A Method for Detrital Corundum Characterization in Sediments: Case Study of the Gem Mountain Mine Placer Sapphire Deposit (Rock Creek, Montana, USA)

Gem corundum (ruby and sapphire) in placer deposits has generally been characterized in larger grain sizes, more specifically in the size fraction of economic interest (generally >2 mm, e.g., Buravleva et al. 2016, Sutherland et al. 2009) and seldom characterized in <1 mm size fractions. In contrast, general heavy mineral surveys typically sample grains in the 0.25–1 mm or 0.25–2 mm size fractions of ∼10 kg sediment samples (McClenaghan 2011), and larger grain sizes are not typically examined due to the rarity of large heavy mineral grains in sediments targeted for region-scale surveying. Quaternary sediment heavy-mineral sampling programs in Canada have noted the presence of corundum grains in tills (e.g., Tremblay et al. 2020) and stream sediments (e.g., Falck et al. 2015). However, there exists little context in regards to placer gem deposits with which to interpret these findings, i.e., knowledge of the characteristics or concentrations of detrital corundum that occur in economically viable gem deposits. Furthermore, studying the fine fraction of corundum in placer deposits could yield important genetic information relating to inclusions, color, zoning, and, more broadly, mineral chemistry. This study details the corundum size distribution, abundance, shape, color, zoning, transparency, and inclusions from the sapphire-producing sediment layer at the Gem Mountain Mine (Anaconda Bench, N 46.272800°, W 113.6215°), Gem Mountain, Rock Creek, near Philipsburg, Montana, USA. The goal of the study is to: (1) outline a comprehensive and systematic method for characterizing corundum in sediment samples, (2) provide benchmark corundum concentration data in an economic placer deposit for future regional heavy-mineral studies and gemstone exploration, and (3) apply the knowledge gained from detrital corundum grains in all size fractions to gain insights into Rock Creek placer sapphire corundum geology and mineralogy.

The Rock Creek placer sapphire deposits are located in SW Montana, USA, northwest of the town of Philipsburg. They are located within the Sapphire tectonic block, an allochthon consisting of folded and faulted low-grade Mesoproterozoic metasedimentary rocks (including limestone, dolostone, quartzite, and argillite) which was intruded by late Cretaceous to early Tertiary plutons of diorite and granite (Lonn et al. 2010). The rhyolitic Rock Creek volcanic field is located within the Philipsburg area. It consists of sills (intruding Tertiary siliciclastic sedimentary rocks, e.g., siltstones, tuffaceous sedimentary rocks), rhyolite flows, tuffs, and volcaniclastic rocks (Berg 2014, Lonn et al. 2010). At Rock Creek, corundum var. sapphire (henceforth sapphire) occurs in placer deposits over an 11 km² area associated with rhyolite float (Lonn et al. 2010, Berg 2014).

The sapphire-bearing colluvium, which is the focus of this study, is typically less than 3 m in thickness (Berg 2014), shows rudimentary bedding, and consists of poorly sorted sediments with a mud matrix containing an abundance of angular clasts (Barron & Boyd 2015). Sapphire has also been produced from alluvial deposits in streams within some of the gulches downhill from the colluvial placers (Berg 2014).

The source rock for sapphire has been suggested to be rhyolite, based on the rare presence of adhering rhyolitic material on the surface of some sapphire grains (Berg 2014). Garland (2002) alternatively proposed that sapphire in the colluvial deposit on Anaconda Bench is of metamorphic origin and comes from a re-worked Tertiary paleoplacer which mixed with volcaniclastic rock and ash during mass wasting. Using evidence from primary and secondary melt inclusions, Palke et al. (2017) suggest that sapphire formed via a peritectic melting reaction during partial melting of a hydrated plagioclase-rich protolith (such as an anorthosite), and that the sapphire was then transported to the surface by trachytic, rhyolitic, or dacitic magmas. At the Silver Bow sapphire occurrence (∼80 km SE of Rock Creek; Berger & Berg 2006), two rock fragments, interpreted as xenoliths within felsic-lapilli-tuff-rich debris flows, were noted to contain corundum. These included an amphibolite-facies, biotite-sillimanite-andesine schist containing light (“pastel”) colored corundum and an igneous albite-biotite-glass rock containing small idiomorphic dark blue corundum crystals. Berger & Berg (2006) proposed that pastel-colored corundum which has a corroded surface appearance is of metamorphic origin (i.e., xenocrysts in the felsic volcanic rocks), while the uncommon, small, dark blue corundum represent magmatic phenocrysts.

Two sediment samples were collected at the Gem Mountain Mine by placer mine operators following instructions from the author. The first, sample M1 (10.4016 kg dry weight), was collected using a clean shovel from the producing layer of colluvium 1 m below the ground surface, avoiding plant roots and organic matter. Due to the sample size and collection method, large cobbles and boulders are not represented. The second, sample M2 (11.1110 kg), is a sample of colluvium that was processed at the mine using a mechanized 2-inch-minus sieve. It was used as a secondary sample for picking of individual grains in the 0.125–0.250 mm fraction of corundum, whereas the same size fraction in sample M1 was analyzed with SEM-MLA (scanning electron microscopy-mineral liberation analysis).

A 2500 g split was taken from sample M1 and used for granulometric analysis (aperture sizes in mm: 0.063, 0.125, 0.25, 0.5, 1, 2, 4, 8, 16, 32, 64). Sample M2 and the remaining portion of sample M1 were processed using standard heavy-mineral concentration procedures (McClenaghan 2011). A blank was run before and after each of the samples was processed. Samples were screened in water using a 2 mm sieve aperture. The >2 mm fractions were sieved at the following aperture sizes (mm) and dried: 4, 8, 16, 32, 64. Heavy minerals were separated from the <2 mm fraction using a gravity separation shaking table (Wilfley table), and the resulting heavy-mineral fraction was further separated using methylene iodide, resulting in a concentrate of particles with a specific gravity greater than 3.3 g/cm³. This concentrate was separated into ferromagnetic and non-ferromagnetic fractions and then sieved at apertures of 0.125, 0.250, 0.50, and 1.0 mm.

