Diamond formation — Where, when and how?

Lithos 220–223 (2015) 200–220

Contents lists available at ScienceDirect

Lithos journal homepage: www.elsevier.com/locate/lithos

Invited review article

Diamond formation — Where, when and how? T. Stachel ⁎, R.W. Luth Department of Earth and Atmospheric Sciences, University of Alberta, Edmonton, AB, T6G 2E3, Canada

a r t i c l e

i n f o

Article history: Received 25 August 2014 Accepted 29 January 2015 Available online 14 February 2015 Keywords: Inclusion in diamond Diamond formation Solidus Redox reaction Peridodite Eclogite

a b s t r a c t Geothermobarometric calculations for a worldwide database of inclusions in diamond indicate that formation of the dominant harzburgitic diamond association occurred predominantly (90%) under subsolidus conditions. Diamonds in eclogitic and lherzolitic lithologies grew in the presence of a melt, unless their formation is related to strongly reducing CHO fluids that would increase the solidus temperature or occurred at pressure–temperature conditions below about 5 GPa and 1050 °C. Three quarters of peridotitic garnet inclusions in diamond classify as “depleted” due to their low Y and Zr contents but, based on LREEN–HREEN ratios invariably near or greater than one, they nevertheless reflect re-enrichment through either highly fractionated fluids or small amounts of melt. The trace element signatures of harzburgitic and lherzolitic garnet inclusions are broadly consistent with formation under subsolidus and supersolidus conditions, respectively. Diamond formation may be followed by cooling in the range of ~ 60–180 °C as a consequence of slow thermal relaxation or, in the case of the Kimberley area in South Africa, possibly uplift due to extension in the lithospheric mantle. In other cases, diamond formation and final residence took place at comparable temperatures or even associated with small temperature increases over time. Diamond formation in peridotitic substrates can only occur at conditions at least as reducing as the EMOD buffer. Evaluation of the redox state of 225 garnet peridotite xenoliths from cratons worldwide indicates that the vast majority of samples deriving from within the diamond stability field represent fO2 conditions below EMOD. Modeling reveals that less than 50 ppm fluid are required to completely reset the redox state of depleted cratonic peridotite to that of the fluid. Consequently, the overall reduced state of diamond stable peridotites implies that the last fluids to interact with the deep cratonic lithosphere were generally reducing in character. A further consequence of the extremely limited redox buffering capacity of cratonic peridotites is that redox reactions with infiltrating fluid/melt likely cannot produce large diamonds or high diamond grades. Evaluating the shift in maximum carbon content in CHO fluids during either isobaric cooling or ascent along a cratonic geotherm, however, reveals that isochemical precipitation of carbon from CHO fluids provides an efficient mode of diamond crystallization. Since subsolidus fluids are permissible in harzburgites only, and supersolidus melts in lherzolite we suggest that CHO fluid metasomatism may explain the long observed close association between diamonds and harzburgitic garnets. In the absence of thermodynamic data we cannot evaluate if supersolidus carbonatebearing melts, stable at fO2 conditions below EMOD, would experience a similar decrease in maximum carbon solubility during cooling or ascent along a geotherm. The absence of a clear association between diamond and lherzolitic garnets, however, suggests that this is not the case. A very strong association between diamond and eclogite likely relates to the fact that the transition from carbonate to diamond stable conditions occurs at redox conditions that are at least about 1 log unit more oxidizing than EMOD. At this time we cannot quantitatively evaluate the redox buffering capacity of cratonic eclogites but given their much higher Fe contents it has to be significantly higher than for peridotites. Alternatively, diamond in eclogite may precipitate directly from cooling carbonate-bearing melts that may be too oxidizing to crystallize diamond in olivine-bearing lithologies. © 2015 Elsevier B.V. All rights reserved.

Contents 1. 2.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Diamond substrates in Earth's mantle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

⁎ Corresponding author. E-mail address: [email protected] (T. Stachel).

http://dx.doi.org/10.1016/j.lithos.2015.01.028 0024-4937/© 2015 Elsevier B.V. All rights reserved.

201 201

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

3.

Pressure–temperature conditions of diamond formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Peridotitic suite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Eclogitic suite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3. Comparison of temperatures derived from non-touching and touching inclusion pairs: evidence for diamond formation during transient heating events? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1. Peridotitic inclusion pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.2. Eclogitic inclusion pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4. Pressure–temperature conditions of diamond formation and solidi of diamond host rocks . . . . . . . . . . . . . . . . . . . . . 3.5. Evidence from trace elements in support of melt-present and melt-absent diamond formation . . . . . . . . . . . . . . . . . . . 4. Timing of diamond growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Diamond forming reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Diamond formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Oxidized and reduced mineral inclusions in diamond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3. Oxidation state of Earth's upper mantle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4. Fluids in the upper mantle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. Ascent along lithospheric PT–fO2 path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6. “Isochemical” ascent or isobaric cooling of fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.7. Diamond precipitation from ascending or isobarically cooling melts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.8. The redox buffering capacity of cratonic peridotite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. Diamond forming processes based on co-variations in δ13C–N and on fluid inclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

201

. . . . . .

201 201 203

. . . . . . . . . . . . . . . . . . .

204 204 204 204 205 206 207 207 207 208 210 213 213 213 215 215 216 217 217

. . . . . . . . . . . . . . . . . . .

Following the discovery of the first kimberlite hosted diamond deposits in South Africa in 1870–1871, diamond formation was initially linked to reaction of the kimberlite magma with abundant carbonaceous shale fragments (crustal xenoliths) present in these diatremes (Lewis, 1887). The subsequent proposal of diamond representing a highpressure phenocryst in kimberlite was widely accepted (e.g., Williams, 1932) till the advent of geochemical studies of inclusions in diamond (Meyer, 1968; Meyer and Boyd, 1972; Sobolev et al., 1969) and radiometric dating of diamond formation ages (Kramers, 1979; Richardson et al., 1984), both implying crystallization in Earth's mantle unrelated to host kimberlite magmatism. The seminal suggestion of a xenocrystic origin for diamond in kimberlite (based on the observation of diamondiferous eclogite xenoliths), however, already dates back to Bonney (1899). Since the 1970s, numerous studies covering all important deposits around the globe have provided a detailed picture on the mineralogical and chemical environment for diamond formation, including the attendant pressure–temperature conditions (reviewed in Gurney, 1989; Meyer, 1987; Stachel and Harris, 2008). Following the recognition of diamond as mantle-derived xenocrysts in kimberlite, the concept emerged that diamond forms via redox reactions (Eggler and Baker, 1982; Luth, 1993; Rosenhauer et al., 1977) that relate to migration of a fluid or melt through a mantle host rock (Haggerty, 1986) and in consequence, diamond is now generally viewed as a metasomatic mineral (e.g., Stachel and Harris, 1997; Taylor et al., 1998). Little, however, is known about the exact composition and the redox character (carbonate- versus methane-bearing) of the fluids or melts that precipitate smooth-surfaced monocrystalline diamonds. In this contribution we use the large body of published data on diamonds and their mineral inclusions and less plentiful Fe3 +/Fe2 + determinations on minerals in cratonic garnet peridotites to discuss the “where, when and how?” of diamond formation and to place constraints on possible modes of diamond precipitation that invalidate some popular models.

studies have shown that diamonds derive from subcontinental lithospheric mantle extending into the diamond stability field (e.g., Boyd and Gurney, 1986) or may have originated at an even greater depth (extending to at least 700 km) in the sublithospheric mantle (Harte and Harris, 1994; Moore and Gurney, 1985; Scott Smith et al., 1984). Sublithospheric diamonds are, however, rare and the subcratonic lithospheric mantle represents the primary source of over 99% (by mass) of the worldwide diamond production (Stachel and Harris, 2008). In addition, the sublithospheric diamonds studied so far relate to the recycling of oceanic lithosphere into the deep mantle (Harte et al., 1999b; Stachel et al., 2000a, 2000b; Tappert et al., 2005; Walter et al., 2011) and hence can provide only very limited insights into diamond formation and storage in pyrolitic upper and lower mantle. For this contribution we will, therefore, focus exclusively on lithospheric diamonds. Based on their mineral inclusion content, diamonds from the lithospheric mantle (N = 2837) are divided into peridotitic (65%), eclogitic (33%) and websteritic (pyroxenitic) suites (2%). Using garnet compositions (N = 685), the peridotitic inclusion suite can be subdivided into harzburgitic (56% of all diamonds), lherzolitic (8%) and wehrlitic (0.7%) parageneses. The 86:13:1 harzburgite:lherzolite:wehrlite split of the peridotitic suite is virtually unchanged from the original pioneering work of Gurney (1984). The 2:1 ratio of peridotitic:eclogitic suite diamonds is based on destructive studies on diamonds generally b3 mm in size; it has, however, been speculated that among larger diamonds the relative proportion of the eclogitic suite may increase (e.g., Gurney, 1989; Stachel and Harris, 2008). In any case, 33% of all diamonds hosted in eclogite far exceed the b 1 to 5% estimated volumetric abundance of eclogite in subcratonic lithospheric mantle (Dawson and Stephens, 1975; McLean et al., 2007; Schulze, 1989). Equally, the ratio of harzburgitic:lherzolitic paragenesis diamonds (~ 7:1) reverses the relative proportions of harzburgite to lherzolite in diamond stable lithospheric mantle (ca. 1:4 for the Western Kaapvaal Craton; Griffin et al., 2003). This suggests that compared to lherzolite, harzburgite and eclogite are strongly preferred substrates for diamond (Grütter et al., 2004; Gurney, 1984).

