The new mineral crowningshieldite: A high-temperature NiS polymorph found in a type IIa diamond from the Letseng mine, Lesotho

Nickel monosulfide (NiS) has two known polymorphs. Millerite, or β-NiS, with the trigonal space group R3m, is the polymorph more familiar to geologists. The second form is α-NiS, with the hexagonal space group P6₃/mmc and so-called NiAs-type structure, which is the stable form above 379 °C for pure NiS (Kullerud and Yund 1962; Sowa et al. 2004). Here, we describe the first known natural occurrence of α-NiS, the new mineral named crowningshieldite, with an empirical formula (Ni_0.90Fe_0.10)S. It was found as part of a multiphase inclusion assemblage in a diamond from the Letseng mine, Lesotho. This inclusion is likely altered from its primary mineralogy. The diamond was investigated previously for its (unaltered) metallic Fe-Ni-C-S inclusions and belongs to a newly recognized variety of sublithospheric diamonds called CLIPPIR diamonds (Smith et al. 2016, 2017).

Many occurrences of natural millerite appear to have formed in relatively low-temperature geological settings, during hydrothermal alteration, and develop a prominent acicular habit. However, some examples of massive NiS in magmatic sulfides, at Norilsk, Russia, and Temagami, Canada, for instance, are suspected to have formed originally as crystals of α-NiS (crown-ingshieldite) that inverted to millerite upon cooling (Borovskii et al. 1959; Kullerud and Yund 1962). Synthetic α-NiS can be quenched to room temperature, but it is metastable and transforms to β-NiS, typically over months depending on temperature and composition (Kasper et al. 2000; Wang et al. 2006). Interestingly, this α-NiS to β-NiS transition is of concern for the glass industry, because float glass can contain inclusions of NiS particles. Heat treatment and rapid cooling to produce tempered or toughened glass can quench these inclusions as α-NiS. Later, after cooling, stresses arising as α-NiS inverts to the ~3% larger β-NiS phase may lead to localized micro-cracking that initiates spontaneous glass failure (Barry and Ford 2001; Swain 1981).

The new mineral is named in honor of G. Robert Crowning-shield (1919–2006). He was a central figure in research at the Gemological Institute of America (GIA) for more than 50 years, helping it to become a place of unique research opportunities such as the present mineral discovery. Crowningshield advocated for the use of spectroscopy as a valuable tool in gemology and meticulously recorded the absorption patterns of many transparent gems by hand, ultimately publishing them as a key reference for gem mineral identification. In 1956 he described a reliable spectroscopic method to identify yellow irradiated diamonds and discriminate them from naturally colored yellow diamonds, thereby resolving a serious concern in the gem trade at the time (Crowningshield 1957–1958).

The mineral and the name crowningshieldite have been approved by the Commission on New Minerals, Nomenclature and Classification (IMA2018-072). The holotype specimen described herein has been deposited at the Mineralogical Museum of the University of Padova, Italy (catalog number: MM 20501). A second, smaller piece of the original specimen is deposited in the collections of the Gemological Institute of America Museum, Carlsbad, California, U.S.A. (catalog number: 41800).

The Letseng mine in Lesotho has a kimberlite eruption age of 85.5 ± 0.3 Ma, based on U-Pb dating of groundmass perovskite (Stamm et al. 2018). The diamond in which crowningshieldite was discovered (Fig. 1) was studied previously by Smith et al. (2016) under the sample name Letseng_889 (here shortened to sample 889). It is a 2.14 carat (0.428 g) inclusion-bearing offcut, trimmed from a larger type IIa (no detectable nitrogen by infrared spectroscopy; less than about 5 ppm) colorless rough diamond during the gem manufacturing process. The size of the larger parental rough diamond is unknown, but probably had dimensions exceeding 1 × 1 × 1 cm and a weight >10 carats, based on the size of the offcut (Fig. 1). Remnant rough surfaces on the offcut are strongly resorbed and suggest an irregular morphology, as opposed to a symmetrical octahedral or dodecahedral diamond habit. These features are typical of the type IIa gem diamonds regularly found at Letseng (Bowen et al. 2009), as well as of other high-quality type IIa diamonds recovered globally. Such diamonds are collectively termed CLIPPIR diamonds (Cullinan-like, Large, Inclusion-Poor, Pure, Irregular, Resorbed). This sample, 889, is part of a larger suite of diamonds that was used to investigate the geologic origin of CLIPPIR diamonds (Smith et al. 2016, 2017).

