Experimental Petrology Applied to Natural Diamond Growth

Perhaps the first point to address in this chapter is what role, or roles, experimental petrology can play in understanding the formation of natural diamond. As can be seen from the other chapters in this volume, there is a rich diversity of diamonds—even confining ourselves to terrestrial examples, there are diamonds that form in impacts, those that are found in ophiolites and in UHP metamorphic rocks, as well as those that are found in mantle-derived volcanics such as kimberlites and lamproites. Some of these—such as those found in ophiolites (e.g., Farré-de-Pablo et al. 2018 and references therein; Litasov et al. 2019a,b)—may form under P, T conditions at which diamond is metastable rather than stable, although such occurrences are controversial (Farré-de-Pablo et al. 2019; Massonne 2019; Yang et al. 2019). Others, such as mantle-derived monocrystalline, fibrous, and coated diamonds, presumably form at P, T conditions at which diamond is the stable polymorph of carbon. This diversity presents a rich P, T composition space for experimentalists to explore to provide some insights into diamond formation.

Experiments allow researchers to explore first-order questions such as

• Can diamond form from this particular composition of fluid or melt?

• Can diamond form from a specific redox reaction, such as carbonate reacting with a reduced fluid?

Experiments can also address questions such as how the diamond-forming melts or fluids originally form, and how they evolve as they interact with different lithologies in the Earth.

The vast majority of natural diamonds form in the Earth’s lithospheric mantle (See Nimis 2022, this volume; Stachel et al. 2022a, this volume), and so unsurprisingly most of the experimental work to date has focused on this pressure range. There has been some work at much higher pressure and temperature conditions extending to those of the deep Earth, however, and we shall touch on those as well in this review.

In the one-component carbon system, diamond is the high-pressure polymorph stable at pressures above the graphite—diamond transition (Kennedy and Kennedy 1976; Day 2012). Adding more components makes the situation more complex: in C–O, diamond would be stable in a region of P, T, fO₂ space bounded by the reactions

and

If the fluid phase contains H as well, then the relevant system is C–O–H (Fig. 1), in which the diamond saturation curve separates a two-phase field of diamond + fluid from a field of C-undersaturated fluid. It should be noted that the fO₂ of the fluid increases from left to right along the diamond saturation curve, and for context the value of fO₂ for the iron-wüstite buffer (IW), and the locations of fluids with values of log(fO₂) three and four units below the calculated (metastable) location of the fayalite-magnetite-quartz buffer (FMQ) are shown.

The situation in the Earth’s mantle is more complex, given that CO₂ becomes unstable in the presence of olivine (Newton and Sharp 1975) because of the carbonation reaction

A consequence of the stability of this reaction is the well-known EMOD reaction

(Eggler and Baker 1982), which further constrains the stability of diamond. Schematically, Figure 2 shows the fields of stability for carbonate, diamond, and fluid in T–log(fO₂) space at constant pressure within the diamond P, T stability field. Under conditions where a carbonate melt rather than crystalline magnesite is stable, an analogous reaction to (3) where MgCO₃ is a component in the melt would apply, as discussed by Stagno and Frost (2010).

Under more reducing conditions such as those proposed to occur with increasing depth in the Earth’s mantle (see overview by Stagno 2019), carbon can dissolve in the alloy or form carbides such as cohenite, (Fe, Ni)₃C, or Fe₇C₃. Another complication is the presence of sulfur—with a bulk silicate Earth estimated concentration on the same order as that for carbon (~250 ppmw, McDonough and Sun 1995). Although there is ongoing debate about the relative importance of alloys, carbides, and sulfide melts as hosts of carbon in the Earth’s mantle, the key point for the experimentalist is that all three of these are at least potential sources of carbon that could precipitate diamond upon oxidation—a process tractable to experimental study.

Considerations of the large-scale carbon cycle (e.g., Lee et al. 2019) reveal two basic categories of carbon in the Earth’s mantle: primordial carbon surviving from early Earth accretion and differentiation, and carbon re-introduced into the Earth’s interior via subduction. The present carbon content for the primitive mantle is ~90–130 ppm, although as outlined in the recent review by Dasgupta and Grewal (2019) estimates range from ~100 to > 500 ppm. Subsequent to accretion and large-scale initial differentiation leading to core formation, it is not clear how carbon is hosted in the mantle, but it seems reasonable that this carbon would be present in its thermodynamically stable form(s) at the P, T and oxidation state of the ambient mantle. This presupposes that ferric–ferrous equilibria will serve as a sink or source for oxygen, which is reasonable at these carbon concentrations.

Subduction transports carbon down into the mantle, but the amount of carbon that survives sub-arc processing to be subducted into the mantle remains an active area of research (e.g., Kelemen and Manning 2015; Galvez and Pubellier 2019; Lee et al. 2019). Carbon would be present in both subducted sediment and altered oceanic crust (AOC) as carbonate, both biogenic and abiogenic, as well as some fraction as organic matter. The relative contribution of sediment and AOC likely depends on the specific thermal regime of a subducting slab, but a recent study by Li et al. (2019) implicates the latter as the dominant carrier of carbon as both abiogenic and biogenic carbonate into the mantle. Whether all this carbonate is reduced to graphite during subduction (Galvez et al. 2013) or survives as carbonate until partial melting ensues is unclear, so the experimentalist is faced with addressing both possibilities as carbon sources for diamond formation.

Diamond can be formed directly from graphite in the absence of a solvent or catalyst, but this transformation requires pressures above ~10 GPa and temperatures in excess of 1800 °C (e.g., Irifune et al. 2004 and references therein). Diamond growth at lower pressures and temperatures requires some sort of solvent or growth media, and thus the first question is what that medium might be. From the examination of natural diamonds, it is well accepted that diamonds grow through metasomatic processes in C–O–H–N–S–Cl bearing mobile fluids, melts or supercritical fluids, and occurs over times and depths.

