Carbonado diamond is found only in Brazil and the Central African Republic. These unusual diamond aggregates are strongly bonded and porous, with melt-like glassy patinas unlike any conventional diamond from kimberlites-lamproites, crustal collisional settings, or meteorite impact. Nearly two centuries after carbonado’s discovery, a primary host rock compatible with the origin of conventional diamond at high temperatures and pressures has yet to be identified. Models for its genesis are far-reaching and range from terrestrial subduction to cosmic sources.
Discovered in 1841 in Brazil, carbonado was named by Portuguese diamond prospectors for its resemblance to charcoal (Leonardos, 1937; Dominguez, 1996). Carbonado was found later in the Central African Republic. These two localities, now separated by the Atlantic Ocean and situated on the São Francisco and the Congo cratons, respectively, previously shared a common geological setting for more than a billion years (De Waele et al., 2008) on the supercontinent of Rodinia (figure 1) and its precursor Nuna, also known as Columbia.
Carbonado was prized by the French as a superior polishing material. It was used for drilling during the construction of the Panama Canal and formed part of the U.S. strategic mineral stockpile as recently as 1990. At the height of alluvial mining in Brazil (1850–1870), some 70,000 carats were produced by an estimated 30,000 artisanal miners (Svisero, 1995). A conservative estimate of the recovery from Brazil and the Central African Republic is approximately 2 metric tons (Haggerty, 2014). Four of the five largest diamonds reported from Brazil, ranging in weight from 726 to 3,167 ct, are carbonado (Svisero, 1995). The largest of the five, the Sergio, recovered in 1905, is 61 ct heavier than the largest single-crystal diamond ever reported (the 3,106 ct Cullinan rough).
While earlier investigations of carbonado focused on physical and chemical properties and synthesis, more recent studies have introduced dating techniques, high-resolution microscopy, spectroscopy, and an emphasis on origin (see Haggerty, 2014, for a more comprehensive view). The present study offers a detailed examination of about 800 carbonados from Brazil and the Central African Republic (figure 2), ranging from <1 to 730 ct. These samples showed no significant differences in their texture, superficial appearance, and physical and chemical properties (Haggerty, 2014). This article describes the unusual textural features of carbonado, namely their pores and the presence of glassy diamond as a surface patina, with the aim of assessing the origin of carbonado.
CHARACTERISTICS OF CARBONADO
Carbonado is typically found in five major size categories: >200 ct, 75–95 ct, 25–35 ct, 8–15 ct, and 0.25–1.25 ct (see figure 5 of Haggerty, 2014). Sand-sized particles (<1 mm) also occur, and melon-size objects larger than the Sergio are reported but unconfirmed (M. Ozwaldo, pers. comm., 1996). Carbonados are typically equidimensional (in millimeter to centimeter sizes), although some are elongated (figure 3); they are seldom rounded.
Carbonado is opaque, composed of randomly oriented diamond crystallites that impede light refraction and increase absorption. Color varies from black and putty gray to shades of brown (figure 3), deep purple to pink, rusty red, and the occasional olive green. Pores (figure 4), an unusual glassy patina (figure 5), highly irregular surfaces (figures 6 and 7), and polycrystallinity (figures 8 and 9) distinguish carbonado from conventional diamonds.
Porosity. As the porosity of an object increases, its apparent density decreases, because the voids take up more and more of the volume. In carbonado, the number of exposed microdiamond cutting points increases with porosity. This was a sought-after property that made carbonado more expensive by weight than diamond at the turn of the twentieth century (Haggerty, 2014). Densities as low as 2.8 g/cm3 and as high as 3.45 g/cm3, with most around 3.05 g/cm3 (Trueb and De Wys, 1969; Haggerty, 2014), are in contrast to gem diamond at 3.52 g/cm3. Calculated pore concentrations vary between 5% and 15% in volume. The pores persist into the interior of the carbonado and are either spherical or oblate. Some are inferred to be interconnected (Ketcham and Koeberl, 2013), but the material’s permeability is very low because the pores are free of infiltrating hydrothermal precipitates that abound in surface pores (again, see figure 4). The spherical pores in carbonado are unlike those in other polycrystalline diamond such as framesite, where the open spaces are at adjoining crystal faces and the shapes are irregularly polyhedral. In other polycrystalline diamonds, the open spaces are microns in width and either radial (in non-gem-quality ballas) or parallel (in fibrous cubes).