In the previous investigations (Smith et al. 2016, 2017), which sought to document inclusions in CLIPPIR diamonds, the focus was on the nature of their recurring metallic Fe-Ni-C-S melt inclusions that are predominantly made up of cohenite [(Fe,Ni)₃C], Fe-Ni alloy, and pyrrhotite. The average bulk composition is estimated as (Fe_60–80Ni_8–15C_10–16S_4–14) (Smith et al. 2016). These are the most abundant kind of inclusions in CLIPPIR diamonds, with the metallic melt being observed in 67 of 83 inclusion-bearing diamonds (Smith et al. 2017). Inclusions of breyite (CaSiO₃) plus other Ca-silicates, together interpreted as the breakdown products of former CaSiO₃-perovskite, along with inclusions of majoritic garnet indicate that CLIPPIR diamonds are sublithospheric (super-deep) in origin, derived from a depth of about 360–750 km (Smith et al. 2016, 2017).

The offcut studied here (sample 889) contained more than 10 primary metallic Fe-Ni-C-S inclusions (see Fig. 1c) of varying size, two of which were shown to contain the cohenite, Fe-Ni alloy, and pyrrhotite assemblage (Smith et al. 2016, 2017). These metallic inclusions have methane (CH₄), detectable with Raman spectroscopy, that is trapped at the walls between the solid inclusion and the surrounding diamond host (Smith et al. 2016). The methane confirms these inclusions are fully enclosed in the diamond. Any fracture reaching from the inclusion to the diamond exterior would allow the methane to leak out. It is important to note that these pristine metallic Fe-Ni-C-S inclusions do not have any reddish iron-oxide staining associated with them.

In contrast, the largest inclusion in sample 889 had a large, reddish, iron-oxide stained fracture extending between the inclusion and the exterior surface of the diamond (Fig. 1c). Based on the fracture, it is strongly suspected that this inclusion has been exposed to fluids that have altered its primary mineralogy, possibly during the time of kimberlite eruption. It was in this largest inclusion that crowningshieldite was discovered, by X-ray diffraction analysis. Crowningshieldite occurs as part of a multiphase assemblage along with magnetite-magnesioferrite, hematite, and graphite. Given the reddish stained fracture, the crowningshieldite-bearing inclusion is interpreted to be an altered, partly oxidized version of the metallic Fe-Ni-C-S melt inclusions described above. The as-received diamond offcut (sample 889) was already mostly broken open along the large fracture, such that the fracture surface was mostly exposed and the crowningshieldite-bearing inclusion was also partly exposed along its side (Fig. 1b). The surface of the exposed inclusion was tested with a multimeter and found to be electrically conductive, suggesting graphite interconnectivity.

The original 2.14 carat offcut (Fig. 1) has been trimmed down further for other ongoing studies of the Fe-Ni-C-S inclusions. The remaining diamond fragment shown in Figure 2 is the holotype sample for crowningshieldite, from which high-quality diffraction data was collected, and from which the composition was measured on a flat surface. All analytical work presented here focuses on this holotype sample (Fig. 2).

The multiphase inclusion in sample 889 (Fig. 2) is reasonably flat and smooth because it sits flush with the polished surface of the adjacent diamond host. This flat exposed surface, still embedded within its diamond host, is shown in secondary electron (SE) and backscattered electron (BSE) images in Figure 3. Crowningshieldite occurs as small grains, 10–20 μm, along with magnesioferrite and graphite. Due to the fine-grained and fractured nature of the crowningshieldite it was not possible to isolate a single crystal for measurement of optical and physical properties. Also, the inclusion material could not be safely removed from the diamond host to polish it separately, due to its rarity and probable fragility. Therefore, a well-polished crowningshieldite surface was not available for reflected-light microscopy.

A nickel sulfide phase was also found within a second Letseng diamond (sample Letseng_893) by SEM-EDS during a prior preliminary investigation. This nickel sulfide is presumed to be crowningshieldite, but diffraction data could not be obtained to confirm its crystal structure, so it is not discussed further. Again, a fracture connected this particular inclusion to the diamond's exterior surface and, accordingly, it had no detectable methane fluid jacket by Raman spectroscopy.

It is proposed that when cracks develop in the host diamond, connecting metallic Fe-Ni-C-S inclusions to the exterior of the diamond, they allow fluids to interact with the inclusion and alter it to a variable assemblage including iron oxides, iron-bearing nickel sulfide, and graphite. It is unclear when alteration of the original inclusion occurred, but kimberlite volcanism and eruption is considered the most likely timing, when the diamond would have been subject to stresses that may have caused cracking and also exposed to volatiles/fluids at elevated temperatures.