As outlined in other chapters in this volume, study of the inclusions in natural diamonds led to the identification of syngenetic inclusions, those that form as the diamond was growing (Harris 1968). The mineral inclusions enabled researchers to distinguish between diamonds that were growing in peridotitic and eclogitic lithologies, for example. Other inclusions were thought to sample the growth medium from which the diamond is precipitating. Fibrous diamonds in particular were fruitful to study in this context because they contain the so-called “HDF” (high-density fluid) inclusions. Among fibrous diamonds, the coated ones, with a monocrystalline pure core surrounded by a fibrous rim, inspired the idea that both kind of diamonds may share the same parent fluids and that a single diamond may grow in a few or several events. Chrenko et al. (1967) reported the first indication for the presence of carbonates and water in inclusions in diamond, but Navon et al. (1988) were the first to propose that coated diamonds sample mantle melts or volatile-rich fluids (enriched in H₂O, CO₃²⁻, K₂O, Cl, depleted in MgO and including several other elements) from which they grow. Numerous publications have followed (see Weiss et al. 2022, this volume), including some significant studies such as: the finding of solid carbon dioxide inclusions in fibrous diamonds (Schrauder and Navon 1993), the discovery of brine inclusions in cloudy diamonds (Izraeli et al. 2001), the chemical and isotopic connection of fluid compositions trapped in coated diamond’s inclusions from various cratons showing that fluids derived from mantle source would not be affected by local heterogeneities (Klein-BenDavid et al. 2004), the idea of miscibility/immiscibility processes linking mantle fluids of different compositions or “endmembers” during diamond growth (Klein-BenDavid et al. 2007), and many others.

In brief, the bulk compositions of these HDFs can be described in terms of four endmembers: (i) silicic HDFs that are rich in Si, Al, K, and water with some carbonate; (ii) saline HDFs that are rich in Cl, K, Na, water, and carbonate; (iii) and (iv) hydrous carbonatitic HDFs that are Mg-rich and Mg-poor, respectively.

Along with the relationship between these endmembers and their origin, a long-standing question has been whether the more common monocrystalline diamonds form from the same type of fluids. Strong evidence for this idea has been established from a variety of studies. For example: the common sinusoidal REE patterns between garnet inclusions trapped in the monocrystalline area of coated diamonds and HDFs trapped in the fibrous rims (Weiss et al. 2009); the similarity of hydrosilicic HDFs that precipitated the rims of coated diamonds with the fluids that precipitated most monocrystalline diamonds (Rege et al. 2010); the first finding of HDF micro-inclusions in a monocrystalline diamond, similar in major and trace elements compositions to those of fibrous diamonds (Weiss et al. 2014), suggesting that these HDF are involved during the growth of many monocrystalline diamonds. As a last example, Nimis et al. (2016) discovered thin “fluid” hydrous silicic films around solid mineral inclusions in gem-quality monocrystalline diamonds from peridotitic and eclogitic suites that were interpreted as potential vestiges of the original fluid present during diamond growth.

Are fibrous, coated, and monocrystalline diamonds growing by the same process? In our view, the debate was definitively closed in the affirmative—at least for lithospheric diamonds, by the study of Jablon and Navon (2016), who found micro-inclusions of HDFs trapped in the twinning planes of twin gem diamonds (macles) together with silicate minerals. They concluded that the mechanism of diamond formation is similar for most diamonds. It is worth noting that this process also includes the growth of micro diamonds in ultra-high-pressure metamorphic rocks (e.g., Stöckhert et al. 2001; Dobrzhinetskaya et al. 2007; Frezzotti et al. 2014).

Most of the studies based on fluid inclusions document the important role played by water during diamond growth, a role possibly not only limited to subduction zones where water is likely recycled from slabs, but also in the cratonic mantle. Although most of the HDFs observed in lithospheric diamonds are oxidized, some diamonds such as mixed-habit ones containing octahedral and cuboid sectors also host inclusions with reduced fluids such as CH₄ (Smit et al. 2016). Diamond growth in fluids evolving in redox state is also able to reconcile the apparent contradictory messages delivered from diamond stable isotope studies, because depending on the carbon source, CO₂ (Boyd et al. 1994) or CH₄ (Thomassot et al. 2007), the core to rim evolution of the diamond’s δ¹³C signature will be reversed. Furthermore, both eclogitic and peridotitic diamonds can be derived from the same isotopically homogeneous carbon source with metasomatic growth (Cartigny 2005).

These HDFs may not be limited to the lithosphere; some evidences of volatile-bearing inclusions are found in sublithospheric diamonds, such as ice VII inclusions (Tschauner et al. 2018); CH₄ and H₂ associated with inclusions of solidified iron–nickel–carbon–sulfur melts in large gem diamonds (Smith et al. 2016); or unknown fluids in association with inclusions of iron carbides (Kaminsky and Wirth 2011). These metallic melts may take part in diamond growth in the lower mantle and transition zone, but possibly also in the lithosphere, in both cases in very reduced environments.

In this review, we will try to guide the reader into the voluminous literature on experimental studies of diamond formation. To provide some structure to this, we will look at different growth media in turn, moving from “simple” to more complex systems. In each section, we will tabulate studies chronologically, and the reader will recognize how the experimental studies have evolved as studies of natural diamonds have continued to provide insights and constraints to help ground the experiments.