All grains in the 0.250–8 mm size fractions that had an appearance similar to that of corundum (any color, monocrystalline or glassy polycrystalline, excluding obvious examples of quartz and feldspars in fractions not processed for heavy-mineral separation) were picked and their identity verified with energy-dispersive spectrometry (EDS). Corundum was picked from sample M1 from the 0.250–2 mm heavy-mineral separates, the 1–4 mm granulometry splits, and the 2–8 mm washed fractions. No corundum was found in the >8 mm sediment fractions. A selection of grains in the 0.125–0.250 mm fraction were picked from a secondary sample (M2), mounted, polished, and verified with SEM-EDS. Picking was done under the binocular microscope (<2 mm) and under a 10× magnifying glass (>2 mm).

Non-ferromagnetic heavy minerals in the <0.125 mm and 0.125–0.250 mm fractions were spread as even monolayers and set in 25 mm epoxy pucks (one puck for the finer fraction and three for the coarser fraction). Samples were flat polished using diamond grit and ground to a depth to maximize the surface areas of mineral grains. Due to the variation in grain thickness, particularly with regards to corundum (some grains being very thin and others blocky), grain size and shape data measured from the polished surface is not considered to be adequately representative of the grains, so has thus not been reported.

Corundum grains picked from the coarser size fractions of sample M1, and a selection of grains from the 0.125–0.250 mm heavy-mineral separate from sample M2, were mounted in 25 mm epoxy pucks and polished.

The present study presents a new approach to the characterization of detrital corundum grains. It has been modified from standard sediment-particle characterization criteria (i.e., shape, form, roundness, surface texture, adhering materials) and includes the incorporation of additional characterization criteria previously used on Montana sapphire (e.g., Berg 2014, Garland 2002). These criteria are outlined in Table 1 and emphasize qualitative determinations, in some cases, using well-established visual comparison charts. The use of qualitative, as opposed to quantitative, criteria is done to facilitate the collection of varied data on a large number of grains under different conditions (e.g., under no or low magnification, using a binocular microscope, and an electron microscope) at a practical speed. Colors were assessed under a 5000K LED light source.

Non-ferromagnetic heavy minerals in the <0.125 and 0.125–0.250 mm size fractions were mounted as monolayers in epoxy pucks, polished, carbon coated and characterized by automated mineralogy using a scanning electron microscope (SEM) equipped with an energy dispersive X-ray spectrometer (EDX/EDS). The data obtained were processed using mineral liberation analysis (MLA) software, which has been previously used for automated quantitative mineral identification in sediments, particularly fine-grained sediments (Sylvester 2012). The instrument used, at Memorial University of Newfoundland, is a FEI Quanta 650 SEG SEM instrument with field emission source, secondary electron detector, and dual Bruker XFlash 5030 EDX silicon drift detectors. The MLA was performed using the XBSE method, where the software identifies areas of different brightness in backscattered electron images and collects a single X-ray spectrum from each distinct mineral, even within single grains (Sylvester 2012). The instrument was run at high voltage (25 kV) with an X-ray acquisition time of 12 ms. The frame size was 1.5 × 1.5 mm with a resolution of 800 × 800 pixels per frame.

Chemical compositions were obtained with a JEOL JXA-8230 electron microprobe (University of Ottawa) in wavelength-dispersion (WD) mode. The operating voltage for silicate analyses was 20 kV, with a 20 nA beam current and 5 μm beam diameter (20 μm for phengite). Counts were collected for 20 s for each element, with the exception of F (100 s). The following standards and X-ray lines (all Kα, except for Zn, Sr and Ba, for which Lα lines were used) were employed: sanidine (Si, Al, K), albite (Na), diopside (Ca, Mg), hematite (Fe), synthetic cochromite (Co), chromite (Cr), rutile (Ti), tephroite (Mn), pentlandite (Ni), gahnite (Zn), sanbornite (Ba), celestine (Sr), tugtupite (Cl), and synthetic F-rich phlogopite (F). The operating conditions used for rutile analyses were 20 kV, 40 nA beam current, and 5 μm beam diameter. Counts were collected for 20 s. The following standards were used for rutile analyses (all Kα except for Zn, Zr, W, Nb, Ta, for which Lα lines were used): columbite-(Mn) (Nb, Mn), synthetic NiTa₂O₆ (Ta), zircon (Si, Zr), chromite (Al, Cr), hematite (Fe), vanadinite (V), rutile (Ti), synthetic CoWO₄ (W), and gahnite (Zn). Matrix corrections were made using the Armstrong ϕ(ρZ) method (Armstrong 1988).

Zircon and apatite chemical compositions were obtained with a JEOL JXA-8230 electron microprobe at the Memorial University of Newfoundland. Operating voltage was 15 kV, a beam current of 50 nA (zircon) or 20 nA (apatite), and 1 μm (zircon) and 10 μm (apatite) beam diameters. Count times and background collection times were 40 s for zircon (except 20 s for Na, K, Zr, Si) and apatite (except 20 s for Na, P, Cl, 30 s for Ca, 10 s for F). The following standards were used: a set of synthetic REE+Y phosphates (Ce, La, Nd, Y), synthetic UO₂ (U), synthetic ThO₂ (Th), synthetic zircon (Zr), synthetic hafnon (Hf), apatite (P), synthetic fluorapatite (F, Ca in apatite), synthetic chlorapatite (Cl), diopside (Si, Mg, Ca in zircon), almandine (Al, Fe), rhodonite (Mn), sphalerite (S), albite (Na), arsenopyrite (As), celestine (Sr), and orthoclase (K). Matrix correction calculations were done using the PAP method (Pouchou & Pichoir 1985).

Sample granulometry was determined for sample M1. A 2.5 kg archive split was used for sieve analysis, while the remainder of the sample (7.9016 kg) was used for (1) a sieve analysis of the granule/pebble size fractions and (2) the concentration of heavy minerals, including corundum, from the <2 mm fraction. Granulometry data is reported in Table 2, and the grain-size distribution in <2 mm size fractions is recalculated for the sieve analysis split and the total sample based on the grain-size distribution measured in the granulometry split. Colluvium sample M1 consists of a poorly sorted sediment with a mud-rich matrix (36.8 wt.% mud; Fig. 1).

Monocrystalline plagioclase and quartz grains are common in the <4 mm grain-size fraction. Plagioclase shows typical cleavage faces and grains are angular. Quartz grains in the 1–4 mm size range are typically glassy and transparent and range from angular to slightly rounded (due to their being subhedral). These quartz grains are identical to those found in pebbles of rhyolite tuff. None of these grains show surface textures characteristic of abrasion.