2. Diamond substrates in Earth's mantle

3. Pressure–temperature conditions of diamond formation

The mineralogy and the mineral compositions of diamond host rocks in Earth's mantle are very well characterized through studies on mineral inclusions in diamond (reviewed in Gurney, 1989; Meyer, 1987; Meyer and Boyd, 1972; Shirey et al., 2013; Stachel and Harris, 2008). These

3.1. Peridotitic suite

1. Introduction

Diamond represents a closed system, with even the mobility of hydrogen being very low (Connell et al., 1998; Saguy, 2004). Non-touching

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

diamond transition (Fig. 1a). The regression line indicates a typical diamond window – representing the thickness of diamond stable lithospheric mantle bracketed by the intersections of the local geotherm with the graphite–diamond transition and the mantle adiabat – of 95 km thickness (110–205 km depth). Linear regression of the PNT 00–TNT 00 dataset of Grütter (2009) for garnet lherzolite xenoliths from the Central Slave Craton yields an almost identical local geotherm, underscoring that the Central Slave Craton represents a very appropriate choice as a reference geotherm for cratonic regions with high diamond potential (Grütter, 2009).

T [˚C] 600 3.0

800 ap

Dia

hit

e

1200

1400

Harzburgitic

mo

nd

4.0

Lherzolitic

5.0

Solidus Harzb+Carb+H2O

P [GPa]

1000

Solidus Lherz+CH4+H2O

Gr

b 6.0 Solidus Lherz+Carb+H2O Solidus Lherz+H2O

7.0 3.0

4.0

5.0

6.0

7.0

Adiabat TP=1300°C

Gr a Dia phite mo nd

45

a PNT 00-PNT 00 Lherzolitic PBKN-THarley Lherzolitic Harzburgitic

40

35

inclusions (i.e., isolated single-phase mineral inclusions) in diamond are, therefore, assumed to have remained unchanged since encapsulation — rare cases with microscopically visible exsolution features aside (Leost et al., 2003). Disequilibrium has been documented among non-touching inclusions incorporated during diamond growth in chemically evolving environments (e.g., Bulanova, 1995; Griffin et al., 1988; Rickard et al., 1989; Stachel et al., 1998). Multiple inclusions in single diamonds suitable for the application of independent geothermometers allow testing for possible disequilibrium. Worldwide, 13 diamonds have been studied containing inclusions of garnet + orthopyroxene + olivine (database of Stachel and Harris, 2008), allowing to derive temperature estimates based on the Mg–Fe exchange between both garnet–olivine (O'Neill, 1980; O'Neill and Wood, 1979) and garnet–orthopyroxene (Harley, 1984); accounting for the systematic deviations between the two thermometers documented by Brey and Köhler (1990), the two methods yield temperature estimates agreeing within uncertainty for 12 of the diamonds (i.e., N 90% show no evidence for internal disequilibrium; Supplementary Fig. 1). Adding further evidence derived through detailed examination of tie-lines connecting coexisting mineral pairs (e.g., Otter and Gurney, 1989; Rickard et al., 1989) and observation of common trace element equilibrium between multiple inclusions (e.g., garnet– clinopyroxene pairs; Stachel et al., 2000a), disequilibrium between nontouching inclusions is considered the exception rather than the rule (Gurney, 1989). On this basis, the pressure–temperature conditions of diamond formation can be derived through mineral exchange geothermobarometry employing silicate inclusions. In its primary stability field and at temperatures ≥1000 °C, diamond is not a rigid pressure vessel but will transfer changes in ambient pressure to included minerals by means of plastic deformation (De Vries, 1975). Touching minerals will thus re-equilibrate under changing pressure–temperature conditions and, therefore, reflect conditions of the final mantle storage. The geothermobarometers applicable to a comparatively large number of inclusions are: (1.) for garnet–orthopyroxene pairs, the Al exchange barometer of Brey and Köhler (1990; PBKN) combined with the Mg–Fe exchange thermometer of Harley (1984; THarley) and (2.) for clinopyroxene inclusions assumed to be in equilibrium with orthopyroxene and garnet, the single crystal geothermobarometer of Nimis and Taylor (2000; PNT 00 and TNT 00). The former is applicable to the dominant harzburgitic and the lherzolitic inclusion paragenesis; the latter can only be applied to comparatively rare diamonds with lherzolitic inclusions. The combination of PBKN–THarley owes its relatively large uncertainty (1 sigma of ~100 °C and 0.5 GPa) mainly to systematic errors in the thermometer calibration (above 1000 °C the thermometer increasingly underestimates temperature; Brey and Köhler, 1990). The experimental calibration of the PBKN barometer extends to 6.0 GPa; extrapolation to higher pressures adds additional uncertainty. The better precision of PNT 00–TNT 00 (~50 °C and 0.3 GPa) only holds to pressure of up to ~ 4.5 GPa; at higher pressure the barometer increasingly underestimates pressure (Nimis, 2002). Application of PNT 00–TNT 00 requires rigorous filtering of clinopyroxene analyses (using cation totals as a quality criterion and then excluding all compositions outside the experimental ranges; see Grütter, 2009 for details), which for the current study eliminated 40% of the available data (N = 134). Fig. 1 illustrates P–T estimates (including 26% touching inclusion pairs) using the two geothermobarometer combinations discussed above. Based on 157 independent estimates (82 diamonds with lherzolitic and 75 with harzburgitic inclusions), the average peridotitic suite diamond derives from 5.3 ± 0.8 GPa and 1130 ± 120 °C. Lherzolitic and harzburgitic inclusions yield similar means (5.1 ± 0.8 GPa and 1120 ± 140 °C, and 5.6 ± 0.8 GPa and 1140 ± 90 °C, respectively). The lherzolitic data, however, extend to low temperature conditions (640–970 °C) that are not observed for the harzburgitic inclusion set. A regression line through the combined dataset is slightly oblique to the continental heat flow geotherms of Hasterok and Chapman (2011), crossing from a 39 mW/m2 model geotherm at the intersection with the mantle adiabat to a 36 mW/m2 geotherm at the graphite

P [GPa]

202

600

800

1000

1200

1400

T [˚C] Fig. 1. (a): Pressure–temperature estimates for peridotitic suite diamonds based on garnet– orthopyroxene inclusion pairs (PBKN–THarley) and clinopyroxene inclusions (PNT 00 and TNT 00). The latter combination is only applicable to the lherzolitic inclusion paragenesis. The filled red circle and surrounding pale red error ellipse represent the average and 1 standard deviation for the entire dataset (5.3 ± 0.8 GPa and 1130 ± 120 °C; N = 157). A linear regression through the dataset (r2 = 0.53) is shown as a red line and represents an average geotherm for global diamondiferous lithospheric mantle. Continental geotherms (blue dashed lines with surface heat flow values in mW/m2; Hasterok and Chapman, 2011), a mantle adiabat for a potential temperature of 1300 °C (Hasterok and Chapman, 2011), and the graphite–diamond transition (Day, 2012) are shown for reference. (b): Fields of P–T conditions for lherzolitic and harzburgitic inclusions in diamonds compared to a range of solidus temperatures in the presence of CHO-fluids. The solidus for hydrous melting of carbonated lherzolite (red short dashed line) is from Wyllie and Ryabchikov (2000). The solidi of lherzolite in the presence of a hydrous fluid (red long dashed line) and a reduced CHO fluid (red dotted line) are from Wyllie and Ryabchikov (2000) and Litasov et al. (2014), respectively. The solidus of harzburgite in the presence of carbonate and H2O (blue dashed-dotted line) is taken from Wyllie (1987). The error ellipse around the average value of P–T estimates for peridotitic suite diamonds is shown as a black dashed line.

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

The significance of the relatively small difference in average P–T conditions for peridotitic suite diamonds of lherzolitic and harzburgitic paragenesis is difficult to evaluate as it likely results from iterations in two different geothermobarometer combinations (e.g., Nimis and Grütter, 2010). Assuming an ~5.0 GPa average pressure, the temperature conditions for diamonds hosted by lherzolite and harzburgite can be compared based only on the garnet–olivine Mg–Fe exchange geothermometer (O'Neill, 1980; O'Neill and Wood, 1979; T O'Neill). Based on TO'Neill (Fig. 2a), harzburgitic (N = 144, average = 1170 ± 120 °C, median = 1170 °C) and lherzolitic (N = 23, average = 1150 ± 170 °C, median = 1170 °C) inclusions yield the same median values and an equality of means cannot be rejected (Student's t-test at α = 5%). This suggests that harzburgitic and lherzolitic paragenesis diamonds grew in different substrates but under comparable P–T conditions.