There have been extensive studies over the last 30 years of diamond nucleation and growth from initially graphite-saturated C–O–H fluids. Diamond grows from a variety of these fluids, from CH₄- to CO₂-rich (Table 1). The studies tabulated here are those in which graphite was present in the starting material (or generated upon breakdown of the organic fluid source); studies examining the ability of fluids to participate in oxidation-reduction reactions, such as those in which carbonate is reduced to diamond, are discussed in a subsequent section.

In general, these studies demonstrate that diamond growth is higher in H₂O-rich systems, lowest in CH₄-rich systems, reflecting a dependence on the activity of H₂O in the fluid (which is clearly fO₂ dependent). Diamond growth is characterised by temperature-dependent induction times, certainly for nucleation of new diamond crystals, and in some cases even for growth on pre-existing seed diamonds. More recent studies such as Matjuschkin et al. (2020) have demonstrated the ability of these fluids to grow diamonds at lower temperatures (e.g., that are more realistic for lithospheric mantle geotherms). In this study, extensive efforts were devoted to improving experimental design to minimize changes in fluid composition over the course of the experiments.

In experiments at diamond-stable conditions, C–O–H volatiles are added to the capsule as either a fluid by micro-syringe, or as solid materials that break down to fluids at the conditions of the experiment. C–O bearing materials such as oxalic acid dihydrate (H₂C₂O₄·2H₂O) and silver oxalate (Ag₂C₂O₄) have a long history of use in experimental petrology (e.g., Holloway et al. 1968; Boettcher et al. 1973). These materials, as well as others such as anthracene (C₁₄H₁₀) and glucose (C₆H₁₂O₆), have been used to generate C–O–H fluids for diamond-synthesis experiments. A key assumption is that the solid starting material break downs to the expected fluid composition at experimental conditions; progress is being made on both in situ studies of these fluids (McCubbin et al. 2014) and careful characterisation of fluids following quenching of the experiment (e.g., Tiraboschi et al. 2016; Sokol et al. 2017). Other issues arise for specific compounds; for example, the elemental silver liberated by decomposition of silver oxalate can alloy with the sample capsule, which can lead to melting, capsule rupture, and loss of fluid (e.g., Brey et al. 1991).

Because of the potential for hydrogen diffusion across the capsule wall during the experiment, evolution of the composition of the fluid phase during the experiment is an issue relevant to these experiments (see discussions in Sokol et al. 2004, for example; Palyanov et al. 2010a). This long-standing experimental issue has led to various approaches to buffer the fluid composition (see overviews by Rubie 1999; Stagno 2019), and Table 1 notes studies in which buffering of the fluid phase by an external hydrogen buffer was employed.

Continued focus on experiments with well-constrained redox conditions to maximize experimental run durations and to maintain constant fluid compositions will undoubtably improve our understanding of the dependence of diamond growth rate on fluid composition. Given the solubility of silicates in C–O–H fluids at high P and T, continued study of diamond growth in fluids saturated with peridotitic or eclogitic mineral assemblages would seem to be the most fruitful in terms of direct relevance to diamond growth in the Earth’s mantle. Finally, studies of diamond growth in C–O–H fluids will benefit from continued efforts to improve thermodynamic models of fluids at high P, T. As an example, the solubility of carbon in CO₂ fluid at 6.3 GPa and 1250–1400 °C measured by Palyanov et al. (2010a) is ~1 order of magnitude larger than that predicted by the GFluid thermodynamic model (Zhang and Duan 2010). Furthermore, fluid models such as GFluid are restricted to fluid constituents in the C–O–H system. Extending these models to quantitatively model the effects of the solubility of silicates and oxides is sorely needed—and will require an enormous amount of experimental data to calibrate.

Like C–O–H fluids, carbonate melts have attracted significant attention of experimentalists as potential growth media for diamond over the past three decades. A rich variety of alkali- and alkaline earth carbonates have been explored in these studies (Table 2), including those most likely to be relevant to diamond growth in the Earth’s mantle, such as those in the Na₂CO₃–K₂CO₃–CaCO₃–MgCO₃–FeCO₃ system. All studies in Table 2 have graphite present in the starting material (or likely formed by breakdown of oxalate or other solid sources in the initial stages of the experiment) and use carbonate melts as media for graphite transformation into diamond. Experiments that explore diamond nucleation and growth as a result of redox reactions involving carbonates are described in another section.

Review of Table 2 reveals that the early experiments were conducted at temperatures well above those reasonable for diamond growth in the mantle for two reasons: (1) diamond nucleation and growth require temperatures significantly overstepping of the graphite–diamond transition and (2) diamond growth requires the carbonate to be partially molten in order to serve as a viable growth medium. The solidus temperatures of alkaline-earth carbonates are in excess of ~1400–1600 °C at ≥ 6 GPa for successful growth of diamond. Alkali carbonates, with their lower melting temperatures, allow melt-present experiments at lower temperatures, which facilitates crystallization of diamond at lower temperatures. The addition of volatiles such as water and CO₂ to the systems further encourages diamond crystallization to proceed at lower temperatures.

Carbonate melts are characterised by low viscosity at least to pressures of ~6 GPa (c.f. Jones et al. 2013 and references therein) although their viscosity may increase at higher pressures (Wilding et al. 2019). In the experiments, possible escape of the liquid through a graphite or MgO capsules must be considered—most of the experiments use Pt or Au in part for this reason.

Oxygen (hydrogen) fugacity control is also a consideration; hydrogen diffusion through the capsule wall can drive carbon precipitation via 2H₂ + CO₂ (melt) = C + 2H₂O, which also introduces H₂O in the system (e.g., see discussion in Palyanov et al. 2016). However, it is worth emphasizing that use of an external oxygen buffer in a double-capsule configuration with hydrogen diffusion across the capsule wall equalizes the chemical potential of hydrogen, not the fO₂ (e.g., Whitney 1972), and these buffers are best used as hydrogen sinks unless there is a fluid coexisting with the carbonate melt in the experimental charge.