Graphite geothermometry, while more often used for determining peak metamorphic temperatures of carbonaceous rocks, may provide a sense of the temperature conditions during the formation of this crowningshieldite-bearing assemblage. Using Raman spectroscopy to assess graphite crystallinity affords an estimate of the peak temperature that graphite was exposed to (Beyssac et al. 2002). It is not an ideal tool for this purpose, but given the simplicity of the approach, graphite geothermometry is applied here as a qualitative indicator under the assumption that graphite and crowningshieldite formed at comparable conditions (during a single episode of alteration). Graphite associated with crowningshieldite has a Raman spectrum (Fig. 4) with D1, G, and D2 peak area ratios corresponding to a temperature of 357 ± 50 °C, according to the geothermometer of Beyssac et al. (2002). In an iron free system, the NiS inversion temperature between β (millerite) and α (crowningshieldite) ranges from 379 °C for stoichiometric NiS down to 282 °C for more nickel deficient compositions, Ni_1–xS with x reaching 0.09 (Kullerud and Yund 1962). Taken qualitatively, the graphite geothermometry for this crowningshieldite-bearing inclusion suggests alteration took place at a temperature a few hundred degrees or more above ambient surface temperatures, which is consistent with the interpretation that alteration occurred prior to the complete cooling of the kimberlite after emplacement. However, although the temperature overlaps the experimental stability field for α-NiS (crowningshieldite) formation, it does not provide an independent confirmation that alteration in this assemblage occurred at a higher temperature than the millerite stability field.

In general, the crowningshieldite-bearing inclusion still retains its proposed original iron-, nickel-, carbon-, and sulfur-rich composition but is more oxidized, from cohenite–Fe-Ni alloy–pyrhhotite to magnesioferrite–hematite–graphite–crown-ingshieldite. This latter mineral association may be the typical outcome for metallic inclusions with surface-reaching cracks in CLIPPIR diamonds. If such particular circumstances are needed for the formation of crowningshieldite, it is understandable that it has not been found previously. CLIPPIR diamonds are rare, making up only about 1% of gem diamonds worldwide, and often carry a high gem value, so they are challenging to access for research and have only recently been examined in detail (Smith et al. 2016, 2017).

Crowningshieldite is opaque, with a metallic luster, and bronze yellow color. Its occurrence as small anhedral grains in the only available sample (Fig. 3) prohibited direct determination of some properties. The calculated density of crowningshieldite is 5.47(1) g/cm³, based on the empirical formula and refined unit-cell parameters, which is slightly greater than the 5.38 g/cm³ density of millerite, as predicted from the more compact structure of crowningshieldite compared to millerite. Crowningshieldite is predicted to have a hardness, between 3 and 3.5 on the Mohs scale, similar to millerite, based on the crystal-chemical similarity between crowningshieldite and millerite.

Attempts to measure the Raman spectrum of crowningshieldite proved unsuccessful. The spectra recorded were that of millerite, even at minimal laser power and long integration times, despite the fact that no traces of millerite were found by X-ray diffraction analysis of this multiphase inclusion assemblage. The unavoidably poor surface polish on the crowningshieldite, as it is still embedded in diamond, could be a reason for the Raman observations. Mechanical polishing of ore minerals has been shown to affect Raman spectra, in some cases leading to loss of crystallinity in the surface layer to perhaps tens of nanometers deep (Libowitzky et al. 2011). Another possible explanation is that the laser excitation caused inadvertent localized, surficial transformation of crowningshieldite to millerite at the point of analysis. The Raman spectrum of Bishop et al. (1998) that is reportedly for synthetic α-NiS (crowningshieldite), which is the only published spectrum for α-NiS, also appears to suffer from this problem where all major spectral features essentially correspond to millerite and do not coincide with calculated α-NiS Raman modes (Goel et al. 2019). Raman analysis of millerite alone shows even modest laser power can degrade the sample surface, causing broadening and shifting of Raman bands (Guillaume et al. 2008). Laser induced changes like this are a relatively common difficulty encountered for Raman spectroscopy of opaque materials susceptible to oxidation or structural transformation (e.g., de Faria et al. 1997).