3.2. Eclogitic suite The most widely used and tested thermometer for mantle eclogites is the garnet–clinopyroxene Mg–Fe exchange thermometer of Krogh (1988; TKrogh 88). Brey and Köhler (1990) found that TKrogh 88 reproduces their lherzolite system experiments within about 40 °C (1 sigma) over the

TKrogh 88 [°C] 800 40

900

1000

1100 1200 1300 1400

Eclogitic: N=144 30 Mean=1170 ±110 °C Median=1160 °C

N 20

b

203

entire 900–1400 °C experimental range. Purwin et al. (2013) conducted eclogite system experiments specifically designed to evaluate the effect of highly variable Fe3 +/Fe2 + in their starting materials over an 800–1300 °C temperature range. They found that, if total iron is used as input for TKrogh 88, their experiments are reproduced over the entire temperature range within 60 °C, whilst significant systematic temperature estimation errors occur when measured Fe2+ is employed. Consequently, results obtained for eclogitic inclusions in diamond using total Fe measurements via EPMA and the Krogh (1988) thermometer can be considered both accurate and precise. A reliable barometer suitable for mantle eclogites and eclogitic inclusions in diamond is currently not available (Nimis and Grütter, 2010). Assuming that diamonds hosted by eclogite form at similar pressure as peridotitic suite diamonds, a fixed value of 5.0 GPa is applied for thermometric calculations. On this basis, 144 eclogitic garnet–clinopyroxene pairs (including 11% touching inclusion pairs) yield an average equilibration temperature of 1170 ± 110 °C (median = 1160 °C; see Fig. 2b). This compares very well to temperatures derived from equally abundant (N = 164) garnet–olivine pairs (TO'Neill) in peridotitic suite diamonds, which yield a mean of 1160 ± 110 °C (median = 1140 °C; see Fig. 2a) for the same assumed pressure. This excellent agreement suggests that diamonds hosted by peridotite and eclogite indeed share a common origin within diamond stable lithospheric mantle. In the absence of a reliable barometer, pressure estimates for mantle eclogites are usually based on projections of calculated temperatures on independently derived local geotherms. Following a similar approach (Fig. 3), the average temperature and one sigma range (1170 ± 110 °C) for eclogitic inclusions in diamond are projected on both the average Precambrian shield geotherm of Hasterok and Chapman (2011; 40 mW/m2) and the peridotitic inclusion-based diamond geotherm (Fig. 1a). A depth of origin for 70% (samples within 1 sigma) of eclogitic suite diamonds between 135–190 km is obtained for the average shield geotherm, and 145–200 km for the peridotitic inclusion-based diamond “geotherm” (Fig. 3). The latter projection results in P–T conditions that are about 0.5 GPa and 70 °C higher than those for peridotitic suite diamonds.

3.0

10

4.0

Peridotitic: N=164 Mean=1160 ±110 °C 30 Median=1140 °C

Harzburgitic

Peridotitic Diamond P-T 5.0 Regression

45

6.0 40

a

N 20

1 sigma

7.0

Solidus Eclogite+Carbonate 35

Lherzolitic 10

600

800

1000

1200

Solidus Eclogite+CH4+H2O

P [GPa]

0

GraphiteDiamond Transition

Adiabat TP=1300°C

Solidus Basalt+H2O

1400

T [˚C] 0 800

900

1000

1100 1200 1300 1400

TO’Neill [°C] Fig. 2. Temperature estimates for diamonds with peridotitic (a; garnet–olivine pairs, TO'Neill) and eclogitic inclusions (b; garnet–clinopyroxene pairs, TKrogh 88). Calculations assume a fixed pressure of 5.0 GPa. Temperatures above 1400 °C were excluded as they exceed the mantle adiabat (Fig. 1a) by more than 50 °C. For the eclogitic inclusion set (b), all data from the Argyle Diamond Mine were excluded as they represent atypically “hot” conditions (Stachel et al., in press).

Fig. 3. Average temperature and 1 sigma range (1170 ± 110 °C; TKrogh 88 at 5 GPa) for 144 eclogitic garnet–clinopyroxene inclusion pairs shown as solid red circles and bold red lines, in a pressure corrected projection onto two geotherms: the 40 mW/m2 average Precambrian shield geotherm of Hasterok and Chapman (2011; as labeled) and the peridotitic suite diamond geotherm at slightly higher pressures (from Fig. 1a). The solidi for hydrous eclogite (Kessel et al., 2005; solid green line) and carbonated eclogite (Dasgupta et al., 2004; dashed green line) occur at lower temperatures, implying melt present conditions for diamond precipitation in eclogite, except under reducing conditions (IW buffer, with CH4–H2O present; Litasov et al., 2014; long dashed blue line). The hydrous solidus terminates at a second critical endpoint (star) between 5–6 GPa (Kessel et al., 2005). Mantle adiabat, conductive geotherms and graphite-diamond transition as in Fig. 1.

204

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

3.3. Comparison of temperatures derived from non-touching and touching inclusion pairs: evidence for diamond formation during transient heating events? 3.3.1. Peridotitic inclusion pairs An extensive geothermobarometric study to evaluate differences between the conditions of diamond formation (8 non-touching pairs) and final mantle residence (27 touching pairs) was conducted on garnet–orthopyroxene inclusions in diamonds from the De Beers Pool kimberlite pipes in South Africa (Phillips and Harris, 1995; Phillips et al., 2004). Based on a combination of PBKN and THarley it was found that diamond formation (average non-touching inclusion pairs: 6.2 GPa and 1200 °C) was followed by cooling (average touching pairs: 5.4 GPa and 1080 °C) by about 120 °C (Phillips et al., 2004). Recalculating the entire dataset of Phillips et al. (2004) at a fixed pressure of 5.0 GPa, however, strongly reduces the average temperature difference to about 60 °C (Fig. 4a). The decrease of 0.8 GPa in pressure may indicate that the residence history of diamonds in the lithospheric mantle beneath Kimberley (Western Kaapvaal) was more complicated than simple cooling and involved differential uplift of almost 30 km as well (Phillips et al., 2004). This number compares well to the loss

T O’Neill or Harley [˚C] 900 4

1000

Panda

1100

1200

1300

Nontouching

in lithosphere thickness for the Western Kaapvaal Craton of ~20 km occurring between kimberlite emplacement at Finsch (124 Ma) and Kimberley (94–86 Ma; Griffin et al., 2003) and may suggest that the uplift reflected in the inclusion in diamond data was related to extension of the lithospheric mantle. For diamonds from the Panda kimberlite in the Central Slave Craton, Canada, thermometric data obtained from non-touching (4 pairs) and touching (4 pairs) garnet–olivine (TO'Neill) and garnet–orthopyroxene (THarley) pairs indicate that diamond formation was followed by cooling (Stachel et al., 2003) by about 150 °C (Fig. 4b). Nitrogen aggregation based thermometry (Leahy and Taylor, 1997; assuming 3 Ga mantle residence) indicates that time averaged mantle residence temperatures for Panda diamonds with non-touching pairs (giving diamond formation temperatures) are on average about 30 °C lower than the mineral exchange temperatures but about 120 °C higher for diamonds containing touching inclusion pairs (giving final residence temperatures). This suggests that slow lithosphere-scale cooling (100 s of million to billion year time scale) rather than a short lived local heating event associated with diamond formation (million year time scale) may have occurred in the Central Slave Craton. Evidence that diamond formation in peridotite is not invariably followed by cooling can be drawn from two diamonds each containing a clinopyroxene and a clinopyroxene–orthopyroxene pair (diamonds MW-80 from Mwadui in Tanzania and Nam-211 from Namibia). Calculating TNT 00 at 5.0 GPa, in both cases the single clinopyroxene inclusion records temperatures about 40 °C lower than the touching pair. These temperature differences are slightly larger than the error on the temperature estimates (1 sigma of ±30 °C) and suggest that thermal regimes may also slightly increase following diamond formation.

Frequency

3

Touching 2

b

1

0

De Beers Pool

Touching

Frequency

15

10

a Nontouching

5

0 900

1000

1100

1200

1300

T Harley [˚C] Fig. 4. Temperature estimates for non-touching (blue) and touching (red) peridotitic inclusion pairs in diamonds from the De Beers Pool Mines (a) and the Panda kimberlite of the Ekati Mine (b). Temperatures were calculated assuming a fixed pressure of 5.0 GPa and are based on garnet–orthopyroxene (THarley; De Beers Pool and Panda) and garnet–olivine (TO'Neill; only Panda) inclusion pairs. For De Beers Pool diamonds (a), the average temperature for non-touching inclusions is 1120 °C versus 1060 °C for touching inclusions (implying cooling by 60 °C from diamond formation to final mantle residence). For Panda diamonds (b), the average temperature for non-touching inclusions is 1190 °C versus 1040 °C for touching inclusions (150 °C drop in temperature from diamond formation to final mantle residence).