Patina. In carbonado, patina surfaces are pervasive (figure 5). Pores in contact with surface patinas are reduced in size, and at 50× magnification they can no longer be distinguished. Glass-like in appearance and similar to synthetic carbon glass (de Heer et al., 2005), these veneers may be dimpled or furrowed, with mounds and flow structures (figures 5 and 6). These textures are akin to those seen in melts in volcanic rocks or in slags from metal processing. But in carbonado, the veneers are diamond that appear to have formed directly from the underlying porous substrates, although diamond coating at a later time is also possible. Contact boundaries between pore-present and pore-absent surfaces are poorly defined, except in cases where patina crusts have splintered off where the contact is sharp, as seen in the lower-right images of figure 6 and in figure 7. Secondary pits and microcraters are pervasive and, in many cases, younger than the patina (figure 7). While pores tend to have sharp outlines (figure 4), craters are rounded with bubbly surfaces or rimmed by smooth ridges (again, see figures 6 and 7). The evidence of flow in both types of voids implies differences in origin. Solid melt marbles are typical. Microcraters, free of ornamentation, grade into texturally soft plastic walls (figures 5 and 6). Slickensides, the striated surfaces known to form on rocks that have been forced to slide along a fracture surface at high pressure as in a fault (figure 7), are of interest because these could only have developed on frictional contact with a body whose hardness was equivalent to another diamond. On the other hand, the patina itself may represent frictional melting (e.g., de Heer et al., 2005; Mitchell et al., 2016; Shumilova et al., 2016a,b; Shiell et al., 2016). Standard diamond testers that measure thermal conductivity give a sharp response to glassy diamond surfaces, less so to the ridges and mounds. The pore-rich surfaces are distinctly sluggish and erratic in response, possibly due to crystal discontinuities of microdiamond grain boundaries.
PHYSICAL AND CHEMICAL PROPERTIES
Octahedra, dodecahedra, tetrahexahedra, and fibrous cubes, all typical of conventional diamond (e.g., Orlov, 1977), are not observed in carbonado. Polycrystalline cubes measuring 5 to approximately 20 µm are common. Encased in very fine diamond (<1–5 µm), the matrix is tightly fused with angular interstices and rounded pores (figure 8). Scanning electron microscopy (SEM) images illustrate the distribution of diamond cleavage surfaces, hopper crystals, skeletal crystallites, re-entrant intergrowths, and layers in single crystals in the open-space pores of carbonado (figures 8 and 9). Trueb and De Wys (1969) and Petrovsky et al. (2010) suggest that the closest analogy to carbonado textures is in synthetically compressed nanodiamond aggregates. Because these structures are found in pores, a more reasonable comparison is with vapor deposition of diamond. The preferred crystal habit of these diamonds is cuboidal, either as single solid cubes or as interpenetrating twins on  that follow the fluorite twin law (figure 8). The solid cubes are colorless and, although fine grained, appear to be translucent. Diamond cubes and cuboctahedra are routinely synthesized in metallic catalysts at high pressure and temperature (Burns and Davies, 1992) or by chemical vapor deposition (CVD) under high vacuum and at plasma temperatures (Sato and Kamo, 1992).
X-ray diffraction (XRD) data on crushed carbonado grains are similar to conventional diamonds. Hardness is also similar, but there are data indicating that carbonado is slightly harder (Haggerty, 2014). Its toughness and tenacity, stemming from the random orientation of microdiamonds, are clearly superior to monocrystalline gem diamond, to the point that carbonado can only be cut by lasers.
Yet another unusual feature of carbonado is the presence of an exotic array of metals (Fe, Ni, Cr, and Ti), metal alloys (Fe-Ni, Fe-Cr, Ni-Cr, and W-Fe-Cr-V), and very unusual minerals, specifically moissanite (SiC) and osbornite (TiN). These phases occur as primary intergranular inclusions or as crystal-controlled oriented intergrowths. They are only stable at the very low oxidation states (Gorshkov et al., 1996; De et al., 1998; Makeev et al., 2002; Jones et al., 2003) that would occur deep within Earth’s mantle or other reducing environments such as outer space. By contrast, surface pores and fractures are filled by secondary, low-temperature minerals such as quartz and highly oxidized magnetite, goethite, florencite, and goyazite (Trueb and Butterman, 1969), typical of a more oxidized terrestrial surface growth environment.