Given that an ideally polished crowningshieldite surface could not be prepared without risking its destruction, it was not possible to perform quantitative electron microprobe analysis. However, the surface sitting flush to the polished diamond surface was sufficiently flat and smooth to permit semi-quantitative analysis by energy-dispersive X-ray spectroscopy (EDS), electron (BSE) detector, and an EDAX EDS system for micro-analysis, at the Department of Geosciences at the University of Padova. Analysis was performed with a 20 kV accelerating voltage, ~13 nA filament current, and 27 mm working distance. The non-ideal crowningshieldite surface is the largest source of uncertainty, which cannot be rectified with the use of standards. However, this instrumental setup has shown accurate results, compared to known compositions, for analysis of other minerals with poorly polished surfaces and relatively simple compositions, like NiS.

Figure 5 shows the EDS spectrum of crowningshieldite, constituted only by Ni, Fe, and S. No other elements were detected. Cobalt, which is often present as a minor component in millerite, was not detected, even with extended counting times. Although the crystal size is limited, the relatively simple nature of the chemical composition permitted its reliable determination by EDS, based on four spot analyses (Table 1). The empirical formula calculated from EDS data on the basis of one sulfur atom per formula unit is (Ni_0.90Fe_0.10)S. The simplified formula is NiS, requiring 64.67 wt% nickel and 35.33 wt% sulfur.

The crystal structure of crowningshieldite was determined using X-ray powder diffraction methods, in situ, due to the nature of the sample. Single-crystal diffraction was not possible due to its limited crystal size and micro-texture. The instrumentation consisted of a Rigaku-Oxford Diffraction Supernova single-crystal diffractometer (Kappa-geometry) equipped with an X-ray micro-source (MoKα) with 120 μm beam spot, operating at 50 kV and 0.8 mA, and a 200 K Pilatus detector (Dectris), at the Department of Geosciences at the University of Padova. Data were collected in powder diffraction mode. The multiphase inclusion (Fig. 3), exposed to air, was centered under the X-ray beam and X-ray data were collected in φ scan mode. The X-ray data were converted to a format readable by HighScore Plus software (Panalytical) to examine and plot the data.

Figure 6 shows the X-ray diffraction results, indicating that a polycrystalline mixture of phases makes up the inclusion. The pattern was subsequently evaluated by Rietveld methods (Table 2), which made it possible to quantify the relative amount of each phase present in the analyzed inclusion volume: magnesioferrite 57%, crowningshieldite (labeled NiS) 28%, graphite 10%, and hematite 6%. No traces of millerite were recorded. To emphasize the clear differences between crowningshieldite and millerite, Figure 7 shows the X-ray diffraction pattern again, with only the labels for crowningshieldite and added reference peaks for millerite.

Rietveld analysis for crowningshieldite (Table 3), using the starting model given by Rost and Haugsten (1969), gave reasonable values of R_Bragg = 2.34% and R_WP = 4.21%. The simple structure of crowningshieldite permitted refinement of B_iso, but attempts to refine the occupancies failed as the X-ray scattering power of Ni and Fe are virtually the same (Table 4). Overlap between the peaks of crowningshieldite and other phases prohibits a reliable estimation of standard deviation for refined occupancies and B_iso values. Therefore, occupancies were fixed based on EDS chemical analyses and B_iso values were fixed to 0.5 Ų. Considering the low R_Bragg obtained, such a chemical constraint appears to be reliable, which reinforces the conclusion that the composition obtained by EDS analysis is robust. Table 5 lists the d-spacings of crowningshieldite and, for comparison, the published d-spacings of millerite and synthetic α-NiS are given in Table 6.

Crowningshieldite follows the archetype structure of nickeline, NiAs (Wyckoff 1963). A CIF is on deposit¹. Each Ni atom is bound to six S atoms, with a bond length of 2.397(2) Å. Its unit cell is shown in Figure 8 along with a visualization of NiS₆ octahedra, which share faces along the c axis. Crowning-shieldite and nickeline have similar unit-cell volumes, 55.0(2) and 56.3 ų, respectively, but their unit-cell parameters are significantly different owing to the different ionic radius of S compared to As (Table 7). In crowningshieldite the two groups of six S–S distances are somewhat similar, being 3.443(3) and 3.336(4) Å, whereas in nickeline, the two groups of six As-As distances are more dissimilar, being 3.602 and 3.255 Å. Their orientations with respect to the a and c axes cause a longer a axis for crowningshieldite and a longer c axis for nickeline.

Compared to crowningshieldite, where the Ni atoms are sixfold-coordinated, millerite has Ni atoms that are fivefold-coordinated, and the space group changes to R3m. The fivefold-coordinated Ni in millerite is relatively uncommon in nature.