3.3.2. Eclogitic inclusion pairs For diamonds hosted by eclogite the presence of touching and nontouching pairs of garnet and clinopyroxene inclusions can only be confidently identified in literature data for the kimberlites at De Beers Pool (Phillips et al., 2004), George Creek, USA (Chinn, 1995), and Jagersfontein, South Africa (Tappert et al., 2005). For De Beers Pool and George Creek, touching inclusions record temperatures that are distinctly lower (differences in average TKrogh 88 of 100 and 180 °C, respectively; Fig. 5a and b) than obtained from non-touching pairs, suggesting that diamond formation was followed by cooling. These results are consistent with a ~ 170 °C cooling estimate (recalculated using TKrogh 88) for eclogitic clinopyroxene and touching garnet–clinopyroxene in an unsourced diamond initially described by Meyer and Tsai (1976), based on analyses of Prinz et al. (1975). In the case of Jagersfontein, the temperatures calculated for touching and non-touching pairs agree within error of the thermometer (touching inclusions indicate temperatures that are 20 °C higher; Fig. 5c). Similarly, at Argyle, for a diamond containing one touching garnet–clinopyroxene pair and two separate garnets, calculated temperatures (TKrogh 88) agree to within 10 °C (i.e., within error). So similar to peridotitic suite diamonds, in some cases formation of eclogitic diamonds is followed by 100–180 °C cooling whilst in other cases diamond formation and ultimate storage take place in a thermally stable mantle environment. 3.4. Pressure–temperature conditions of diamond formation and solidi of diamond host rocks Based on a layer-by-layer growth mode, smooth-surfaced octahedral diamond is presumed to grow into open space (Sunagawa, 1984), implying the presence of a melt or a high density fluid. This premise can be tested by comparing inclusion-based temperature estimates to experimentally determined solidus temperatures for possible diamond host lithologies. Diamond forms as a consequence of the influx of carbon-bearing fluids/melts into pre-existing substrates (e.g., Haggerty, 1986; Stachel and Harris, 1997; Taylor and Green, 1988). Since carbonation reactions

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

800 3

900

1000

1100 1200 1300 1400

Jagersfontein

Frequency

Nontouching 2

c 1

Touching

0

George Creek

Frequency

6

Nontouching

4

b

Touching

2

0

De Beers Pool

Frequency

6

Nontouching 4

a Touching 2

0 800

900

1000

1100 1200 1300 1400

T Krogh 88 [˚C] Fig. 5. Temperature estimates for non-touching (blue) and touching (red) eclogitic inclusion pairs in diamonds from the De Beers Pool Mines (a), George Creek, USA (b) and Jagersfontein, South Africa (c). Temperatures were calculated assuming a fixed pressure of 5.0 GPa and are based on garnet–clinopyroxene (TKrogh 88) inclusion pairs. For De Beers Pool diamonds (a), the average temperature for non-touching inclusions is 1170 °C versus 1070 °C for touching inclusions (i.e., a 100 °C drop in temperature from diamond growth to final mantle residence). For diamonds from George Creek (b), the average temperature for non-touching inclusions is 1110 °C and for touching inclusions pairs 930 °C, implying cooling by 180 °C. For Jagersfontein diamonds (c), the average temperature for non-touching inclusions is 1140 °C versus 1160 °C for touching inclusions (implying an increase in temperature by 20 °C from diamond genesis to final mantle residence). Excluding the one touching pair yielding a very high temperature estimate (1380 °C), cooling by 30 °C is derived. Both temperature differences for Jagersfontein are within error of the thermometer.

prevent the migration of CO2 through olivine-bearing rocks at high pressure (Wyllie and Huang, 1976), the spectrum of oxidized to reduced fluid species during peridotite melting ranges from H2O–carbonate through pure H2O and H2O–CH4 to H2O–CH4–H2 (Foley, 2011; French, 1966; Taylor and Green, 1989). With respect to solidus temperature,

205

peridotite plus a H2O–carbonate fluid and carbonated peridotite plus H2O are equivalent. The solidus for hydrous melting of carbonated lherzolite shown in Fig. 1b is taken from Wyllie and Ryabchikov (2000) and based on extrapolation of experiments b3.5 GPa. However, there is excellent agreement between this extrapolation and the hydrous solidus of carbonated lherzolite between 4.0–6.0 GPa experimentally determined by Foley et al. (2009). In the pressure region shown in Fig. 1b, the hydrous solidus of lherzolite (from Wyllie and Ryabchikov, 2000; principally based on Kushiro et al., 1968) is only about 50 °C higher. P–T estimates for lherzolitic inclusions dominantly fall above the hydrous solidi for lherzolite and carbonated lherzolite. At lower pressures (≤5.0 GPa equivalent to ≤160 km depth) there is, however, a group of “low temperature” (b1050 to b1000 °C) lherzolitic inclusions that were encapsulated by diamond under sub-solidus conditions. Because of the low solubility of CH4 and H2 in silicate melts, reduced fluids inhibit melting (Green, 1990). In mixed CH4–H2O fluids, due to strong non-ideality, with increasing CH4 the activity of H2O decreases more rapidly than the actual H2O content in the fluid, leading to a rapid increase in CH4 activity and solidus temperature (Foley, 2011). The solidus for lherzolite in the presence of a reduced CHO fluid shown in Fig. 1b (Litasov et al., 2014) is for very reducing conditions (IW buffered), resulting in a methane dominated fluid with a water content increasing from 13% at 3 GPa to 25% at 6 GPa. Temperatures in conductively equilibrated peridotites are below the solidus of lherzolite in the presence of a reduced fluid (Fig. 1b). The solidus temperature for harzburgite in the presence of carbonate and H2O (Fig. 1b; estimate of Wyllie, 1987) is also high and exceeded by only a few harzburgitic inclusions pairs derived from deeper portions of lithospheric mantle (N 6.0 GPa equivalent to N 190 km). We conclude that subsolidus fluids (H2O–CO2− to H2O–CH4) are the principal diamond forming 3 agent in harzburgitic substrates but only play a similar role in lherzolites if either they are reducing (CH4-rich) or diamond formation occurs below the H2O–carbonate solidus (i.e., at b 5 GPa along model geotherms b40 mW/m2; Fig. 1). Under diamond stable conditions the solidus of anhydrous and noncarbonated metabasalt distinctly exceeds the mantle adiabat (Yasuda et al., 1994). However, similar to their peridotitic counterparts, the formation of diamonds in eclogite is related to the influx of fluids and/ or melts (Ickert et al., 2013; Keller et al., 1998; Taylor et al., 1998). Contrary to peridotite, in the presence of a hydrous fluid and along typical cratonic geotherms CO2 is not buffered in bimineralic eclogite (Knoche et al., 1999; Luth, 1993). The experimentally determined solidi of metabasalt in the presence of water (Kessel et al., 2005) or carbonate (Dasgupta et al., 2004) provide the two most relevant end-member conditions (Fig. 3). Kessel et al. (2005) found that the hydrous eclogite solidus terminates at a second critical end-point at 5–6 GPa, after which the onset of melting cannot be defined anymore. Experiments on melting of eclogite under strongly reducing conditions (Litasov et al., 2014; IW buffered), i.e. in the presence of a methane dominated fluid (b 20% H2O in the pressure range of Fig. 3), document an ~200 °C increase in solidus temperature relative to hydrous and carbonated eclogite at 5.0 GPa. Our thermometric results for diamonds in eclogitic substrates indicate that in the presence of water or carbonate, their formation occurred in a source that was partially molten (Fig. 3). In the presence of a reducing fluid, eclogite melting would only occur in the deeper portions of subcratonic lithospheric mantle (N 5.5 GPa equivalent to N175 km). 3.5. Evidence from trace elements in support of melt-present and meltabsent diamond formation Fluid and melt metasomatism have distinctive trace element signatures, with the ratios of highly to mildly incompatible elements (e.g., LREE/HREE or Zr/Y) decreasing strongly from fluids to melts (e.g., Griffin and Ryan, 1995; Stachel et al., 2004; Stosch and Lugmair, 1986). Griffin and Ryan (1995) empirically devised fields in a diagram of Y versus Zr content in garnet that permit discrimination between

206

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

low temperature (fluid) metasomatism and melt metasomatism (Fig. 6a). The majority (about 75%) of garnet inclusions analyzed for these elements fall into the “depleted” field (Fig. 6a), where no clear distinction of metasomatic style can be made below Y b5 ppm and Zr b20 ppm. Plotting the HREE content (YbN, with N = C1-chondrite normalized) versus the degree of sinuosity of REEN patterns (NdN/ErN; Fig. 6b), nevertheless allows to derive evidence for ubiquitous metasomatic overprint for the strongly depleted group of garnets (as predicted by Frey and Green, 1974). This implies that formation of diamond is typically associated with mild metasomatic events reflecting low fluid–rock or melt–rock ratios. For HREE depleted garnets, LREEN–HREEN ratios near or greater than one imply overprint by a strongly LREE enriched fluid (e.g., Stachel et al., 2004). For depleted garnets showing some re-enrichment in HREE, sinuosity of REE patterns is comparatively low (NdN/ErN b 10; Fig. 6b), suggestive of overall mild metasomatism involving a melt. If all depleted garnets with NdN/ErN N 20 are interpreted to relate to fluid metasomatism and all garnets with YbN N 2 to indicate mild melt metasomatism, then (with one exception) depleted lherzolitic garnets reflect melt metasomatism whilst depleted harzburgitic garnets reflect both mild fluid and mild melt metasomatism (Fig. 6b). About half the depleted garnets (with YbN b 2 and NdN/ErN b 20), although clearly derived from a metasomatized source (otherwise NdN/ErN should be ≪1), cannot be assigned to a particular style of overprint. For the more strongly metasomatized inclusions (falling outside the “depleted” field in Fig. 6a) it is evident that lherzolitic garnets have high Y/Zr, typical of melt infiltration into their host rocks, whereas harzburgitic garnets predominantly show low Y/Zr (typical for low-temperature metasomatism) but with a number of exceptions (as predicted from the position of the hydrous solidus; Fig. 1b). Overall, the trace element evidence hence supports the conclusions made in Section 3.4. The low-temperature (fluid) metasomatism evident in Fig. 6a is, however, not a simple extension of the fluid metasomatic trend visible in Fig. 6b, since sinuosity of garnet REEN patterns actually decreases towards high Zr contents (e.g., all garnets with Zr N 50 ppm have NdN/ErN b 5). This suggests that increasing Zr contents are accompanied by decreasing LREE–HREE fractionation in metasomatic fluids. It should, however, be noted that such an approach will only identify the dominant re-enrichment event affecting a diamond source region,