Relative to mantle-derived diamonds, carbonado is isotopically light, with δ13C = –24 to –31‰ (Ozima et al., 1991; Shelkov et al., 1997; De et al., 2001). Nitrogen concentrations are low (~20 to 500 ppmw), and δ15N ranges from –3.6 to 12.8‰ with an average of 3.7‰ (Shelkov et al., 1997; Vicenzi and Heaney, 2001; Yokochi et al., 2008). The coupled isotopic distribution of C and N shows that the compositional field for carbonado is distinctly different from that of conventional diamonds (figure 10).
Figure 11 shows photoluminescence (PL) spectra of carbonado, which are similar to those of irradiated and heated CVD diamond (Clark et al., 1992). The characteristic peaks at 1.945 eV and 2.156 eV are attributed to nitrogen vacancy (NV) defects in type Ib diamonds. Wang et al. (2009) report a substantial amount of nonaggregated N in type Ib diamonds with H2 and H3 defects. Hydrogen-containing defects (H1) and NV defects are also reported by Nadolinny et al. (2003).
Cathodoluminescence of large (approximately 200 µm) monocrystals of diamond in carbonado exhibit orange and green tones (Magee and Taylor, 1999; De et al., 2001; Yokochi et al., 2008). However, blue luminescence in large diamonds, embedded in an orange luminescent matrix of submicron diamond, are also reported (Rondeau et al., 2008). The range in colors is attributed to various N-V (nitrogen-vacancy) defects.
Synchrotron infrared measurements of carbonado have shown the presence of single nitrogen (type Ib) substitution and hydrogen (Garai et al., 2006), in contrast to aggregated N typical of conventional type Ia diamonds that have undergone prolonged high P-T annealing in the mantle.
Carbonado has been dated by Ozima and Tatsumoto (1997) and Sano et al. (2002), on samples derived from conglomerates (again, see figure 2) that have been reworked over a period from at least 1.7 Ga to approximately 3.8 Ga (Pedreira and De Waele, 2008). It is relevant to note that, unlike the dating of conventional diamond, which is based on trapped mineral inclusions (garnet, pyroxene, and sulfides), the age of carbonado discussed in this review was determined directly on diamond. Following a robust chemical protocol of acid dissolution to remove all nondiamond material, the cleansed carbonado was subjected to two different instrumental methods of analyses. Ozima and Tatsumoto (1997) used high-resolution mass spectrometry on carat-sized samples from the Central African Republic, while Sano et al. (2002) employed an ion probe that allowed for micron-sized spot analyses on larger samples from Brazil. Both studies report ages of 2.6–3.8 Ga on implanted radiogenic lead. Although this method of age determination is unconventional, it is important to note that the Archean result is consistent with trapped crustal inclusions (Sano et al., 2002) of zircon (1.7–3.6 Ga), rutile (3.9 Ga), and quartz (3.2 Ga), and with the antiquity of the basement in the São Francisco craton, which is 3.3–3.7 Ga (Barbosa and Sabate, 2004).
In summary, the chemical and physical characteristics of carbonado point to marked similarities with rapidly quenched type Ib diamonds and CVD diamond, both of which contain significant hydrogen. But there are also major differences: carbonado has pores and patinas with distinctions in C and N isotopes, an absence of mantle minerals, and the presence of exotic metal inclusions. Carbonado is unquestionably one of the most unusual forms of diamond ever reported. Because it has never been found in typical diamond-bearing rocks, the many proposed origins are varied, and none are uniformly accepted.