Based on previous experiments, quenched samples of α-NiS (crowningshieldite) are metastable and eventually tend to revert to β-NiS (millerite). For this reason, the long-term stability of the holotype specimen is somewhat uncertain. For example, Wang et al. (2006) observed that a stoichiometric pure synthetic crowningshieldite stored for six months at room temperature exhibited 3–4% transformation to millerite. As suggested by Kullerud and Yund (1962), massive millerite in some magmatic sulfide deposits is suspected to have crystallized initially as crowningshieldite (α-NiS) and subsequently inverted to millerite during cooling. It is therefore noteworthy that crowningshieldite in the present sample has remained preserved on a geological timescale, although the precise mechanism for its preservation is not understood. Potential factors are discussed below.

The possibility that the observed crowningshieldite might have been generated during sample preparation as a result of heat, by transforming millerite above ~379 °C, is ruled out for two reasons. First, the crowningshieldite-bearing multiphase inclusion was not exposed to any such high temperatures. The sample surface in Figure 2 has been lightly polished after the diamond was cut, which may have led to some limited local heating at the interface, but the crowningshieldite occurs throughout the bulk of the inclusion, most of which was protected by diamond during these processes (Fig. 1). The second reason is that the polishing and preparation work was actually completed two years prior to the present analyses. The fact that no traces of millerite were detected after this period (Figs. 3 and 7) reinforces the notion that this crowningshieldite is reasonably stable. Two potential factors that might contribute to the apparent stability are composition and confining pressure. In terms of composition, it is possible that there is some trace element not detected in EDS that stabilizes the structure. Another possible compositional effect could stem from the Fe content in this crowningshieldite, Ni_0.9Fe_0.1S (5 at% Fe). This Fe content is relatively high compared to typical millerite Fe contents, with the maximum measured in synthetic millerite being 5.5 at% at 230 °C and in natural millerite being below 1.6 at% (Misra and Fleet 1973). Substantial incorporation of Fe into crowningshieldite is not unexpected, given that, in the NiAs-type structure, there is complete solid solution between Fe_1–xS and Ni_1–xS (i.e., monosulfide solid solution, or MSS) above approximately 400 °C (Misra and Fleet 1973; Ueno et al. 2000). The evident maximum Fe content of millerite, however, might suggest a limit to the ease with which Fe can be accommodated during the α-NiS to β-NiS transition, which involves structural reorganization as the Ni coordination changes from sixfold to fivefold, respectively (Wang et al. 2006). Experiments clearly demonstrate that even small amounts of Fe slow the kinetics of the transformation of α-NiS (crowningshieldite) to β-NiS (millerite) (Kasper et al. 2000; Yousfi et al. 2010). Further work is needed to assess the effect of Fe to slow or perhaps present an energy barrier for the transformation of α-NiS (crowningshieldite) to β-NiS (millerite).

An alternative factor that may be relevant for the preservation of the examined crowningshieldite has to do with the confining pressure of the host diamond. Some inclusions in diamond have substantial remnant confining pressure, for example in the case of diamond containing ringwoodite with a residual pressure of ~2 GPa that has prevented its inversion to olivine (Pearson et al. 2014). Experiments show that with increasing pressure, the temperature stability field of α-NiS increases (Sowa et al. 2004), suggesting that pressure may help stabilize crowningshieldite. The actual pressure within the examined crowningshieldite-bearing inclusion (prior to being exposed by any cutting) is not known, but might not have been particularly high due to its relatively large size and the large natural fracture around it. Nevertheless, pressure within the inclusion may be a relevant consideration for crowningshieldite preservation. Ultimately, the factors that have inhibited crowningshieldite from transforming to millerite are not known in this case, but these conditions are likely rare in nature.

With the ongoing study of Fe-Ni-C-S inclusions in CLIPPIR diamonds, further examples of altered inclusion assemblages containing crowningshieldite may yet be found. The present finding illustrates not only that the α-NiS polymorph occurs in nature, but also that particular circumstances can permit its long-term preservation. Inclusions in diamonds are unique among geological materials accessible at Earth's surface. Several otherwise exotic minerals, such as ringwoodite (Pearson et al. 2014) and CaSiO₃-perovskite (Nestola et al. 2018), have been found preserved within them. The discovery of crowningshieldite shows even altered inclusions can host unexpected phases.

Sincere thanks go to A. McDonald and F. Kaminsky for their constructive reviews. The authors also acknowledge K. Smit and A. Palke for helpful feedback on this manuscript, T. Gore for valuable discussion, and T. Moses for help with sample acquisition with support from the Gemological Institute of America. F.N. was supported by the European Research Council supported FN (INDIMEDEA, no. 307322).