Alkali carbonates, particularly K₂CO₃, are notoriously hygroscopic and special care must be taken in experiments with these carbonates. Figure 7 of Shatskiy et al. (2015a) illustrates the effect of adsorbed H₂O on phase relationships: the lowest-T eutectic in the K₂CO₃–CaCO₃(–H₂O) drops from ~1200 °C to ~1000 °C at 6 GPa when the samples were dried at 100 °C rather than 300 °C. Some of the differences in the results of the studies involving alkali carbonates may result from this issue. Parenthetically, the temperature effect Shatskiy et al. (2015a) observed also implies high solubility of H₂O in carbonate melts at high pressures—a field relatively unexplored except at much lower pressures (0.025–0.225 GPa; Keppler 2003).

Further studies of diamond growth in carbonate melts could focus on exploring more of the composition space relevant to natural melts. These studies will benefit from insights provided by recent experimental studies of phase relationships in carbonate systems at mantle pressures (see review by Shatskiy et al. 2015b, for example). The study of systems such as K₂CO₃-CaCO₃-MgCO₃ (Arefiev et al. 2019) and Na₂CO₃–CaCO₃–MgCO₃ (Podborodnikov et al. 2019) hold particular promise in this regard.

Given that it has been firmly established that diamond can indeed grow in carbonate melts, another fruitful line of research is to take advantage of these melts to grow diamond to continue to address issues such as nitrogen incorporation into diamond (e.g., Khokhryakov et al. 2016) and isotopic fractionation between diamond and coexisting melts or fluids (Reutsky et al. 2015a,c, 2018; Bureau et al. 2018) (see later sections).

On another front, there remains another unresolved question: How does carbon dissolve in carbonate melts? Can it dissolve as a neutral carbon species, or does it dissolve as carbonate? A key challenge in resolving this question is the fact that almost all carbonate liquids do not quench to glass, with the intriguing exceptions of some K–Mg carbonates and some more complex melts in the BaSO₄–La(OH)₃–Ca(OH)₂–CaF₂–CaCO₃ system (see Carbonate glasses in Jones et al. 2013 and references therein). This issue naturally complicates both measurements of solubility and solution mechanisms. The available data on C solubility in carbonate melts show low solubility: <0.3 wt.% at 6.8 GPa and 1700 °C in Na₂CO₃ melt (Sokol et al. 1998). On the other hand, CO₂ appears to dissolve readily in carbonate melts: Palyanov et al. (2016) observed a solidus depression of ~100 °C at 6.3 and 7.5 GPa in the system Na₂CO₃-CO₂–C compared to Na₂CO₃, and argued that the 16.4 wt.% of CO₂ produced upon decomposition of sodium oxalate (Na₂C₂O₄) dissolved completely in the melt at these conditions. This solubility is higher than the 11.5 wt.% and 6.5 wt.% CO₂ dissolved in CaCO₃ and MgCO₃ melts, respectively, at liquidus conditions at 2.7 GPa (Huang and Wyllie 1976). To our knowledge, there are no spectroscopic data on solution mechanisms of either C or CO₂ in carbonate melts at diamond-stable P, T. This topic would of course be an ideal in situ study, such as those done at lower pressures (~700 MPa e.g., Mysen 2018), when technology has evolved sufficiently to make measurements at these conditions possible.

The existing experimental data indicate that volatile-free silicate melts cannot provide either diamond nucleation or growth on seeds in the range of P, T-parameters that are of interest for natural diamond formation. Silicate and oxide systems become diamond-forming only when H₂O is added (Table 3). Clearly, part of this effect is a result of water lowering the solidus and liquidus temperatures to stabilize the melt to lower temperatures more representative of conditions of natural diamond formation. In most systems, however, we lack basic information about the solubility of H₂O in the melt as a function of P and T—or indeed the basic phase equilibria—needed to differentiate the effects of melt fraction and melt composition on diamond growth. Nevertheless, the studies in Table 3 show that the nucleation and growth of diamond on seeds was established at temperatures of 1500–1600 °C in water-bearing silicate melts and silicate-bearing aqueous fluids. An increase in the H₂O content in such systems results in a significant increase in the degree of graphite into diamond transformation and is accompanied by a reduction in the induction period preceding diamond nucleation and an increase in carbon mass transfer. As a result, the most favorable conditions for nucleation and growth of diamond occur in the water-rich fluid phase, containing small amounts of silicate or oxide solute. A decrease in temperature leads to a decrease in the diamond-forming ability of silicate–aqueous media, which is usually attributed to more sluggish kinetics at lower temperature. The minimum nucleation temperature for diamond was 1500 °C at 6.3 and 7.5 GPa. The minimum temperature for diamond growth on seeds is 1400 °C at 6.3 GPa (Table 3).

The overwhelming majority of experiments in the silicate-water systems were carried out in Pt ampoules without special buffering. The presence of graphite and water among the initial reagents due to H₂ diffusion through the walls of Pt ampoules, in the absence of buffering, can lead to the formation of an uncontrolled amount of CO₂ in a predominantly aqueous fluid.

One of the unsolved problems is the determination of the solubility of carbon depending on temperature and on the content of H₂O. Experimental studies in this direction will make it possible to evaluate the real scale of diamond formation in the processes of evolution of the system composition and with a decrease in the P, T-parameters.

As seen in Table 3, most experiments in the silicate–water systems were carried out without the participation of alkalis. The addition of alkaline components can change the pH of the medium and bring the model system closer to the eclogitic fluid and, possibly, allow to reduce the minimum P, T-parameters.