which may not necessarily be the diamond forming event; e.g., during a thermal pulse a harzburgitic diamond substrate may have been affected by melt metasomatism, but diamond only formed later, after thermal relaxation, during a mild fluid metasomatic event that essentially remains invisible in the trace element signature of inclusions. In fact, Richardson et al. (1984) concluded that a melt metasomatic event must have preceded diamond formation in the lithospheric mantle beneath Kimberley and Finsch (Western Kaapvaal) by about 300 Ma, based on unsupported highly radiogenic Sr in harzburgitic garnet inclusions. 4. Timing of diamond growth Cathodoluminescence imaging reveals that diamonds often have complex internal growth structures, including oscillatory zoning or the presence of sharp boundaries that may show evidence for intermittent periods of diamond resorption (e.g., Harte et al., 1999a; Wiggers de Vries et al., 2013a). In situ studies of micron scale variations in nitrogen content and carbon isotopic composition demonstrate that such sharp internal boundaries are often associated with abrupt compositional changes, which may include switches in the redox character (carbonate versus methane bearing) of the precipitating fluid/melt (Palot et al., 2013; Peats et al., 2012; Wiggers de Vries et al., 2013a). Grouping sulfides from Mir and 23rd Party Congress (Yakutia) based on their composition, the nitrogen characteristics and carbon isotopic compositions of the associated growth zones in their host diamonds, and coexisting silicate inclusions, Wiggers de Vries et al. (2013b) inferred that these sulfide groupings likely document protracted diamond growth spanning up to ~1 Ga. In this context, diamond formation ages that are not based on the dating of single inclusions have to be viewed with some caution. However, the consistency of certain diamond formation ages within and across cratons (see below) and, in particular, the observed close agreement of Re–Os sulfide and Sm–Nd silicate isochron ages for diamonds from individual occurrences (reviewed in Stachel and Harris, 2008) cannot be coincidental and suggest that the temporal extent of individual diamond growth events is usually well contained within the uncertainty of the age dates. Isotopic dating using garnet, clinopyroxene and sulfide inclusions (reviewed in Gurney et al., 2010; Pearson and Shirey, 1999; Stachel

40

6 Undepleted Garnets

Lherzolitic Harzburgitic

a

Only garnets with Y<5 and Zr<20 ppm 4

m

is

at

om

as

Y 20

Melt

30

b

Yb N

et

M lt

e

2

lo go is pit m e)

M

10

at

h (P Low-T om s Meta

0

Fluid 0

0

Depleted 50

100

Zr

150

0

20

40

60

80

NdN/ErN

Fig. 6. (a): Y versus Zr (wt. ppm) content in peridotitic garnet inclusions in diamond from worldwide sources. Empirical compositional fields and trends are from Griffin and Ryan (1995). The formation of subcratonic lithospheric mantle (SCLM) involved intense melt extraction and consequently, if no major re-enrichment event occurred, garnets from SCLM plot in the depleted field (Y b7 ppm and Zr b30 ppm). Low temperature (fluid) metasomatism involves preferential re-enrichment in Zr and is restricted to harzburgitic garnet inclusions. Simultaneous addition of Zr and Y, attributed to melt metasomatism, is observed for lherzolitic and some harzburgitic garnet inclusions. This is consistent with the observation (see Fig. 1b) that diamond formation mostly occurs above the solidus of lherzolite (in the presence of a CHO fluid) but below the solidus of harzburgite. The majority of garnet inclusions, however, plot into the depleted field, and for garnets with Y b5 ppm and Zr b20 ppm assignment to extrapolated metasomatic trends is no longer practical. For such strongly “depleted” garnets (b, right), ubiquitous metasomatic overprint is nevertheless documented by LREE–HREE ratios (NdN/ErN) invariably near or greater than one. Modest HREE enrichment at low NdN/ErN is interpreted to represent mild melt metasomatism (trend along the Y-axis) whilst strongly sinusoidal REEN at low HREE (trend along X-axis) document overprint by a strongly fractionated fluid Normalization to C1-chondrite (N) after McDonough and Sun (1995).

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

or carbonate melts, so precipitation of diamond from such melts would require an oxidation or reduction reaction. In general, such reactions could involve reduction of oxidized species such as CO23 − or CO2, or oxidation of reduced species such as CH 4 . In addition to being dissolved in silicate or carbonate melts, oxidized species could also be present as crystalline carbonate or in a fluid. Reduced carbon could be present in a melt, a fluid, or under very reduced conditions in phases such as moissanite (SiC; see Di Pierro et al., 2003; Shiryaev et al., 2011; Ulmer et al., 1998). Fluids at mantle conditions may contain oxidized carbon-bearing species such as CO 2 or reduced species such as CH4, depending on the oxidation state. There is a constraint on CO2 -bearing fluids in peridotitic lithologies, however. Since the landmark work of a number of research groups in the 1970s, it has been clear that CO2-rich fluids would not be stable in oxidized peridotitic mantle, but would react to form carbonates such as dolomite or magnesite by reactions such as forsterite + diopside + CO2 → enstatite + dolomite and forsterite + CO2 → enstatite + magnesite (Fig. 7). Another reaction relevant to peridotitic mantle is the exchange reaction by which enstatite and dolomite react to form the higher-pressure assemblage magnesite and diopside. Comparing these reactions with possible geotherms for lithospheric mantle (Fig. 7), it is apparent that magnesite would be the stable carbonate under most circumstances in oxidized lithospheric peridotitic mantle, and thus carbon would be present in magnesite rather than in a CO2-rich fluid. At more reduced conditions, magnesite becomes unstable relative to graphite or diamond via the reaction enstatite + magnesite → forsterite + graphite/diamond + O2 (the EMOG/EMOD reactions of Eggler and Baker, 1982). Whether the mantle is sufficiently oxidized to stabilize carbonate requires evaluation of the evidence that constrains the oxidation state of diamond substrates in the upper mantle. 5.2. Oxidized and reduced mineral inclusions in diamond Mineral inclusions in diamond place constraints on the range of redox conditions associated with natural diamond formation. The upper limit of

5. Diamond forming reactions

1 2

3

100

5.1. Diamond formation

P [GPa]

adiabat

4

3

140 45

5

180 6 40

7

220 35

Diamond may form in Earth's mantle by a variety of processes: recrystallization of the low-pressure graphite polymorph, precipitation from a fluid or melt saturated with carbon, or by oxidation–reduction reactions involving carbonate or methane. The direct conversion of graphite to diamond is a reconstructive phase transition, and in the absence of a flux or solvent requires considerable overstepping of the reaction boundary; Irifune et al. (2004) found that 15 GPa and 1800 °C, or 12 GPa and 2000 °C were required for this reaction to proceed in the laboratory in a static compression mode. On the other hand, graphite is commonly used as a carbon source in experiments that grow diamond in the presence of metal, carbonate, and silicate melts, as well as CHO fluids at pressure–temperature conditions much closer to the graphite–diamond reaction curve. The operative mechanism in these cases would be one of dissolution of graphite into the melt or fluid, and re-precipitation as diamond resulting from the higher solubility of the metastable polymorph relative to the stable one at the experimental P–T conditions. Industrial syntheses of diamond at high P and T use metallic melts as growth media. Similar metallic melts may be responsible for diamond growth in the mantle at depths greater than ~250 km where conditions become sufficiently reduced to stabilize metal (Frost and McCammon, 2008; Rohrbach et al., 2007, 2011). In the shallower lithospheric mantle, conditions are typically too oxidized for metal stability. There is no evidence at present for dissolution of elemental carbon in either silicate

Gr Dia

Depth [km]

and Harris, 2008) documents that formation of diamonds occurred through most of Earth's history (from the Paleoarchean to at least the Mesozoic). Diamond forming episodes in subcratonic lithospheric mantle occur on regional to global scales in response to tectonothermal events such as suturing, subduction and plume impact (e.g., Aulbach et al., 2009; Gurney et al., 2010; Shirey et al., 2002). Individual diamond forming episodes may be associated with particular substrates, with harzburgitic paragenesis diamonds generally yielding Paleoarchean (3.6–3.2 Ga) ages and lherzolitic paragenesis diamonds forming mostly in the Paleoproterozoic at about 2 Ga (see references in Gurney et al., 2010; Stachel and Harris, 2008). Harzburgitic garnet inclusions from Udachnaya, however, yield a 2 Ga isochron age (Richardson and Harris, 1997), coinciding with the worldwide peak in lherzolitic diamond formation and younger episodes of peridotitic diamond growth are established through Mesoproterozoic isochron ages for lherzolitic sulfide inclusions in diamonds from Ellendale (Western Australia, 1.4 Ga; Smit et al., 2010), Mir and 23rd Party Congress (the second diamond growth event in Yakutia, 1.0–0.9 Ga; Wiggers de Vries et al., 2013b), a Mesozoic isochron age obtained from two peridotitic sulfide inclusions in a single diamond from Koffiefontein (within error of kimberlite emplacement at 90 Ma; Pearson et al., 1998), and an inferred similar age for a peridotitic sulfide inclusion from Jagersfontein (Aulbach et al., 2009). Formation of diamonds hosted by eclogite is documented from the Mesoarchean to the Neoproterozoic (2.9 and 0.6 Ga) and may well continue up to the present. For fibrous coats (overgrowths over pre-existing monocrystalline diamonds) from Aikhal, Siberia, an Ar–Ar study suggested that the coat-forming event occurred close to the time of host kimberlite eruption (Burgess et al., 2002). This is consistent with infrared spectroscopic studies on fibrous diamonds (coats and cuboids) from southern and western Africa, Siberia and Australia, documenting low nitrogen aggregation states (pure Type IaA; Boyd et al., 1987, 1992) and thereby indicating that their formation likely occurred penecontemporaneously with and probably initiated by host kimberlite activity (Boyd et al., 1987, 1992; Gurney et al., 2010). Fibrous coats on diamonds from tertiary kimberlites at Lac de Gras, Canada (Gurney et al., 2004), consequently, may represent the youngest diamond samples available.