Theories on the genesis of carbonado fall into five categories:
- Meteoritic impact (Smith and Dawson, 1985)
- Growth and sintering in the crust or mantle (Burgess et al., 1998; Ishibashi et al., 2012; Chen and Van Tendeloo, 1999; Heaney et al., 2005; Kagi and Fukura, 2008; Ketcham and Koeberl, 2013)
- Subduction (De Carli, 1997; Irifune et al., 2004)
- Radioactive ion implantation of carbon substrates (Kaminsky, 1991; Ozima et al., 1991; Shibata et al., 1993; Kagi et al., 1994; Daulton and Ozima, 1996; Ozima and Tatsumoto, 1997)
- Extraterrestrial (Haggerty, 1996, 2014)
Meteoritic Impact. This model was based on a correlation with the Bangui magnetic anomaly in the Central African Republic. Originally thought to be a buried iron meteorite, it was subsequently shown to be a crustal-derived banded iron ore body (Regan and Marsh, 1982), similar to the magnetic anomaly and giant iron ore deposit in Kursk, Russia (Taylor et al., 2014). Because the C-isotopic composition of carbonado is very light (δ13C = –21 to –34‰), the presence of biologically derived organic material in the target rocks is assumed. The impact model is unlikely because the C substrate, necessarily of cyanobacteria at ~3.8 Ga, would have been inordinately large (estimated at several cubic km and uncontaminated by crustal material), to account for the estimated two metric tons of carbonado recovered to date (Haggerty, 2014). In addition, the known occurrences of meteorite-impact diamonds (Arizona, United States; Ries, Germany; and Popigai, Russia) are discrete microdiamonds rather than carbonado (Frondel and Marvin, 1967; Hough et al., 1995; Shelkov et al., 1997).
Growth and Sintering in the Crust or Mantle. Some models propose catalytically assisted C-saturated “fluids” in the crust or the mantle. Such fluids provide a source of carbon and a medium capable of drastically decreasing the P-T stability limits of diamond from the traditional 5–6 GPa and 1200–1300°C, at a depth of 200 km or more (Shirey and Shigley, 2013). These “fluids” are hydrous, supercritical (i.e., beyond the point of coexisting fluid + vapor), and intensely oxidized so that diamond crystallization is unlikely, and diamond survival even less so. An analogy with loosely aggregated framesite, found in mantle-derived kimberlites, has also been suggested, but is unsatisfactory because the diamonds are semiprecious, free of pores and patina, and lack the highly reduced mineral suite of metals, carbides, and nitrides.
Subduction. Although carbonado is present in meta-conglomerates (again, see figure 2), these robustly cemented diamonds are very different from the ultra-high-pressure, subducted, metamorphic diamonds found in continental collision zones in Norway, China, Kazakhstan, Greece, and Germany (Ogasawara, 2005). The diamonds at these localities are single crystals and are armored by zircon, garnet, pyroxene, and amphibole that acted as insulating capsules. Sintering would be necessary to form carbonado. This is possible at high pressures and temperatures in the mantle, but the process would have incorporated one or more mantle minerals such as olivine, garnet, pyroxene, and spinel, none of which are observed. Moreover, the inferred subducted plates are oceanic and basaltic in composition and on transformation at high P-T would produce large concentrations of garnet + pyroxene (namely eclogite), which again is not encountered. Transport to Earth’s surface is either not considered or is tentatively ascribed to deep mantle volcanic plumes in both the subduction and radiation models (below).
Radioactive Ion Implantation of Carbon Substrates. Radiation-induced diamond is on the scale of nanometers and cannot account for larger diamonds in the micrometer to millimeter size range found in carbonado. Once diamond is formed, low-energy implantation alters the atomic structure and turns the diamond green; high-energy ion doses produce graphite rather than additional diamond (Kalish and Prawer, 1995). There were no coal deposits at 2–3 Ga, and the radiation-induced diamonds recovered from very rare carburanium (U-rich hydrocarbon) are low in abundance and nanometer in size. Proposals of radiation sintering, and even pore formation, are equally untenable.