Early studies showed that diamond can grow in anhydrous alkali halides (Wang and Kanda 1998; Litvin 2003), but such studies are unlikely to be directly relevant to diamond growth in the Earth because of the absence of H₂O. Brines, or hydrous solutions of alkali halides, on the other hand, do form one of the “endmembers” of HDF inclusions in natural diamonds. Remarkably few studies address diamond growth in such systems. The two studies in Table 4 demonstrate that both KCl + H₂O and NaCl + H₂O fluids can nucleate and grow diamonds at 7.5 GPa and high temperatures (1500–1600 °C). The more extensive work in the former system documents a composition-dependence on the degree of transformation of graphite to diamond (see Table 4). Subsequent work in brine-containing systems has shifted to systems coexisting with silicates or carbonates (see subsequent sections).

There has been limited study of chloride-carbonate systems (Table 5). From these results, it is clear that diamond nucleation and growth can proceed in either anhydrous or hydrous carbonate-chloride melts. An unresolved issue is the minimum temperature at which this occurs: Tomlinson et al. (2004) observed growth as low as 1050 °C at 7 GPa, whereas Palyanov et al. (2007a) did not see diamond growth at 1400 °C at 7.5 GPa. Given the limited data, it is impossible to say whether the presence of H₂O in the system allows diamond nucleation and/or growth at lower temperatures.

These studies explore the efficacy of carbonate–silicate melts as a medium for diamond nucleation and growth. In effect, these studies can be considered extensions of carbonate-melt studies that explore the effect of dissolution of silicate constituents into the melt. At the same time, these studies mark a significant step towards simulating systems representative of those potentially present in the Earth’s mantle. Examination of Table 6 shows that these melts are effective growth media for diamond, particularly at higher temperatures, which in part reflects the temperatures required for melt to be present, but also the temperature-dependence of diamond nucleation and growth as seen in the previous systems.

Unresolved questions in these systems include whether the effect of adding carbonate is strictly a result of stabilising a melt to lower temperatures, a compositional effect of the presence of carbonate, or a result of decreasing viscosity of the melt with increasing carbonate content. Systematic studies to tease out these possibilities may be warranted if such systems are viewed to be useful models for natural diamond growth.

The main regularities in diamond crystallization, as revealed in simple model systems (above), are also valid for multicomponent media and are as follows:

• Silicate melts do not facilitate nucleation and diamond growth; it is the addition of water makes the silicate + H₂O systems diamond-forming. The silicate component acts as a diamond formation inhibitor (e.g., Sokol and Pal’yanov 2008), with all other parameters being constant.

• In carbonate–silicate environments, the silicate component also inhibits diamond formation.

• The most effective diamond-forming media are alkaline carbonate melts as well as H₂O- and CO₂-bearing fluids; nucleation and growth of diamond in these media are realized at the lowest P, T, probably reflecting their stability as melts or fluids at these conditions.

• The induction period preceding the nucleation and growth of diamond depends on P, T and the composition of the crystallization medium.

• With an increase in P and T, the diamond-forming ability of all media increases. A decrease in P and T in all media leads to crystallization of metastable graphite.

• An increase in pressure from 5.7 to 12 GPa and above allows the reduction of the minimum diamond nucleation temperature in all systems, including multicomponent carbonate–silicate media.

• All other parameters being equal, the nucleation of a diamond requires higher temperatures than the growth of a diamond on seeds.

• Crystallization of diamond or graphite in a carbonated eclogite at 9–23 GPa was suggested to result from oxidation of Fe²⁺ to Fe³⁺ in coexisting garnet (Kiseeva et al. 2013).

Experiments in multicomponent carbonate–silicate systems and using the “diamond-bearing” rocks were carried out mainly in Pt ampoules (Table 7). The effect of the ampoule material (for example, Pt and Au) on the diamond nucleation process in non-metallic solvents has been studied insufficiently. For example, in long-term experiments (tens of hours), the dissolution of Au and its crystallization on diamond in the C–O–H fluid were shown (Sokol et al. 2001a). The effect of Pt on diamond nucleation was established in long-term experiments in the kimberlite–graphite system (Palyanov et al. 2015); however, the real scale of the effect of Pt is not clear. It can be most significant at high P and T, especially when using small diameter Pt ampoules.

New experimental studies need to be performed in these fields:

• Despite a large number of experiments carried out in a wide range of P, T-parameters and in various media, there is still no clarity and experimental confirmation of how diamond is formed in eclogites and peridotites in the pressure range of 4–7 GPa and 900–1400 °C, characteristic of formation of most natural diamonds (Stachel and Harris 2008).

• There is still insufficient experimental data to assess the effect of various components of the C–O–H–N–S system fluid on the crystallization of diamond in multicomponent media. The data on the defect-impurity composition of diamonds synthesized in various model media are still scarce, which does not allow us to fully assess the overall picture of the effect of the composition of diamond-forming media on the real structure of diamond.

The studies in this section (Table 8) document that diamond can nucleate and grow in iron-sulfide melts, but typically at temperatures higher than those appropriate for corresponding pressures in the mantle. From these studies, sulfide melts appear to be less effective at mediating diamond growth than either H₂O-rich fluids or carbonate melts.

To this point, all of the studies had graphite in the starting materials, and the primary focus was on whether a particular melt or fluid would mediate the transformation of graphite to diamond at conditions approaching those appropriate for the Earth. In contrast, the studies in this section all contain carbonate as a carbon source and use different reducing agents to produce diamond by reduction (Table 9). The early studies demonstrated the viability of this mechanism, using metal, metal carbides, or metal hydrides as reducing agents. Subsequent studies looked at the possible role of C–H fluids, sulfides and oxides. Overall, reduction of carbonate has been shown to be a viable mechanism for formation of diamond; the key question in terms of applicability to the Earth is what the appropriate reducing agent might be in different environments, particularly in the lithospheric mantle at depths shallower than those at which iron metal or carbide are stabilized.