207

700

900

1100

1300

1500

T [°C] Fig. 7. Pressure–temperature diagram showing the location of the carbonation/ decarbonation reactions (1) enstatite + dolomite = forsterite + diopside + CO2 (Wyllie et al., 1983) and (2) enstatite + magnesite = forsterite + CO2 (Newton and Sharp, 1975). In both cases, CO2 is present on the high-temperature, low-pressure side of the reaction boundary. (3) is the exchange reaction diopside + magnesite = enstatite + dolomite (Brey et al., 1983) that stabilizes magnesite on the high-pressure side of the reaction. The 35, 40, and 45 mW/m2 geotherms for lithospheric mantle and the 1300 °C mantle adiabat are from Hasterok and Chapman (2011). Gr–Dia is the graphite–diamond transition from Day (2012). Magnesite would be the stable carbonate in peridotitic lithospheric mantle along most geotherms.

208

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

diamond stability with respect to oxygen fugacity is defined by the presence of carbonates. Primary inclusions of carbonate (versus secondary precipitates from original fluid or melt inclusions) are rare but have been observed in diamonds hosted by both peridotite (magnesite: Harris et al., 2004; Phillips et al., 2004; Wang et al., 1996) and eclogite (calcite and dolomite: Meyer and McCallum, 1986; Sobolev et al., 2009). On the reducing side, diamond stability in mantle peridotite is limited through metal saturation (Frost and McCammon, 2008; Wood, 1993). Ni-free to Ni-poor native iron inclusions have been observed in diamonds from Yakutia (coexisting with silicates of harzburgitic paragenesis; Sobolev et al., 1981), Sloan (Meyer and McCallum, 1986) and Mwadui (rimmed by wüstite; Stachel et al., 1998). Oxygen fugacities distinctly below the iron–wüstite buffer (Ulmer et al., 1998) are documented by the occurrence of moissanite as inclusion in diamonds from Monastery (Moore and Gurney, 1989), Argyle (Jaques et al., 1989), and George Creek (Chinn, 1995). In the latter two cases, moissanite coexisted with silicate inclusions of the eclogitic suite. Attempts to constrain the fO2 conditions associated with common inclusions in diamond have so far been limited to the application of the olivine–orthopyroxene–spinel oxybarometer to diamonds of peridotitic paragenesis from several kimberlites on the Kaapvaal Craton (Daniels and Gurney, 1991) and the Limpopo Belt (Kopylova et al., 1997). Considering all uncertainties, including significant differences between the various existing calibrations (Ballhaus et al., 1991; O'Neill and Wall, 1987), the results provide no additional constraints beyond the limitation of diamond stability to fO2 conditions bounded by carbonate and Fe–metal forming reactions.

fO2 (FMQ) = log fO2 (sample) − log fO2 (FMQ), changing the T-log fO2 diagram to one expressing T versus Δlog fO2 (FMQ) one (Supplementary Fig. 2b). These buffer reactions also depend on pressure to varying extents (Supplementary Fig. 3), so using values of Δlog fO2 (FMQ) to compare samples from different pressures should be done with the awareness that such values do not normalize out the pressure dependencies as effectively as the temperature dependencies. Turning to the oxidation state of the mantle in the diamond stability field, the most direct evidence comes from garnet-bearing peridotites. All four minerals in a garnet peridotite (olivine, orthopyroxene, clinopyroxene, and garnet) contain Fe2 +; the pyroxenes and garnet can accommodate Fe3+ to variable extents as well. Various reactions have been proposed as “oxybarometers” for garnet peridotites; Luth et al. (1990) outlined five: 2þ

3þ

2þ

2þ

Fe3 Fe2 Si3 O12 ¼ 2 Fe2 SiO4 þ Fe

SiO3 þ 0:5O2

3þ

2þ

Mg3 Fe2 Si3 O12 ¼ MgSiO3 þ Mg2 SiO4 þ Fe2 SiO4 þ 0:5O2 3þ

2þ

2Ca3 Fe2 Si3 O12 þ 6MgSiO3 þ 4Fe ¼ 6CaMgSi2 O6 þ 3þ

2þ 4 Fe2 SiO4

ð1Þ ð2Þ

SiO3

þ O2

ð3Þ

2þ

2 Ca3 Fe2 Si3 O12 þ 2 Fe3 Al2 Si3 O12 ¼ 2 Ca3 Al2 Si3 O12 þ 4 Fe2 SiO4 þ 2 FeSiO3 þ O2

ð4Þ

3þ

2Ca3 Fe2 Si3 O12 þ 2Mg3 Al2 Si3 O12 þ 4FeSiO3 ¼ 2Ca3 Al2 Si3 O12 þ 4Fe2 SiO4 þ 6MgSiO3 þ O2 :

ð5Þ

5.3. Oxidation state of Earth's upper mantle Discussion of the oxidation state of natural samples is usually couched in terms of their oxygen fugacity, which stems from the use of oxygen buffers in experimental petrology pioneered by Eugster (1957). These buffers are assemblages that control the chemical potential of oxygen by reactions such as: :: 2x Fe þ O2 ¼ 2 Fex O ðiron–wustite; IWÞ 2Ni þ O2 ¼ 2NiO ðnickel–nickel oxide; NNOÞ : 3 Fe2 SiO4 þ O2 ¼ 2Fe3 O4 þ 3SiO2 ðfayalite–magnetite–quartz; FMQ Þ 4 Fe3 O4 þ O2 ¼ 6 Fe2 O3 ðmagnetite−hematite; MHÞ For each of these reactions, an equilibrium expression of the form KMH ¼

a6Fe2 O3 4 a Fe3 O4 f O2

may be written. Combining this equation with that relating the equilibrium constant to the standard free energy for this reaction o

RT lnK MH ¼ −Δr G

produces an expression for fO2 (by convention expressed as log10 fO2) of the form: log f O2 ðMH Þ ¼

Δr Go þ 6 log a Fe2 O3 −4 log a Fe3 O4 ; 2:303RT

which in the case of pure magnetite and hematite reduces to: o

log f O2 ðMH Þ ¼

Δr G : 2:303RT

The calculated locations of these reactions in T-log fO2 space at 4 GPa are shown in Supplementary Fig. 2a. Because these reactions are nearly parallel in this diagram, much of the relative temperature dependence is removed by expressing the values of fO2 relative to a reference reaction. The most common reference in use is the FMQ reaction, such that Δlog

The thermochemical data available to Luth et al. (1990) restricted their analysis to the reactions involving the Ca 3 Fe 2 Si 3 O12 endmember. Following the determination of the free energy of formation of the Fe3Fe2Si3O12 end-member by Woodland and O'Neill (1993), Gudmundsson and Wood (1995) experimentally constrained reactions (1) and (5) at 1300 °C and 2.5–3.5 GPa. They found (1) to be more precise, and until recently all subsequent studies have used their calibration for Δlog fO2 (FMQ) based on reaction (1), with the correction to a typographical error as noted by Woodland and Peltonen (1999). Oxybarometry of garnet peridotites was revisited recently by Stagno et al. (2013), who did experiments at 3, 6, and 7 GPa at temperatures between 1300 and 1600 °C. They proposed a new calibration based on reaction (5), which reproduced their data to better precision than the previous calibration of Gudmundsson and Wood (1995). There have been a number of studies on the oxidation state of cratonic garnet peridotites from various localities (Canil and O'Neill, 1996; Creighton et al., 2009, 2010; Goncharov and Ionov, 2012; Goncharov et al., 2012; Lazarov et al., 2009; Luth et al., 1990; McCammon and Kopylova, 2004; Woodland, 2009; Woodland and Koch, 2003; Woodland and Peltonen, 1999; Yaxley et al., 2012). These studies have produced a dataset of 225 samples from cratons worldwide, with almost half of the data coming from the Kaapvaal Craton in southern Africa. Of the samples studied, very few have been documented to contain diamond or graphite. The one graphite-bearing sample studied by Canil and O'Neill (1996) is a low-pressure sample (2.1 GPa, 710 °C). Two diamond-bearing samples from Finsch have been studied; one (865) by both Canil and O'Neill (1996) and Lazarov et al. (2009). The other (F556) is from the Canil and O'Neill (1996) study. There have also been recent studies of ultrahigh-pressure orogenic peridotites interpreted to be samples of mantle wedges (Malaspina et al., 2009, 2010) that yield fascinating insights into the oxidation state in these environments. However, given the focus of this review on diamond formation in cratonic lithospheric mantle, space prohibits a full exploration of variations in oxidation state with tectonic environment. To place the available data on cratonic samples in context in terms of their depths of origin and the geothermal conditions recorded by their mineral compositions, they are shown on a P–T diagram (Fig. 8) along