The coarser-grained fractions of sediment (>2 mm) are primarily composed of angular to slightly rounded clasts of rhyolite tuff (porous and commonly with clay alteration), rhyolite, and a weakly metamorphosed K-feldspar-quartz siliciclastic rock, though complete clast characterization is beyond the scope of the present study.

In sample M1, corundum grains generally increased in abundance with decreasing grain size. In the coarser-grained fractions (total sample including granulometry split), corundum grain counts are as follows: 4–8 mm (1 grain), 2–4 mm (13 grains), 1–2 mm (10 grains, granulometry split picking combined with grains from heavy-mineral separation). In the finer grain sizes (heavy-mineral separation split only), corundum grain counts are: 0.5–1.0 mm (34 grains), 0.25–0.50 mm (72 grains), 0.125–0.250 mm (383 grains, determined using SEM-MLA, below), and <0.125 mm (44 grains, determined using SEM-MLA). The grain counts, normalized to 10 kg of total sediment (Fig. 2A) and to 1 kg of the same size fraction (Fig. 2B), are summarized in Table 3. Smaller corundum grains (<2 mm) outnumber larger grains (>2 mm) in the sediment by a factor of ∼50 times (Fig. 2C).

Automated mineralogy was used to quantify corundum grain counts in heavy-mineral separates of the 0.125–0.250 mm (383 grains) and <0.125 mm (44 grains) size fractions. Associated non-ferromagnetic heavy minerals include (in order of decreasing abundance): allanite, tourmaline, almandine, hematite, probable pyroxene, apatite, zircon, Al-silicate, spessartine, epidote, pyrite, ilmenite, TiO₂, titanite, and monazite. Black, euhedral, stubby crystals of allanite are by far the most abundant non-ferromagnetic heavy mineral. Grains of heavy minerals range from being angular and euhedral, to more rounded subhedral/anhedral grains, or as broken fragments, none of which show surface textures characteristic of abrasion.

The corundum grains consist of single crystals with few exceptions. One grain consists of a cluster of multiple blue idiomorphic corundum crystals (0.25–0.50 mm fraction), and nine grains (seven in 0.125–0.250 mm, one in <0.125 mm, one in 0.5–1.0 mm) consist of multi-mineral intergrowths containing corundum with biotite, K-feldspar, and/or plagioclase, the latter being the most common. Multi-mineral corundum grains are described under “Corundum grain properties” below.

The principal trends observed in corundum grain properties, their frequency, and select important examples are outlined below. A complete set of descriptions and photographs of the corundum grains examined are available as supplementary materials (Appendices 1–6)¹.

The dominant form (i.e., the tri-dimensional characteristic of a grain defined by the ratios of its three dimensions; see Sneed & Folk 1958) of corundum grains in the >2 mm size range is compact bladed, where subhedral/euhedral grains of corundum tend to be of shortest dimension along the c axis and have two differing proportions along a. Overall, corundum grains primarily vary from compact to bladed, platy, and elongated, with bladed and platy crystals (generally with c < a) being the most abundant (Fig. 3). However, the frequency of forms in the 0.250–0.500 mm fraction is more evenly distributed than in the others. The dominant forms associated with 35 grains from the 0.125–0.250 mm fraction in sample M2 are bladed, platy, very bladed, and very platy. The presence of thin broken fragments, particularly in the 0.125–0.250 mm size range, suggests that very bladed and very platy grains were more likely to fragment during erosion and transport, although the fragmentation of extremely thin grains during sample processing cannot be discounted. Overall, the c axis in euhedral and subhedral crystals appears to be the shortest dimension in most of these grains, and the few examples of elongated grain shapes are elongated parallel to the basal face rather than perpendicular to it (i.e., they are not barrel-shaped or spindle-shaped crystals elongated along the c axis as is common with corundum from numerous localities around the world).

Corundum grains are described according to the extent of crystal-face development, where euhedral crystals have discernible crystal faces that show minor to no rounding, subhedral crystals show some recognizable aspects of crystal form (commonly better developed basal faces but irregular or rounded edges and prism faces), and anhedral grains show no readily recognizable crystal faces. It should be noted that grains are described at their respective sizes, thus a similar irregular surface texture may result in very angular shape in smaller size fractions, whereas the surface topography of this same texture may be insignificant at coarser grain sizes (e.g., in the 2–8 mm size range), and therefore may not be characterized as very angular. Similarly, such an irregular surface texture may result in anhedral character at small grain sizes whereas larger grains with an identical surface texture still appear euhedral.

Three surface textures predominate and correlate with roundness/angularity, with an additional fourth population of small idiomorphic grains. Grains with the Type 1 surface feature include rounded to subrounded grains with smooth surfaces covered in minute, relatively equal sized narrowly spaced “bumps” (Fig. 4A) with a matte (all grain sizes) to frosted appearance (<0.5 mm fractions; see example also showing locally larger protruding bumps in Appendix 5A, grain 7-2). In many cases, the small surface bumps reflect light at the same angle of observation, and some appear to show the presence of tiny basal faces. Grains with the Type 2 surface features include subangular to very angular grains with hillocky to rough surfaces, commonly with irregular protrusions and depressions, and glassy luster. An example of this is shown by the anhedral grain in Figure 4C. Here, an irregular surface on a subhedral grain shows some evidence of crystallographically controlled surface structure. Such features are uncommon and may also include stepped “growth” (?) features or trigonal-shaped depressions or protrusions (e.g., Fig. 4D). Grains with the Type 3 surface are uncommon but include angular to very angular grains (<1 mm) with relatively smooth, glassy surfaces associated with crystals that have curved “faces,” which may be stepped but wherein the step lines are curved. Grains with the Type 4 surface are angular to very angular, with grains (<1 mm) always showing euhedral morphology and generally having relatively smooth, glassy surfaces (Fig. 4E), though rarely showing a slightly rough or matte surface. Larger (>0.5 mm) euhedral grains most commonly have slightly rough surface texture or, less commonly, slight rounding and a smooth matte surface texture. Rarely, grains are of mixed Type 1 and Type 2 textures, where part of the grain has a smooth, matte, and bumpy surface, with localized areas of smooth to hummocky glassy texture (Figs. 4A–B). In the case of the grain shown in Figure 4A, the smooth texture abruptly changes to a hummocky texture, resulting in angular grain features at the transition.