Whereas they are not designed for that purpose for most of the cases, laser heated diamond anvil cells (LHDAC) experiments may give significant information about the formation of superdeep diamonds, and demonstrate progress towards multiphase mantle petrology at conditions of the lowermost mantle.

Most of these studies (Table 10) are looking at high T- and P-dependent breakdown of carbonates like CaCO₃ = CaO + C + O₂ in the context of subduction recycling, or to characterize the formation of new mineralogical species of carbonates in the deep Earth. For example, Dorfman et al. (2018) show that CaCO₃ is preserved at pressure and temperature conditions reaching those of the deep lower mantle, supporting its relative stability in carbonate-rich lithologies under reducing lower mantle conditions. Boulard et al. (2011) provide evidence for the presence of both diamonds and an oxidized C-bearing phase, suggesting that oxidized and reduced forms of carbon might coexist in the deep mantle.

Litvin et al. (2014) took the benefit of these HPHT tools to study the diamond-forming lower mantle systems, by investigating melting phase relations of simple carbonates of Ca, Mg, Na and multicomponent Mg–Fe–Na–carbonate up to 60 GPa and 3500–4000 K (LHDAC and multi-anvil presses). They grew diamonds that showed that ‘Super-deep’ diamonds can crystallize in the system carbonate-magnesiowüstite-Mg-perovskite-carbon.

Recently, Martirosyan et al. (2019) performed LHDAC experiments in the MgCO₃–Fe system combined with in situ synchrotron X-ray diffraction and ex situ transmission electron microscopy. Based on the results they suggest that the interaction of carbonates with Fe⁰ or Fe^0-bearing rocks can produce Fe-carbide and diamond, which can accumulate in the D’’ region (the lowermost portion of the mantle that sits just above the molten iron-rich outer core).

Because of the pressure limitation of large volume presses, DAC experiments may be precious tools to understand the growth of superdeep diamonds in the lower mantle. Another potentially interesting approach is the use of dynamic compression experiments (i.e., laser-compressed materials), to generate diamonds at ultrahigh pressures and temperatures in order to study impact diamonds or the formation of diamonds in planetary interiors conditions (Kraus et al. 2017).

There has been significant interest in exploring the possibility for iron carbides to act as both carbon sources and reducing agents since the general acceptance of the idea that depth-driven reduction in the Earth’s mantle stabilizes Fe–Ni metal and carbide in C-bearing mantle regimes. Studies have examined interaction of carbides, particularly cohenite (Fe₃C), with oxides, silicates, carbonates, and sulfides, as can be seen in Table 11 below. Cohenite has proved to be an effective source of carbon upon reaction, usually forming metastable graphite. Growth of diamond is not typically observed in the absence of a melt. In many of the studies in Table 11, it is notable that the presence of melt in itself is insufficient, and diamond growth is inhibited until a sufficiently high temperature is achieved. In carbonate-bearing experiments, elemental carbon is produced by both breakdown of cohenite as well as reduction of CO₂ produced by decarbonation reactions involving the carbonate.

Inclusions trapped during growth of synthetic diamonds at high pressure and temperature have been observed since the first syntheses in metallic melt solvents. In one of the early reports from the G.E. group, Bovenkerk (1961) stated “It has not been possible so far to grow diamond without trace inclusions.” For industrial purposes, of course, inclusions are something to avoid or minimize (e.g., discussion in Sumiya et al. 2002). On the other hand, given that inclusions in natural diamonds provide valuable clues to the nature of the growth medium, it is of interest to deliberately try to grow diamonds with inclusions to see, for example, if the inclusions provide an accurate picture of the fluid or melt from which the diamond is growing. Early studies of synthetic diamond documented metal inclusions inferred to be from the growth medium (Lonsdale et al. 1959), and, intriguingly, coesite (Milledge 1961). The possibility of formation of single fluid inclusions during diamond crystallization in metal-carbon system was shown by Osorgin et al. (1987). Later work documented the presence of a rich diversity of inclusions (taenite, wüstite, spinel, silicate, diamond, and fluid) in synthetic diamond grown in the Fe–Ni–C system (e.g., Pal’yanov et al. 1997). In this study, the oxide and silicate inclusions resulted from diffusion of elements from the container materials and from impurities in the initial reagents. These examples suggested that detailed study of inclusions in diamond growing in other systems would be fruitful in exploring how representative these inclusions are of the growth medium.

As shown in Table 12, many studies report the presence of inclusions in passing, such as graphite (e.g., Pal’yanov et al. 1999b) and quenched melt, either reduced or oxidized—sometimes in the same experiment (Palyanov et al. 2013b), but there have been also some studies designed specifically to synthesize inclusion-bearing diamonds (e.g., Khokhryakov et al. 2009; Bureau et al. 2012, 2016; Bataleva et al. 2016c). The first of these documented the rich variety of ways in which graphite inclusions in diamond can be formed; importantly it showed that graphite inclusions could form during growth of the diamond crystal from a carbonate melt—in the stability field of diamond. The others showed a good fidelity between the materials trapped as inclusions and those present in the environment in which the diamonds were growing. These studies provide a foundation for further studies on inclusion formation and trapping in a variety of media. For example, Bataleva et al. (2016c) suggested that slower growth rates would decrease the relative proportion of fluid or melt inclusions in the growing diamond, as had been previously documented for inclusions in magmatic minerals (Roedder 1984). Applying these methodologies to other growth media—such as the saline-rich endmember seen in natural fibrous diamonds—would be of considerable interest, as another example.