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

with conductive reference geotherms and the mantle adiabat from Hasterok and Chapman (2011). The samples from the Slave Craton mostly plot between the 35 and 40 mW/m2 geotherms, with a few recording hotter conditions. The Siberia samples scatter more widely, whereas those from the Kaapvaal record conditions near the 40 mW/m2 geotherm, with the deepest samples actually plotting along the extrapolation of that geotherm beyond the mantle adiabat of Hasterok and Chapman (2011). To show the oxidation state of these samples, we can plot them on a depth/pressure versus Δlog fO2 (FMQ) diagram, on which the samples are placed at the pressure derived from their mineral thermobarometry (Fig. 9). Temperature would also be increasing with depth (cf. Fig. 8). For comparative purposes, we also plot the iron–wüstite (IW), EMOG/D, and graphite–diamond reactions on these diagrams. These reactions are all calculated for the P–T conditions along a 40 mW/m2 model geotherm (Hasterok and Chapman, 2011). The IW curve approximates the fO2 at which metal becomes stable; as shown by Frost and McCammon (2008), for example, the fO2 at which Ni-rich metal becomes stable is very close to the location of the end-member IW reaction (after Ballhaus et al., 1991). The EMOG/D reactions represent the upper fO2 bound for the stability of graphite or diamond in olivine-bearing mantle relative to crystalline carbonate. The “SF10” curves shown in Fig. 9 represent the upper bound when the carbonate is present as a carbonatitic melt (Stagno and Frost, 2010). Comparing the oxidation states of these samples calculated using the calibration of Stagno et al. (2013; “S13”) with those from the calibration of Gudmundsson and Wood (1995; “GW95”) (Fig. 9a and b), there is reasonable agreement at low pressure (b 5 GPa), but increasing differences at higher pressure (Fig. 10), with S13 yielding more oxidized values. S13 also yields a wider range of values at a given pressure than GW95 (Fig. 9a and b). For both oxybarometers, calculated Δlog fO2 (FMQ) for a given Fe3 +/ΣFe in garnet becomes more reduced with increasing depth (Supplementary Fig. 4). Separating the global dataset allows us to understand better what is happening on a regional basis. We will consider the two cratons with the most data: the Kaapvaal and Slave Cratons. The results are illustrated in Fig. 9c and d, which show the trend to more reduced values with

45

3

adiabat

35

P [GPa]

4

140

5 Gr Dia

Depth (km)

40

100

180 6 Kaapvaal Siberia

7 700

Slave Tanzania

220

Fennoscan

900

1100

1300

1500

T [°C] Fig. 8. Global dataset of garnet peridotites with measured Fe3+/ΣFe (garnet). Data at less than 2.5 GPa (mostly Vitim, southern Siberia data from Goncharov and Ionov, 2012) are not plotted. The 35, 40, and 45 mW/m2 model conductive geotherms for lithospheric mantle and the 1300 °C mantle adiabat are from Hasterok and Chapman (2011). Temperatures and pressures were taken from the original papers except for those of Luth et al. (1990) and Canil and O'Neill (1996), which were re-calculated as per Creighton et al. (2009), i.e., using PBKN (Brey and Köhler, 1990) in combination with TO'Neill (O'Neill, 1980; O'Neill and Wood, 1979). Blue circles are diamond-bearing peridotites (Canil and O’Neill, 1996; Lazarov et al., 2009).

209

increasing depth. Some localities show trends with less scatter than others (e.g., the Early Cretaceous Group II kimberlite at Finsch versus the Late Cretaceous Group I kimberlites of Kimberley). For the Kaapvaal Craton, there is also a correlation with texture; the “sheared” peridotites (open symbols) tend to be more oxidized than the “coarse” ones (filled symbols). A first order observation is that there is considerable variability in the relative oxidation states recorded even by samples from the same locality. The Kimberley samples from the Kaapvaal Craton and both the Diavik and Jericho suites from the Slave Craton are excellent examples of this variability, in that these samples present ranges in Δlog fO2 (FMQ) on the same order as that of the global dataset. This variability begs the question as to whether these values reflect the differences in oxidation state(s) at the time of diamond formation, or processes that post-date diamond formation. If the latter is the case, then the significance of the current values is as a record of whether the lithospheric mantle was “diamond-friendly” or “diamond-hostile”, at least at the time of sampling by the kimberlite host magma. What causes the variability observed in either the global dataset or in individual suites? To address this question, we first need to understand how a peridotite of a given fixed composition, including its FeO and Fe2O3 contents, would change in relative oxidation state with pressure and temperature along the geotherm. Then, to explain the variation at a given depth, we need to determine how robust the relative oxidation state values are to modification by interactions with fluids or melts. Wood et al. (1990) showed that in an isochemical mantle, the relative oxidation state of garnet peridotite should decrease with depth along the geotherm because of the change in volume for reactions that stabilize Fe3+-bearing garnet. This result has been confirmed by subsequent modeling by a number of workers (Ballhaus and Frost, 1994; Creighton et al., 2009; Frost and McCammon, 2008; Luth and Stachel, 2014) (Supplementary Fig. 5). Superimposing the results of the “Luth/Stachel” calculations for pyrolite with different bulk Fe3+/ΣFe on the global xenolith dataset (Fig. 11), it is noted that a bulk Fe3+/ΣFe greater than 0.05 would stabilize carbonate rather than diamond in a pyrolite mantle composition, and that most samples lie in a field bounded on the low-fO2 side by the 0.02 curve. It is also clear in Fig. 11 that the variability in Δlog fO2 (FMQ) at a given depth can be explained by changing the bulk Fe3+/ΣFe of a peridotite — by adding or removing oxygen, in other words, without requiring any other compositional change. Confounding this straightforward interpretation of the dataset, however, is the fact that many of the natural samples are more depleted than pyrolite (and variably so), and thus the effects of depletion – which would be reflected in both modal mineralogy and mineral composition – need to be evaluated. How will the more depleted nature of cratonic peridotite affect the sensitivity of the calculated Δlog fO2 (FMQ) to addition or removal of oxygen by interactions with fluids or melts? It is reasonable to suppose that decreasing the modal abundance of Fe3 +-bearing minerals – clinopyroxene and garnet in particular – should decrease the amount of oxygen required to shift the fO2 of the rock. To quantify this, Luth and Stachel (2014) calculated the Δlog fO2 (FMQ) of a peridotite of an arbitrary bulk composition as a function of bulk Fe3 +/ΣFe, P, and T. Their model, conceptually based on those of Frost and McCammon (2008) and Stagno et al. (2013), allowed exploration of the effects of changing bulk composition on the relationship between bulk Fe3+/ΣFe and the calculated Δlog fO2 (FMQ). Using the pyrolite mantle composition of McDonough and Sun (1995) as a starting point, they varied the modal mineralogy (and hence the bulk composition) and found – as anticipated – decreasing the amount of clinopyroxene and garnet decreased the change in Fe3 +/ΣFe required to move the fO 2 of a sample from IW to EMOD. For example, at ~200 km depth (approximating the base of the lithosphere), the Fe3+/ΣFe of pyrolite increases from 0.017 at IW to 0.065 at EMOD (i.e., 382%). For comparison, the Fe3+/ΣFe of a peridotite with 5% garnet and 5% clinopyroxene would change from 0.009 to 0.026 (289%)

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

a

b

IW

3

EMOG

210

IW

100 120 140

Gr Dia

6

GW95 180 Calibration

IW

EMOD

Kaapvaal Siberia Slave Tanzania Fennoscan

d

EMOG

c 3

S13 Calibration

EMOG

7

SF10

160

SF10

5

IW

Depth [km]

Gr Dia

EMOD

P [GPa]

4

200 220

100 120

Gr Dia

5

160 Kimberley Lesotho Frank Smith Monastery Cullinan (Premier) Gibeon Finsch (Group II)

180

7 -4

-2

0

Δlog fO2 (FMQ)

EMOD

EMOD

6

-6

140

Depth [km]

SF10

SF10

Gr Dia

EMOD

P [GPa]

4

200 Diavik Gahcho Kue

220

Jericho

-6

-4

-2

0

Δlog fO2 (FMQ)

Fig. 9. Top figures: Comparison of the oxidation states of 225 garnet peridotites calculated with the Stagno et al. (2013) calibration (a) and with the Gudmundsson and Wood (1995) calibration (b). Data for samples with pressures b2.5 GPa are omitted. The curves shown are calculated at the pressure–temperature conditions along a 40 mW/m2 geotherm (Hasterok and Chapman, 2011). Blue circles in (a) are diamond-bearing peridotites (see text). Bottom figures: Comparison of the oxidation states of garnet peridotites from the Kaapvaal (c) and Slave (d) cratons, both calculated with the Stagno et al. (2013) oxybarometer calibration. In (c), the solid symbols are “coarse” textured samples, open symbols are “sheared” samples. Darker blue symbols are diamond-bearing samples from Finsch. Abbreviations: IW is the iron–wüstite buffer reaction, Gr–Dia is the graphite–diamond reaction; EMOG/ EMOD is the enstatite + magnesite = olivine + graphite or diamond reaction; SF10 is the calculated location of the analogous reaction involving carbonate melt rather than crystalline carbonate. IW from Ballhaus et al. (1991). Gr-Dia, EMOG, EMOD reactions calculated from the thermodynamic dataset of Holland and Powell (2011). SF10 curve from Stagno and Frost (2010).