Some Type 1 grains consist of broken fragments also showing smooth, glassy conchoidal fractures, which are angular on fracture edges. One corundum grain with Type 1 surface features has what appears to be a conchoidal fractured surface which shows a matte surface (Fig. 4F).

Euhedral crystals comprise a larger proportion of corundum grains in the coarser size fractions (>1 mm) relative to smaller grain sizes, where irregular surface features more significantly disrupt the extent of crystal face development of the crystals. It should be noted that the increase in the proportion of euhedral crystals below 0.50 mm, relative to the those in the 0.5–1.0 mm fraction, is related to the presence of euhedral, blue corundum grains. Some of these blue corundum crystals are remarkably well-formed, with perfectly flat, lustrous crystal faces and straight, well defined lines of plane intersection (Fig. 4E). The increase in abundance of subhedral grains in the 0.125–0.250 mm compared to the 0.25–0.50 mm size fraction is due to the presence of subhedral crystals with well-developed basal faces and other irregular and corroded faces in the 0.125–0.250 size fraction.

Corundum grains are dominated by angular grain shapes, with ∼30–40% rounded anhedral grains in the 1–2 mm and 2–8 mm size fractions having a matte surface texture (Fig. 5).

Corundum colors (summarized in Fig. 6) were described according to hue, saturation (low, medium, high, and vivid), and lightness (light, medium, dark). Corundum in the coarser-grained fractions (1–8 mm) is predominantly yellowish-green (Fig. 4D), green (Fig. 4C), and colorless (Fig. 4B), with one grain in the 2–4 mm fraction having a golden-yellow core that diffusely grades into a green outer zone (Fig. 4F). The color saturation in yellowish-green and green corundum is low, whereas yellow corundum is low-medium in saturation and has a slight brownish hue. The yellowish-green and green corundum is also light in color and becomes progressively lighter with decreasing grain size (i.e., they are light in the >2 mm fractions and extremely light in the <0.5 mm fractions) due to shorter light path length and thus a proportionally lower absorption. Thus, unsurprisingly, colorless corundum comprises a larger proportion of the total as grain size decreases, where the lightest yellow-green and green corundum would have imperceptible color at very small sizes. Colorless grains do not predominate in the 0.125–0.250 mm size range (sample M2), but since this is a partial pick, it is possible that blue grains are over-represented due to their visibility, despite care being taken to avoid their over-representation during picking. Bluish grains with low color saturation (one bluish gray, one bluish green) are present in small quantity in the 0.5–1.0 mm fraction. Light blue corundum with medium-high color saturation is present in notable quantity (10%; see Fig. 4E) in the 0.25–0.50 mm and 0.125–0.250 mm size fractions.

Three corundum grains in the 2–4 mm size fraction (23%) show color zonation, consisting of the following combinations: yellowish-green and green (1 grain); yellow and yellowish-green (2 grains); and colorless and yellowish-green (1 grain). No color-zoned corundum was observed in the 1–2 mm sample. One color-zoned corundum, gray to bluish gray, occurs in the 0.5–1.0 mm fraction (3%). Seven zoned grains (10%) in the 0.25–0.50 mm size fraction vary in color: blue and colorless (5 grains; Fig. 4E); yellow and yellowish-green (1 grain); and a variation in the lightness of green from light to medium (1 grain). In the partially picked fraction from sample M2 (0.125–0.250 mm), three zoned grains are noted (∼9% of all grains in this fraction): colorless and golden yellow (2 grains) and colorless and blue (1 grain). The color zonation appears to be oscillatory in nature, with diffuse transitions between colors, the only exception being in blue-to-colorless grains, where the zonation is sharp and sectors may be visible.

Corundum grains in all size fractions exhibit a high-degree of transparency. This transparency is generally not significantly affected by the presence of mineral inclusions. A few corundum grains are translucent due to the very high abundance of fluid inclusions which occur along numerous, re-healed fractures throughout the grains (1 grain of 13 in 2–8 mm, 0 of 10 in 1–2 mm, 1 of 34 in 0.5–1.0 mm, and 7 of 72 in 0.25–0.50 mm). Additionally, one grain in the 0.5–1.0 mm size fraction is translucent due to a very high abundance of rutile “silk” (i.e., fine web-like networks of rutile needles; Appendix 4, grain 5-2), and two blue grains in M2 picked 0.125–0.250 mm fraction are translucent, but the nature of the inclusions could not be determined.

Fine, needle-like rutile “silk” is uncommon in finer grained (<1 mm) corundum and more common in grains larger than 1 mm, in which the rutile “silk” occurs in oscillatory zoning bands consisting of idiomorphic growth zones. Some of these rutile-bearing bands are truncated by the grain surface in both smooth, rounded grains (Fig. 7A) and rough, glassy grains (Fig. 7B).

The characterization of the inclusions present focuses on those exposed at the polished surface, from which chemical composition can be measured using EDS and EPMA. As such, inclusion abundances cannot be taken as being wholly representative of those present in the sample set. The most common solid inclusions are biotite, Ti oxide (usually Nb-bearing rutile), and plagioclase. Mineral inclusions analyzed in blue corundum consist of rutile (4), ilmenite (2), alkali-feldspar (3), zircon (2), apatite (1), Mg-Al spinel (1), and muscovite (1). Yellow-green and yellow corundum contain inclusions of plagioclase (4), hercynite (1), Cu-Fe sulfide (1), Fe-sulfide (1), F-rich biotite (1), and fine-grained, porous phengite (1). Colorless corundum contains inclusions of biotite (6), plagioclase (4), rutile (2), Nb-poor TiO₂ (1), and a kaolinite-subgroup mineral replacing biotite (1). Multiple yellow-green and colorless corundum grains contain reddish-brown subhedral rutile prisms, and one grain appears to contain a colorless elongated zircon (?) inclusion (Fig. 7B), but these were not exposed on the polished surface. Some of the larger plagioclase inclusions reach the surface of the corundum grains. The fine-grained phengite is connected to the grain surface by open fractures, so may not be primary. The inclusion of zircon has a large core with sharp growth zones, suggesting growth in a melt, and is rimmed by zircon that does not luminesce. This zircon is included in blue corundum, but also exposed at the grain surface (Fig. 8).