Natural diamonds are known to contain impurities, the most common being nitrogen. The concentrations of nitrogen depend on the paragenesis and vary from below 10 ppm to > 3500 ppm (see Stachel et al. 2022b, this volume). Boron and hydrogen are also naturally present in the diamond lattice (Green et al. 2022, this volume).

Synthesis of commercial HPHT diamonds in, for example Fe–Co melts under standard synthesis conditions, contain ~100–300 ppm N in {111} growth sectors, and usually about half that in {100} growth sectors when grown under “standard conditions” (i.e., 1300–1400 °C, 5.5 GPa, Burns et al. 1999); lower growth temperatures can reverse this difference, and increase N contents up ~1000 ppm because of the decreased solubility of N in the melt at lower temperatures. In these syntheses, nitrogen is introduced as a contaminant in the starting materials, and its concentration in the diamond can be reduced by addition of so-called nitrogen “getters” such as Al, Ti, or Zr, which form nitrides at the experimental conditions.

In general, non-metallic solvents—carbonate, carbonate–silicate, sulfide, and sulfur—tend to crystallize diamond with higher (500–1500 ppm) N contents. Use of N-bearing compounds such as BN, NaN₃, Ba(N₃)₂, CaCN₂, Fe₃N, P₃N₅, or C₃H₆N₆ can produce diamond with > 3000 ppm N (Table 13).

With the methodology for growing diamond with specific amounts of N, the next frontier is to use modern ion probe methods to examine N concentrations and isotopic compositions across diamond crystals, as pioneered by Reutsky et al. (2008b, 2017), for example. An attempt to find the interconnection between morphology and defect-impurity composition of diamond with its formation conditions was performed by Palyanov et al. (2021a).

Boron is another significant impurity found in diamonds, which gives them a blue color. Defined as Type IIb diamonds, they are very rare < 0.1% of the worldwide extracted diamonds and naturally contain almost no N and up to 10 ppm B (Gaillou et al. 2012). Blue diamonds are synthesized for industry because B gives diamonds semi-conductive properties, given that significant amounts of boron (> 1000 at.ppm) can be incorporated in synthetic CVD diamonds (Polyakov et al. 2001). Synthetic boron-doped diamonds are also used as heaters either in large volume presses or in diamond anvil cells for HPHT experimentation (Shatskiy et al. 2009).

The growth of diamonds with or without impurities is of great importance for understanding the mechanisms at depth in the Earth. For example, experiments performed by Palyanov et al. (2013b) show that simultaneous diamond growth in metal-rich and in carbonate-rich parts of the same capsule would form diamonds containing only 100–200 ppm N when in contact with metallic Fe, whereas those formed in the carbonate melt zone contain 1000–1500 ppm N. This difference is explained by the siderophilic affinity of N to metallic liquids at high P, T conditions. The occurrence of Type II diamonds may be explained by growth in the presence of metallic melts (Smith and Kopylova 2013). Further experimental studies are necessary to confirm such a hypothesis.

In general, the bulk carbon isotope composition of HPHT diamonds solely depends upon the initial carbon isotope composition of the diamond-forming system. All the diamonds from HPHT experiments have δ¹³C in a range from −20 to −30‰ PDB, which is inherited from the initial graphite sources used for synthesis (Hoering 1961; Laptev et al. 1978; Ivanovskaya et al. 1981; Taniguchi et al. 1996; Reutsky et al. 2008a, 2015a) (Table 14). In the film-growth (FG) experiments, no significant isotope inhomogeneity is reported for produced diamond material (Hoering 1961; Laptev et al. 1978; Ivanovskaya et al. 1981). When the temperature gradient growth (TGG) method is used, carbon isotope fractionation, which accompanied the diamond crystallization from carbon solution in metal melt, produces a certain isotope profile in the direction of crystal growth (Reutsky et al. 2008a,b). In Fe-Co+Gr, Fe-Ni+Gr and Fe+Gr systems, the carbon isotope fractionation coefficients at 5.5 GPa and 1400–1500 °C are the same and vary from about 2.0 to 4.5‰ (Reutsky et al. 2008b). Therefore, within the distance of 1.5 mm from the seed these diamond crystals show gradual decrease of δ¹³C on a scale of 2–4‰, followed by slight fluctuation around carbon isotope composition of initial graphite (Reutsky et al. 2008a). In the metal–carbon systems the δ¹³C of diamond always equal or higher than δ¹³C of the initial graphite. The carbon isotope fractionation coefficients associated with diamond growth are strongly depends on the linear growth rate and may vary from 0 to 4.5‰ (Reutsky et al. 2012).

Carbon isotope fractionation was also studied in the NaCO₃+CO₂+C model system with a single carbon source also studied (Reutsky et al. 2015a). The carbon isotope fractionation in those experiments show opposite direction, compared to metal–carbon systems, and the diamond here is always depleted in heavy (¹³C) isotope in comparison with carbonate–CO₂ fluid, exactly as expected from calculations of isotope equilibrium (e.g., Richet et al. 1977). At 7.5 GPa temperature-dependence of carbon isotope fractionation coefficient was determined as the function: Δ_{Carbonate fluid-Diamond} = 7.38 × 10⁶ × T⁻² and established to be dependent on carbonate:CO₂ ratio in the fluid (Reutsky et al. 2015a). The scale of this depletion is comparable to that documented for metal-carbon systems and reaches 4‰. Later work on diamond experimental crystallization from carbonate–silicate–water fluid reveal a similar estimate of fluid/diamond carbon isotope fractionation at 7 GPa and 1400 °C (Bureau et al. 2018).

