both features act as good vectors to mineralisation: wavelength positions shift from 2254 to 2249 nm and from 2343 to 2332 nm, and absorption depths decrease from around 35% to 5% respectively, moving from 1.6 km to 500 m away from the orebody. This correlates with a general increase in the Mg# of chlorite from about 1.6 km towards the fringes of mineralisation. Absorption feature depth variations in particular seem to be a new finding and one potentially more applicable to remote sensing than the well--established position shifts because absorption depth changes can potentially be recognised by satellite remote sensing. The exact cause of changes in absorption depth is not entirely clear, but the fact that this can act as a vector to mineralisation is significant nonetheless, especially if these observations can be recognised in other porphyry systems.
The degree to which these findings apply to other porphyry deposits is still unclear. Furthermore, there is limited information at present to suggest that the variations in alteration mineral geochemistry are specific to fertile (well--mineralised) systems rather than weakly mineralised or barren ones. However, we do note that that the 2250 nm peak position and both 2340 nm and 2250 nm absorption feature depths do not appear to highlight the weakly mineralised Sekongkang prospect in the Batu Hijau district. Future work could include investigating chlorite spectral variation in other, variably mineralised, porphyry systems, and investigating in more detail the effects of epidote. We conclude that SWIR spectroscopy offers significant promise as a practical exploration tool in the "green--rock" environment.

Acknowledgements
We thank the AMIRA P765A project team for their previous sampling, research and data collection, and the access and logistical support from Newmont which made the sampling for this project possible. We are extremely grateful to ASD Inc.       (from  TSG). 'Chlorite/Epidote' refers to samples classified as chlorite in one instance, and epidote in the other -it is probable that they contain roughly equal amounts of each. Samples classified as 'Other/Invalid' were not used in any analysis involving spectral features. Base map modified after Garwin (2000).      34 are thought to be outliers possibly due to epidote in the sample (shown in grey). Vertical error bars are 1σ standard error (propagated through the ratio calculation). Horizontal error bars calculated from the signal--to--noise ratio of the spectrometer are not shown as they were determined to be negligible (±0.025%)

Fig. 12. Chlorite Mg# plotted against the 'hull slope index' (reflectance at 1850 nm/reflectance at 1440 nm)
which corresponds to the slope of the hull. Greater values represent steeper slopes. Data points at the highest end appear anomalous (shown in grey) and may reflect the presence of other Fe--bearing minerals in the samples. Vertical error bars are 1σ standard error (propagated through the ratio calculation). Horizontal error bars calculated from the signal--to--noise ratio of the spectrometer and error propagation are not shown as they were determined to be negligible (±0.06% and lower) Fig. 13. Chlorite--dominated (hull--quotient corrected) spectra selected to illustrate changes in features. Spectra are stacked in order of the 2340 nm position and coloured according to Mg#. Distances from the sample site to the centre of the orebody are also labelled. Both the 2250 nm and 2340 nm features show a shift to higher wavelengths with decreasing Mg#. Feature depths also increase with increasing Fe content. A steepening of the hull between 1400 nm and 1900 nm is also observed as Fe content increases.   Larger red circles correspond to shorter wavelengths. A systematic shortening of wavelength position is seen towards the centre of the Batu Hijau deposit along the south--west transect for both absorption features. This is also exhibited more weakly in the west towards the Sekongkang porphyry centre for the 2340 nm absorption feature. The 'strong biotite' alteration zone at Batu Hijau is outlined in black.

Fig. 17. Wavelength position at the minimum for (A) the 2250 nm; and (B) the 2340 nm absorption features
in chlorite plotted against lateral distance to the nearest mineralised centre. Systematic increases are seen outwards to a distance of around 1.6 km with wavelengths decreasing slightly beyond this. Green points indicate samples with distances measured away from the Sekongkang centre rather than Batu Hijau. Grey points correspond to extremely proximal, mineralisation--hosting tonalite samples that are not strictly part of the propylitic halo and may represent a late overprint (see Wilkinson et al., 2015). Vertical error bars are ±0.5 nm reflecting the measurement precision of the spectrometer. Horizontal error bars are smaller than the symbol size (uncertainty in sample position is less than 10 m from GPS and the distance to ore estimate is thought to be accurate to within 25 m).   Chlorite Mg# plotted against lateral distance to the nearest mineralised centre. The data show a sharp decrease in Mg content to a distance of around 1.6 km away from the centre, followed by a steady increase to what may be background levels. No significant relationship is seen in samples measured away from the Sekongkang centre (green). Vertical error bars are 1σ standard error.

Number of channels 2151
Acquisition software Indico® Pro

Hi-Brite Contact Probe
Light source Halogen bulb, 2901 K (±10%) Spot size 10mm Sample data and classification of mineralogy determined from the spectra. Radial distances to the nearest major porphyry centre are given for each sample; this is to Batu Hijau for most of the samples, but to the Sekongkang system for samples in the western traverse.
* Samples collected from close to the Bambu epithermal vein system † (A) and (B) refer to the two separate sets of spectra acquired for each sample ‡ Principal mineral identified in spectra. "Chlorite/Epidote" refers to samples determined to be chlorite in one set of spectra and epidote in the other. It is probable that they contain roughly equal amounts of each. § Samples proximal to the Sekongkang porphyry centre (distances measured away from this) All geochemical data (except ratios) reported in weight percent, and averaged from multiple readings per sample. Outliers were removed prior to averaging (see text for details). Geochemical data are derived from the AMIRA P765A project database.
* Samples collected from close to the Bambu epithermal vein system † LA-ICP-MS data with anomalously elevated Fe and Mg values (not used in analysis) ¶ 'Hull slope index' = reflectance at 1850nm / reflectance at 1440 nm