Nine grains (seven in the 0.125–0.250 mm fraction, one in the <0.125 mm fraction, and one in the 0.5–1.0 mm fraction) consist of multi-mineral composite grains with corundum being a principal component. The largest composite grain consists of corundum with abundant biotite, both as inclusions and intergrowths at the grain surface (Appendix 4, grain 4-1). One grain consists of a single poikilitic biotite crystal containing numerous, randomly oriented, euhedral to subhedral corundum crystals (Fig. 9A). A composite grain consisting of K-feldspar and corundum shows two corundum crystal orientations and a quasi-sieve texture in one of the corundum (Fig. 9B). A second aggregate, rich in K-feldspar, contains irregular shaped corundum with embayed contacts, and both minerals contain inclusions of Ti-bearing biotite (Fig. 9C). Similar embayed grain boundaries are seen in a corundum-plagioclase grain roughly 50 μm across (Fig. 9D), which differs from the lathlike plagioclase-corundum intergrowth seen in a larger grain (Fig. 9E). The plagioclase and biotite chemical compositions (as measured using EDS) are comparable to that of inclusions in other corundum grains. The sample in the 0.5–1.0 mm fraction consists of a corundum crystal with abundant biotite and plagioclase inclusions (described in the section above), some of which are significantly exposed on the grain surface.

In addition to the composite grains described above, other mineral assemblages are observed at the surface of corundum grains. These are not as intergrowths and consist of (1) a mixture of quartz, plagioclase, and K-feldspar, (2) an unidentified hydrous aluminosilicate that frequently contains minute grains (generally <10 μm) of various minerals, and (3) one example of a plagioclase overgrowth on blue corundum (0.5–1.0 mm size fraction).

Mixtures of plagioclase, quartz, and K-feldspar occur on two samples. The first, in the 0.125–0.250 mm fraction of sample M1, shows a larger crystal of plagioclase together with finer grained quartz-plagioclase-K-feldspar in a depression at the surface of a corundum grain (Fig. 10). The second example, in the 0.250–0.500 mm fraction, consists of micron-sized felsic mineral particles that fill a depression or surface-reaching fracture on the edge of a corundum grain. The minerals are too fine-grained (<1 μm) for individual mineral characterization, but EDS spectra and the backscatter image suggest a combination of K-feldspar, plagioclase, quartz, and trace apatite.

The second adhering mineral assemblage occurs on more than a dozen grains. It is composed of an assemblage of aluminosilicate minerals with low relative brightness in BSE (therefore a low Z_average). This assemblage, as characterized with semi-quantitative SEM-EDS, has an atomic Al:Si proportion varying from 1:2 to 3:4, always contains 1–2.4 mol. % Fe, and can contain up to 0.4 mol.% each of Mg, K, and Ca, though these three elements are not always present. It commonly contains grains (<10 μm) of biotite, plagioclase, K-feldspar, quartz, rare corundum, and one example of zircon (1 μm). The material is most common in deeper surface recesses on corundum grains and commonly separates from the grain surface in the epoxy mounts. Some examples also appear to have shrunk in volume, leaving crack-like gaps in the material. This material is interpreted to be adhering sediment consisting of a mixed illite/clay matrix supporting very fine-grained particles of biotite, plagioclase, K-feldspar, quartz, and corundum.

The chemical composition of TiO₂ inclusions is variable. Most have the typical crystal habit of rutile and are Nb-bearing. One TiO₂ inclusion is anhedral, rounded, and brownish orange in color; it differs from the other grains in that it does not contain any minor elements other than Fe above detection limit (Table 4). Rutile in blue corundum contains higher concentrations of V₂O₃ (0.20–0.25 wt.%) compared to rutile in a gray corundum (<0.03 wt.% V₂O₃). Rutile occurs in multiple other grains, including yellow-green coarser-sized corundum, but these were not surface exposed and analyzed.

Non-biotite phyllosilicate inclusion compositions are summarized in Table 5 and include primary Na- and F-bearing muscovite and the alteration minerals kaolinite subgroup and phengite, which may have formed at the expense of pre-existing inclusions.

Apatite in the blue corundum is F-dominant (fluorapatite) and contains 2.01 wt.% MnO, 0.18 wt.% Na₂O, and low concentrations of Y, Ce, Nd, and La oxides between 0.05 and 0.33 wt.% (Table 6).

Spinel-group inclusions observed in corundum are small in size and the EPMA spectra are slightly contaminated by corundum. The spinel (s.s.) inclusion in blue corundum has X_Mg = Mg/(Mg+Fe_Total) ≈ 0.85, while the hercynite inclusion in yellow-green corundum has X_Mg ≈ 0.22.

Zircon in the blue corundum varies in minor/trace element composition and Th/U in the oscillatory zoned core and the non-luminescent rim. Notably, the non-luminescent rim is richer in Th, U, Hf, Y, Ce, and P compared to the large, zoned core (Table 7). The Th/U ratios in the zoned core range between 0.06 and 0.68 and is 0.25 in the rim.

Alkali feldspar occurs exclusively in blue corundum grains, which were not found to contain plagioclase inclusions. Plagioclase on the surface of a blue corundum grain consists of oligoclase and andesine (average An₂₇; Table 8; Fig. 11). Plagioclase inclusions in colorless and yellow-green corundum are more variable in composition, ranging from andesine to labradorite (An₃₄₋₆₂). A plagioclase phenocryst in a pebble of rhyolite tuff varies in composition between oligoclase and andesine (An₂₆₋₃₉), which is chemically distinct from plagioclase associated with corundum: The tuff-hosted phenocrysts contain 0.14–0.31 wt.% SrO, while plagioclase associated with corundum has concentrations of less than 0.07 wt.%.

Biotite mica (Mg-rich annite to Fe-rich phlogopite) occurs as inclusions in colorless and yellow-green corundum and varies considerably in chemical composition, with X_Mg = Mg/(Fe_Total+Mg) of 0.40 to 0.85, TiO₂ concentrations from 0.69 to 1.85 wt.%, and F concentrations from <0.09 to 3.77 wt.% (Table 9). Biotite phenocrysts in a pebble of rhyolite tuff are very chemically distinct from all biotites found in corundum: The tuff-hosted biotite is very Al-poor due to Fe³⁺ substitution and is exceptionally rich in TiO₂ (4.01 wt.%).