Detailed investigation of different crystallographic sectors of HPHT diamonds reveal surface-controlled carbon isotope distribution within diamond crystals showing higher isotope fractionation on faces {100} in comparison with face {111} (Reutsky et al. 2017). This is supported by ab-initio calculations and provide explanation of systematic isotope lightening of growth sectors {100} relative to {111}.

Useful information on diamond growth can be obtained from investigation of paired carbon and nitrogen isotope fractionation during diamond growth in certain crystallographic directions using high spatial resolution technique such as secondary ion mass spectrometry.

Natural diamonds exhibit various surface textures, acquired during their history and corresponding to growth and dissolution events. For a long time, the dissolution textures were attributed to the effect of the kimberlite melts while diamonds were brought up to the surface (Robinson 1979), and eventually to later secondary processes mostly due to transport and erosion.

Experimental studies evidenced that resorption features are also acquired from mantle-stable corrosive fluids when diamonds are residing at depth (Table 15). In the studies of Chepurov et al. (1985) and Khokhryakov and Pal’yanov (1990), it was experimentally demonstrated for the first time that dissolution of flat-faced diamonds in water-bearing melts results in the formation of rounded crystals, with morphology identical to natural dodecahedroids and octahedroids. The reader is invited to read the detailed review of Fedortchouk (2019), which combines the observation of the morphology of natural diamonds with the description of dissolution and etching experiments, and demonstrates the usefulness of the experimental approach. It is shown that the shape and size of the etch pits on diamond surface depends on the temperature and H₂O/CO₂ ratio in the fluid, whereas pressure affects the efficiency of diamond crystal shape transformation from octahedral into rounded resorbed forms. Thanks to experiments, the identification of resorption produced in the mantle source from that in the kimberlite magma is possible.

Resorption processes possibly occur in between multiple growth events, such as observed in fibrous diamonds (Klein-BenDavid et al. 2007), recently it was proposed that such corrosive mantle fluids would likely be of “melt” nature (Fedortchouk et al. 2019). Surprisingly, it is also observed that both dissolution and growth may occur in very similar fluids/melts, such as in the system carbonate–silicate enriched in H₂O–CO₂ (Khokhryakov and Pal’yanov 2007a). Growth and dissolution events may alternate during diamond formation, possibly depending on carbon saturation, water versus CO₂ proportions, or redox conditions.

More experiments are necessary to decrypt the complexity of diamond’s histories and to understand the sequence of growth and dissolution events that may occur to diamonds from their birth in the mantle to the exposure to the near surface.

As can be seen from the preceding sections, there has been an impressive body of experimental work related to diamond formation over the past decades. The widespread adoption of experimental apparatus such as the belt, the multi-anvil, and the laser-heated diamond-anvil systems have allowed researchers to access the necessary P, T space to address a variety of questions with respect to diamonds. It should be noted that the experimental data reviewed here are not always in unambiguous agreement with each other. This is largely resulting from the use in different scientific groups of different equipment and methods for preparing and conducting experiments.

Building on the work done in industrial laboratories growing diamonds in a variety of metal melts, the ability of a rich variety of fluids and melts, including those possibly stable in the Earth’s mantle, to mediate crystallization of diamond has been firmly established. Actually, growing diamonds in these melts or fluids allows researchers to address questions such as isotopic fractionation during diamond growth and the incorporation of impurities into a growing diamond, which will provide valuable constraints to understanding the formation of diamonds in nature.

Despite all our collective efforts, some fundamental questions remain. For example, what triggers diamond formation? What is the carbon source, what is the nature of the diamond’s parents: fluids, melts, supercritical fluids? Is it depending on tectonic setting, or only on depth? How can we use the incorporation of impurities in diamonds to identify their growth mechanisms?

The experimental studies modeling slab-mantle reactions, whereby carbonate-bearing lithologies—representing subducted slab material—interact with reduced, metal-bearing mantle collectively are an excellent example of addressing a process question. But is this how all diamonds form? Classic models of diamond formation by, for example, oxidized fluids or melts interacting with reduced (but metal-free) mantle, or vice-versa, remain elusive to test experimentally. Other models, such as diamond formation by decompression or cooling of fluids (Stachel and Luth 2015), by partial melting in the presence of a hydrous, carbon-bearing fluid (Luth 2017; Smit et al. 2019), by pH changes in a fluid interacting with different mantle lithologies (Sverjensky and Huang 2015), under the influence of an electric field (Palyanov et al. 2021b), or by mixing of different fluids or melts (e.g., Huang and Sverjensky 2020) await experimental testing as well. There are still many experiments to perform to understand the growth of super deep diamonds in the transition zone and lower mantle. Another challenge for the future is the use of diamond growth to understand the mechanism of inclusions trapping: protogenetic versus syngenetic, or synchronous (Nestola et al. 2017) and thereby to provide information to properly decipher the complex messages delivered by inclusions about the deep Earth mineralogy and chemistry.

To conclude, there is still work to do! We need the experimental approach, in constant interactions with the study of natural diamonds and with models, to map the cartography of diamond growth at any depth into the Earth.

The authors gratefully acknowledge support for their research from their respective funding agencies: NSERC Discovery Grants (RWL), Russian Science Foundation under grant no. 19-17-00075 and state assignment of IGM SB RAS (YNP), and the program Emergences 2019 Sorbonne-Université, CNRS, contract 193256 Hydrodiams (HB). Thorough and constructive reviews by Yana Fedortchouk and an anonymous reviewer significantly improved the manuscript. Finally, the patience of the editors was much appreciated.