Extraterrestrial Origin. The extraterrestrial (ET) model was initially proposed because traditional earthbound scenarios failed to account for major characteristics of carbonado, namely diamond porosity, patina, polycrystallinity, rarity, and location (Haggerty, 2014). Pores are incompatible with high-pressure environments; therefore, carbonado cannot have formed under the same conditions in which conventional diamonds form in the mantle at depths of approximately 200 km. The pores in carbonado (again, see figures 3 and 4) are similar to vesicles in basalts that degassed at low pressures under near-surface conditions from a molten or semi-molten magma. This rules out an origin for carbonado in the crust or the mantle, because liquefaction of carbon is not readily accomplished. In fact, diamond is solid in Earth’s core (6,380 km and approximately 350 GPa and 7000 K; Bundy et al., 1996; Oganov et al., 2013). Consequently, none of the interpreted melt-like features in carbonado (figures 5–7) can possibly be of terrestrial origin. Furthermore, not a single carbonado has been reported from kimberlite-lamproite suites in the nearly 700 metric tons of diamond mined since about 1900 (Levinson et al., 1992). As noted above, carbonado differs from conventional diamond in several respects:
- Hydrogen is prominent and N is dispersed, which is the case for <1% of conventional diamonds (i.e., type Ib).
- Combined N and C isotopes are distinctly not terrestrial (figure 10).
- There are remarkable similarities to diamonds formed by carbon vapor deposition (figure 11), a process that requires vacuum conditions and plasma temperatures that cannot possibly be accomplished in any natural environment on Earth.
- Carbonado lacks the characteristic suite of diamond inclusion minerals such as Cr-garnet, Na-Al-pyroxene, Mg-olivine, Mg-chromite, and Fe-Ni-sulfides, and is instead replaced by exotic, reduced metal alloys and minerals.
The ET scenario posits that carbonado originated from carbon-rich, diamond-bearing stellar bodies and/or disrupted C-bearing planets (Haggerty, 2014). All of the characteristic features of carbonado are satisfied: CVD diamond is the sintering glue to microdiamonds in carbonado; the loss of interstellar H produced the pores, and the patina and flow textures are stellar or interstellar high-vacuum melt products. The model further proposes that carbonado was transported to Earth as a large diamond meteorite or as smaller diamond “plums” in a carbonaceous meteoritic matrix, possibly during the Late Heavy Bombardment (3.8–4.2 Ga), in which the inner solar system was pummeled by meteorites (Fassett and Minton, 2013; Abramov et al., 2013). The numerous craters on the moon are considered evidence of the bombardment (Marchi et al., 2013). The theoretical age of this event corresponds to the oldest age determined for carbonado (3.8 ± 1.8 Ga). This would account for its rarity as a single known occurrence on the São Francisco and Congo cratons, which were once joined geologically as the supercontinents of Nuna and Rodinia. Carbonado was undoubtedly widespread during the bombardment, but the carbonado falls were largely into the expansive oceans that existed at that time. Supercontinent disruption and subduction followed, leaving only the preserved remnants of carbonado on an island that is today split between Brazil and the Central African Republic.
The recent discovery of patches of sub-micron diamonds in Libyan desert glass, a high-silica natural glass that is thought to be of cometary origin (Kramers et al., 2013), lends credence to the ET model for carbonado. This view is supported by the growing lines of evidence for (1) synthetically produced diamond-like glass (Shumilova et al., 2016a, b); (2) nanodiamond encased in glassy carbon shells in the interstellar media (Yastrebov and Smith, 2009); and (3) glassy carbon and nanodiamond produced experimentally (Shiell et al., 2016) and in supernova shock waves (Stroud et al., 2011). Another supporting fact is the discovery of asteroid 2008 TC3, which was tracked upon entering Earth’s atmosphere and landed in North Sudan as a fragmented, diamond-bearing ureilite (Miyahara et al., 2015). Unusual in several respects, the meteorite contains diamonds measuring approximately 100 µm. These are exceptionally large for ureilites, whose diamonds typically measure 1–5 µm, and substantially larger than nanodiamonds of pre-solar origin in carbonaceous chondrites. These reports are complemented by the unexpected discovery that Mercury has a crust of graphite, now covered by volcanic rocks but exposed in meteorite craters (Peplowski et al., 2016), that may prove to be diamond bearing.
Carbonado (figure 12) is the most unusual form of diamond on Earth. Despite many mineralogical clues not observed in conventional diamonds, its mode of origin remains largely unexplained. Discovering the origin of carbonado would herald a whole new mode of diamond formation and could represent a remarkable form of extraterrestrial carbon delivery to Earth. The extraterrestrial model, although conceptual and supported by astrophysical data, will only be vindicated by the discovery of carbonado in the asteroid belt by remote sensing, or by an observed diamond meteorite fall that is dark in color, porous, and patinaed.