The present study offers the first glimpse into the distribution of corundum grain sizes across all those sampled in a placer sapphire deposit. Importantly, this study also serves as the first benchmark for corundum concentrations contained in the <2 mm grain sizes in such a deposit. The abundance of fine-grained corundum in the Gem Mountain Mine ore is sufficiently elevated (∼50× more grains in the <2 mm fraction versus >2 mm) to indicate that it could be possible to detect a deposit by sampling more distal sediments, where fine-grained corundum may still occur in measurable quantities despite dilution by sediment from other sources, but where detecting coarser-grained corundum in a ∼10 kg sediment sample would be highly improbable. The more abundant fine-grained corundum could be a potential exploration tool, but data needs to be collected on sediments in the region to establish a background concentration and whether fine-grained corundum abundance correlates with the presence of larger grains of economic interest.

The predominance of inherited grain-surface textures, with only rare, fragmented grains and no observed abraded grains (see discussion below), together with the extremely poor sorting in colluvium, suggests that the grain-size distribution of corundum at Rock Creek is reflective of the corundum size distribution in the source rocks. Other deposits may have different size distribution profiles, which can reflect the original source and the effect of erosion/fragmentation/abrasion and sedimentary transport.

A regional till heavy-mineral survey in the Fury-Hecla Strait area, Baffin Island, Canada, resulted in the discovery of a till sample containing 208 transparent, colorless corundum grains (0.25–0.50 mm size fraction) per 10 kg of sediment (Tremblay 2021, Tremblay et al. 2020). The concentration of corundum in this sample is more than double that found in the colluvium sample, M1, examined in this study (∼90 grains [0.25–0.50 mm] / 10 kg colluvium). Other size fractions in the till sample were not examined for corundum. Such an elevated concentration relative to other till samples, such as those found on Baffin Island (i.e., Tremblay 2021), could be influenced by a high abundance of non-gem-quality corundum in eroded rocks (i.e., corundum that is heavily included, fractured, or of insufficient grain size, and therefore not of economic interest), but could also indicate potential for a gem deposit. The concentration benchmark established by the present study can be used to recommend follow-up research on such samples to identify potential new gem deposit greenfields, although more data is needed to make such assessments with a high degree of accuracy.

Corundum is considered of high gem quality when it occurs as monocrystalline crystals or grains with physical dimensions adequate (ideally >4 mm) for the production of polished gemstones. Additional corundum gemstone qualities include: (1) high transparency with attractive color, suitable for faceted gemstones; (2) translucent to opaque with attractive color, suitable for cabochon gemstones; and (3) translucent to opaque, preferably with attractive color, with strong asterism due to the presence of oriented rutile and/or hematite inclusions (e.g., Hughes 2017). These qualities are not always present in natural stones and can sometimes be achieved through physical and chemical treatment (Hughes 2017).

Red, pink, and blue corundum are by far the most valuable varieties in the gem trade (Hughes 2017), but pinkish orange (“padparadscha”), green, yellow, “parti-colored” (zoned multi-colored), and colorless corundum are also mined for use as gemstones (Hughes 2017, Wise 2016). The color of natural rough corundum can in many cases be modified by high-temperature heat treatment under oxidizing or reducing conditions, which can lighten the color of a darker blue sapphire, increase the purity of the red color in ruby by lightening blue hues, or turn colorless, greenish, or other light-colored corundum blue by reconstituting fine rutile inclusions into the corundum structure (Nassau 1984). In each of the latter examples, the treatment results in an increase in the market value of the stone. Diffusion treatment (Ti, Be; Hughes 2017) of sapphire can also improve its color, but stones treated using these methods do not have considerable monetary value. Therefore, detrital corundum need not be red, pink, or blue to demonstrate gem potential. If it is not naturally occurring in an attractive color of high market value, it should consist of an attractive but too dark color that could be lightened with treatment, or of a light-colored corundum that may react favorably to heat treatment.

Natural Rock Creek gem-quality sapphire is most commonly pale green or yellow-green, but also pink and blue, with rubies being extremely rare (Berg 2014). They are frequently heat treated to achieve blue or more saturated colors but are also sold as untreated stones (Zwaan et al. 2015, Berg 2014). While detrital corundum at Rock Creek is primarily colorless and yellowish-green, they are nonetheless of gem quality and have commercial value.

In sample M1, yellowish-green and greenish corundum grains are lighter in color—and colorless corundum comprises a larger proportion of the total—with decreasing size fraction. This is due to the shorter light path encountered in smaller grains, which results in a fainter observed color or even an apparent absence of color. Gem-quality corundum of medium and light color (see for, example, a natural 7.81 carat light blue sapphire from Nunavut, Canada; Belley et al. 2017) would appear light, very light, or perhaps even colorless in small grain sizes. Overly dark corundum with no gem-quality potential at adequate grain sizes (>2 mm) may appear lighter and more attractive in smaller, non-economic grain sizes, giving the false illusion of gemstone potential in grains discovered by heavy-mineral surveying. Therefore, an abundance of colorless corundum in the finer grained fractions of Rock Creek sediment is correlative with the gem quality of the coarser-grained samples. The light blue grains (<0.5 mm), however, may be considerably darker if they occurred as larger grains of potential economic interest, therefore their commercial value may rely on having a favorable reaction to heat treatment.

The color of corundum grains from heavy-mineral separates (0.125–2 mm) should be used to rule out gem potential rather than indicate it, i.e., dark and unattractive colors with low/no treatment potential reduces the potential for gem-quality corundum, while lighter colored detrital corundum does not.

Corundum with a natural, high degree of transparency and an absence of internal flaws (inclusions, fractures, etc.) is the most valued as a gem, but the clarity of corundum can also be enhanced by treatment. Although treatment of low-quality rough can increase the profitability of a deposit, the abundance of higher quality, transparent rough material is important to the overall profitability of most gem corundum deposits.

The challenge for detrital heavy-mineral exploration directed at discovering deposits of gem corundum is that this transparency and relative absence of inclusions must occur at dimensions suitable for producing faceted gemstones—in the coarser-grained sediment size fractions (>2 mm, ideally >4 mm). While the Rock Creek corundum samples in the present study have a high degree of transparency with relatively few inclusions in all size fractions, high transparency in <1 mm size fractions may not always reflect the properties of detrital corundum of economic interest (>4 mm) at a certain locality, if these large grains are present in a sedimentary system at all. Even in corundum that is heavily fractured in the size range of economic interest, smaller fragments of this material may show high degrees of transparency. Therefore, low-degrees of transparency (including consistent high abundance of fractures or inclusions) in detrital corundum grains could minimize gemstone potential. Instead, this potential would be indicated by high transparency in the majority of detrital grains, including coarser fractions, but transparency of corundum in the finer grained fractions does not indicate gemstone potential per se, although the high degree of transparency in fine-grained detrital corundum correlates with high transparency in the economic size fraction of corundum at Rock Creek. Interpreting detrital corundum transparency in different size fractions may be more difficult in alluvial deposits, where corundum is accumulated from a large region, and thus smaller corundum grains may not always have the same origin as the larger, gem-quality grains.

The Anaconda Bench placer at Rock Creek, where Gem Mountain Mine is located, has been interpreted to be a colluvial deposit and is unlikely to consist of till since the area was reportedly not glaciated in the Pleistocene (Berg 2014). The very poor sorting of the sediment and the morphological and surface characteristics of corundum grains (i.e., lack of abrasional surface features, occasional presence of conchoidal fractures, high angularity) are consistent with transport via mass wasting, typical in colluvium, and follows interpretations of the sediment genesis (Berg 2014, Barron & Boyd 2015). The alternate hypothesis put forth by Garland (2002), in which sapphire originates from the erosion of a Tertiary alluvial fan paleoplacer, is largely based on the interpretation of smooth, matte surface texture (Type 1 in the present study) as being caused by physical abrasion. However, the occurrence of corundum grains exhibiting a combination of Type 1 and Type 2 (i.e., glassy, hummocky) surfaces (see Fig. 4A–B) suggests that Type 1 texture is inconsistent with it having formed as a result of abrasion resulting from sedimentary transport. Further evidence against the formation of Type 1 surfaces by abrasion is provided by (1) light reflection from surface bumps at the same incident light angle, indicating the presence of small crystal faces which would suggest formation by dissolution or retrograde alteration (i.e., forming partially re-equilibrated basal faces at the grain boundary), rather than by physical abrasion; and (2) the occurrence of larger surface bumps that protrude from the grain surface, which are less likely to occur on an abraded surface (e.g., Fig. 7A; Appendix 5A, grain 7-2). The presence of adhered materials (i.e., biotite) with poor abrasion resistance and the abundance of corundum grains with sharp edges or protrusions (Types 2–4) are inconsistent with abrasion. The predominance of porous, partly kaolinitized rhyolitic tuff—which has very little resistance to erosion—in colluvium and the absence of abrasion features on softer monocrystalline grains (e.g., quartz, feldspar, apatite, zircon, allanite) provide further evidence in favor of a low degree of transport in a colluvial sedimentary environment rather than extensive transport in an alluvial sedimentary environment as suggested by Garland (2002).

The present study identified four types of corundum based on surface texture, which could be further grouped into two sub-populations: (a) corundum with pastel colors (green, yellow, colorless) having surface textures belonging to Types 1 to 3 and occurring in all size fractions and (b) blue idiomorphic corundum (Type 4) which was only observed in the <1 mm size fractions. At the Silver Bow sapphire occurrence (hosted in felsic-tuff-rich colluvium similar to that of the locality in the present study), which is located ∼80 km SE from Rock Creek, Berger & Berg (2006) identified two corundum sub-populations which resemble those described here: (a) pastel-colored (green, blue-green, yellow-green, yellow) corundum generally with rough surface texture and (b) small idiomorphic dark blue corundum. Berger & Berg (2006) identified both types of corundum in two xenoliths, a metamorphic spinel-bearing sillimanite-biotite-andesine rock and an igneous albite-biotite-glass rock, respectively.

The mineral inclusions in corundum and the corundum-associated minerals in composite grains identified in this study have been previously reported for this locality as well as for other sapphire deposits in Montana (see Palke et al. 2017 and the inclusion data compilation by Zwaan et al. 2015). In colorless and yellow-green corundum, biotite, rutile, and plagioclase (labradorite and andesine) inclusions predominate, while in the small idiomorphic blue corundum, the most common inclusions are alkali feldspar, rutile, ilmenite, and zircon. Zircon was noted in pastel-colored corundum by previous authors, so is not unique to the blue corundum variety. The predominant mineral inclusions in the pastel-colored (yellow, yellow-green, colorless) corundum—biotite and plagioclase—correlate with their proposed geological origin at the Silver Bow occurrence (∼80 km SE of Rock Creek), where pastel-colored corundum was found in a metamorphic sillimanite-biotite-andesine xenolith (Berger & Berg 2006). The blue-corundum-bearing xenolith found at Silver Bow is primarily composed of albite, glass, and biotite, which differs from the alkali-feldspar-rutile-ilmenite inclusion assemblage at Rock Creek. The presence of alkali feldspar is geologically compatible with an igneous albite-rich assemblage such as that described at Silver Bow, and its presence could suggest that blue corundum crystallized from aluminous melt at relatively high temperature.

Minerals that are present as intergrowths with corundum (i.e., can be described as composite grains) include plagioclase, K-feldspar, and biotite. The textural relationships between the intergrown minerals and the corundum suggest that they were, in some samples, probably in equilibrium with corundum (e.g., euhedral corundum in feldspar, poikilitic biotite containing euhedral corundum) and in others, not showing equilibrated grain boundaries (e.g., significantly embayed grain boundaries in a K-feldspar-biotite-corundum grain and plagioclase-corundum grain).

The high variability in biotite (X_Mg, Ti, F) and plagioclase (An%) chemical composition suggests that pastel-colored corundum originated from a highly heterogeneous source or from multiple proximal, geologically similar sources. The distinctive chemical compositions of biotite and plagioclase phenocrysts in the rhyolite tuff indicate that corundum did not form in the same environment as these phenocrysts.

The presence of adhered rhyolite on corundum noted by Berg (2014) and the present study, together with the persistent association of sapphire deposits with Tertiary volcanic rocks in SW Montana (Berg 2014, Berger & Berg 2006), suggest that felsic volcanic rocks are a likely source. However, these adhering materials are relatively rare, thus evidence remains somewhat limited.

The cross-cutting of oscillatory zones in some corundum grains of Type 1 and 2 demonstrate that these surface textures are the result of corundum volume loss subsequent to euhedral crystal growth. Berg (2014) attributed Type 1 surfaces on corundum grains to magmatic resorption. At the geologically similar Silver Bow sapphire occurrence, Berger & Berg (2006) noted the presence of hercynite and Fe-rich spinel coronae (reaction rims) on rounded “etched” corundum.

The rough surface texture (Type 2) could be explained, at least in part, by corundum dissolution concomitant with grain boundary migration of other mineral grains from the periphery of the corundum inward. The rare examples of composite corundum-bearing grains appear to show both equilibrated grain boundaries and embayed grain boundaries between corundum and biotite, K-feldspar, and plagioclase. It is possible that these minerals were removed from the surface of corundum grains by weathering or magmatic resorption, leaving an irregular surface texture. The grain boundaries between biotite and corundum in one grain bear a resemblance to rough corundum surface features.

The alteration of corundum to spinel group coronae (Berger & Berg 2006) and the embayment of corundum grain boundaries by associated minerals (present study) are possible mechanisms that can explain the loss in volume that Type 1 and Type 2 corundum grains have experienced. At the moment, it is not possible to determine whether these surface textures were created as a result of retrograde metamorphism in hypothetical metamorphic source rocks, metasomatism/alteration within xenolith fragments as a result of magmatic intrusions (i.e., supplying heat, fluids, or chemical potential gradients), from interaction with magma, or a combination of these factors.

The unknown source lithology of the Rock Creek corundum is a major impediment to the development of a genetic model for this mineral. However, evidence from grain shape and surface features, sedimentology, geological associations, inclusions, and mineral chemistry can be compiled from a number of studies to produce a new iteration of the working hypothesis for corundum genesis at Rock Creek. It should be emphasized that without knowing the bedrock source and without geochronological or P-T determinations for corundum formation, the ultimate origin of Rock Creek corundum remains speculative.

The occurrence of both corundum sub-populations (pastel and idiomorphic blue) at two felsic-tuff-rich colluvium deposits—Rock Creek and Silver Bow—separated by a distance of many tens of kilometers suggests a possible genetic connection. It can be speculated that Tertiary magmas, which intruded over a large area, were widely contaminated with aluminous metapelitic rocks that constitute the metamorphic basement rocks. The variation in biotite composition (X_Mg, Ti, Na, and F content) and plagioclase composition (andesine to labradorite) suggests considerable heterogeneity in source rocks, which could be reflective of either mixing of xenoliths/xenocrysts preceding eruption or mixing of corundum from distinct but geologically similar source rocks as a result of sedimentary transport. Primary melt inclusions in Montana corundum suggest a genesis linked to partially melted aluminous rocks (Palke et al. 2017). The inclusion and adhering mineral assemblages in pastel Rock Creek corundum are consistent with a biotite-sillimanite-andesine schist source rock, an assemblage observed in a xenolith at the Silver Bow occurrence (Berger & Berg 2006). As partially melted metapelitic origin would be consistent with the trace element signature of Rock Creek corundum, which is in the compositional range typical of metamorphic and metasomatic deposits (Garland 2002, Zwaan et al. 2015). It is impossible to know from the available data whether the genesis of the observed corundum arose via partial melting that occurred during regional metamorphism of basement rocks or from heating as a result of widespread magmatic intrusions, as suggested by Palke et al. (2017). As magma became contaminated with highly aluminous rock, blue idiomorphic corundum may have formed from locally Al-enriched magma. The pastel- and blue-colored corundum could then have been transported to the surface by felsic volcanic eruptions, as suggested by the presence of rare felsic volcanic adhering material (Berg 2014, present study) and the presence of trachytic, dacitic, and rhyolitic secondary inclusions in corundum (Palke et al. 2017). The volcanic corundum deposits would have subsequently been eroded and re-deposited into colluvial and alluvial deposits.

• A new comprehensive approach to detrital corundum characterization in placer deposits is proposed.

• Application of this method at Rock Creek has led to development of an extensive data set on numerous corundum grains recovered from a relatively small (∼10 kg) sediment sample.

• Corundum grains are significantly more numerous in smaller grain sizes (particularly 0.125–0.250, and 0.25–0.5 mm size fractions), and this is reflective of corundum size distribution in the source rocks.

• Due to its ∼50× higher abundance relative to coarser-grained corundum, sand-sized (<2 mm) detrital corundum would be easier to detect in smaller sediment samples, and thus is potentially useful for heavy-mineral exploration surveys.

• The characteristics of sand-sized detrital corundum can be used to assess gem potential, albeit with some limitations [e.g., they can be used to indicate or rule out gem potential based on color, transparency, and inclusion characteristics, but cannot be used to infer whether or not gem-quality corundum in grain sizes of economic interest (>4 mm) are associated with these sand-sized detrital corundum].

• Two distinct corundum sub-populations, based on differing physical properties and mineral inclusions, were recognized: yellow-green, green, and colorless corundum (3 types of surface textures/morphologies) containing biotite and plagioclase, among other inclusions, and blue, idiomorphic corundum (<1 mm) with a distinct suite of inclusions of (alkali-feldspar, ilmenite).

• Biotite and plagioclase inclusions in corundum show considerable chemical variability, which suggests a chemically heterogeneous source or multiple proximal, geologically similar sources

• The surface textures of the corundum grains examined are considered to be largely inherited from the original source rock. Surface features originating from sedimentary processes include uncommon conchoidal fractures and abrasion is absent, consistent with transport by mass wasting. Most corundum grains have smooth or rough surface textures that are the result of dissolution (likely of magmatic or metamorphic origin) subsequent to euhedral crystal growth. Idiomorphic crystals of blue corundum did not experience dissolution, based on their surface textures.

• The two sub-populations of corundum examined in this study resemble those found at the Silver Bow occurrence, tens of kilometers away, which suggests that the deposits, and the two corundum sub-populations, have common genetic connections.

The author thanks owners and staff from the Gem Mountain Mine for providing sediment samples collected according to special instructions, Dylan Goudie for support with SEM-MLA, Dr. Wanda Aylward and Glenn Poirier for support with EPMA, and Dr. Greg Dunning for fruitful discussions. The manuscript was significantly improved thanks to detailed comments from two anonymous reviewers, and the editorial contribution of Dr. Daniel Layton-Matthews (Associate Editor), Dr. Andrew McDonald (Editor), and Mackenzie Parker (Managing Editor).