over the same range in fO2, and that of a peridotite with 2% garnet and 2% clinopyroxene would increase from 0.007 to 0.016 (228%). They then modeled some real samples for which modal mineralogy, mineral compositions, and Fe3+/ΣFe for garnet are available. Their results are shown for three natural samples and for the model pyrolite composition in Fig. 12. All three of the natural samples contain less garnet and clinopyroxene than pyrolite, which shifts their curves on the Δlog fO2 (FMQ)–Fe3+/ΣFe diagram to the left. The curves for the natural samples are also steeper than pyrolite, so that they require less increase in Fe3+/ΣFe (and therefore less O2 added to the sample) to change their fO2 from IW to EMOD. Applying their model to a larger set of samples, Luth and Stachel (2014) addressed a simple question: how much O2 is required to shift the fO2 of cratonic peridotites from IW to EMOD? Luth and Stachel (2014) demonstrated that only ~ 400 ppm O2 is required to shift the oxidation state of primitive mantle pyrolite from IW to EMOD. For depleted peridotites, less than 200 ppm is needed — and for four samples from Finsch from the study of Lazarov et al. (2009), less than 50 ppm O2 is needed (Fig. 13). The study of Luth and Stachel (2014) quantified the extreme sensitivity of the fO2 of depleted lithospheric peridotites to changes in bulk

Fe3+/ΣFe — and therefore to changes in oxygen content that could be produced by interaction with a melt or fluid, either during the formation of diamond or subsequently. In order to evaluate the effectiveness of such an interaction, however, we need to turn our attention to what we know about possible fluids and melts.

5.4. Fluids in the upper mantle Fluids in Earth's upper mantle can be modeled in the CHO system, although evidence from natural samples suggests that chlorine, sulfur, and nitrogen need to be included in a comprehensive model of mantle fluids. In addition, the solubility of silicates in fluids, especially those that are water-rich, increases with increasing pressure to the extent that a second critical endpoint on the (hydrous) fluid-saturated solidus of model eclogite has been found at ~6 GPa (Kessel et al., 2005; Fig. 3). The second critical endpoint is where there can no longer be any distinction between a solute-rich fluid and a fluid-rich melt — so the solidus would terminate at this point. The existence of a second critical endpoint on the fluid-saturated solidus of either harzburgite or lherzolite

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

211

EMOD 2

-2

Δlog fO2 (FMQ)

4 5 Kaapvaal Siberia Slave Tanzania Fennoscan

7 -1.0

-0.5

MS95 pyrolite PHN 5267 Finsch F-15 Finsch 695

-4

0.5

0.0

1.0

IW 0.00

1.5

0.02

Δlog fO2 (FMQ)S13 - Δlog fO2 (FMQ)GW95

3

4 0.0 5 0.0 0.06 7

0.0

0.0

0.0

2

has proven to be more difficult to constrain (indeed, as has the location of solidi in either lithology — see discussion in Luth, 2014). As a starting point for our discussion, we will use the CHO system, given that there are quantitative models for fluids in this system that allow calculation of composition and speciation under both C-saturated and C-undersaturated conditions (e.g., Huizenga et al., 2012). The most recent thermodynamic model for CHO fluids is that of Zhang and Duan (2009, 2010). The compositions of these fluids are usually depicted in the CHO ternary (Fig. 14), and the speciation in the fluid can be plotted as a function of fO2 (Supplementary Fig. 6). In both cases, pressure and temperature are fixed; these two example figures show the results for conditions at the base of the lithosphere along a 40 mW/m2 geotherm. From these diagrams, we see that fluids coexisting with diamond would be CH4-rich at reduced conditions (IW), nearly pure H2O at the water maximum (labeled “m” on Supplementary Fig. 6), and water-rich H2O + CO2 at the maximum fO2 at which diamond would be stable in olivine-bearing mantle (EMOD). In the absence of olivine, diamond could coexist with more CO2-rich fluids (dominantly H2O–CO2 mixtures) at higher fO2. The carbon–saturation curve (red curve in Fig. 14) shows the change in fluid composition from CH4-rich to CO2-rich with increasing fO2. For

IW

100

5

160

6

Depth [km]

140

Gr Dia

180

EMOD

P [GPa]

120 4

200 -6

-4

-2

0.06

0.08

0

Δlog fO2 (FMQ) Fig. 11. Calculated oxygen fugacities for a primitive mantle composition (McDonough and Sun, 1995) as a function of bulk Fe3+/ΣFe along a 40 mW/m2 geotherm (Hasterok and Chapman, 2011). Numbers at top of curves are the bulk Fe3+/ΣFe of the mantle composition. Oxygen fugacities are calculated with the Stagno et al. (2013) oxybarometer. For reference, the global xenolith database and reaction boundaries from Fig. 9 are shown as well.

Fig. 12. Δlog fO2 (FMQ) versus Fe3+/ΣFe trends calculated for samples from Finsch and Cullinan (Premier) compared to the trend for primitive mantle pyrolite. Data sources: MS95 pyrolite composition of McDonough and Sun (1995), PHN 5267 (Cullinan) data from Boyd and Mertzman (1987) and Canil and O'Neill (1996), and Finsch data from Lazarov et al. (2009).

example, a fluid at IW contains ~8 mol% oxygen, whereas one at EMOD contains ~34 mol% O. Because H and C have much lower molar weights than O, the change is more dramatic in mass units; if the fluids are expressed in terms of wt. % (of C, H2, and O2), the composition of the IW fluid is 33 wt.% O2 versus 87 wt.% O2 in the EMOD fluid. The shape of the carbon–saturation curve also shows how the carbon content of the fluid changes, and hence how the saturation in diamond changes, with oxidation state. With increasing fO2, the carbon content decreases from 20 mol% C in pure CH4 to its minimum value near the O–H sideline at the composition of H2O. Further oxidation increases the carbon content to its maximum value of 33 mol% at pure CO2. The implications for diamond formation are straightforward: oxidation of a reduced fluid with a starting fO2 at IW will precipitate diamond until conditions of the water maximum are reached. Conceptually, this precipitation can be represented by the reaction CH4 + O2 → C + 2 H2O

O2 required for IW to EMOD shift (ppm)

Fig. 10. Differences between the values of Δlog fO2 (FMQ) calculated with the Stagno et al. (2013) and the Gudmundsson and Wood (1995) oxybarometers plotted as a function of pressure for the global dataset.

3

0.04

Fe3+/ΣFe (bulk)

200

60

150

40

100

20

50

300

200

100

PHN 5267 Finsch samples

0 0

2

4

6

8

10

12

0 14

carats diamond per metric ton

6

-3

ppm C

P [GPa]

3

0

Modal garnet (wt.%) Fig. 13. Amount of O2 required to shift a sample from IW to EMOD, plotted against the modal garnet in the sample. The right hand axis indicates the amount of elemental carbon (diamond or graphite in ppm and carats per ton) that is precipitated during reduction of CO2 (C:O = 1:2) in the fluid. Finsch samples are cpx-bearing peridotites from Lazarov et al. (2009) (their samples 695, F-1, F-3, F-5, F-6, F-8, F-11, F-12, F-14, F-15, F-16). Calculated at stated P–T of equilibration for each sample. For comparison to the curves in Fig. 12, samples 695 and F-15 are the most garnet-poor and most garnet-rich points, respectively, of the “Finsch sample” on this diagram. MS95 pyrolite requires ~400 ppm O2 to shift from IW to EMOD (Luth and Stachel, 2014), causing ~150 ppm C (equivalent to 750 carats diamond per metric ton) to precipitate.

212

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

C

Diamond + Fluid CH4

-5 IW -4 -3

H

-0.31 CO2

-1 -2

EMOD

O

H2O

Fig. 14. Molar ternary diagram showing the compositions of fluids (red curves) in equilibrium with diamond at the base of the lithosphere along a 40 mW/m2 geotherm (192 km, 6.1 GPa, 1356 °C; see Huizenga et al. (2012) for the composition of fluids in equilibrium with diamond at 5.0 GPa, 1227 °C). The numbers along the curve are values of Δlog fO2 (FMQ). For reference, the location of the Δlog fO2 (FMQ) values for IW and EMOD are shown as well. Calculated with GFluid (Zhang and Duan, 2010).

(Huizenga et al., 2012; Taylor and Green, 1989). With further oxidation, the fluid will consume diamond rather than precipitate it. Conversely, an oxidized fluid coexisting with diamond would have a starting fO2 no higher than EMOD in a peridotitic mantle, and could only precipitate diamond upon reduction via CO2 → C + O2 until reaching the water maximum. Further reduction of such a fluid would resorb diamond as the fluid became more C-rich. The amounts of oxygen involved are not trivial; 119 g O2 have to be added to 100 g starting IW fluid, and 45 g diamond has to precipitate for the fluid to evolve to the water maximum composition at the P and T conditions of Fig. 14. The amounts of oxygen and carbon removed from the fluid in reducing an EMOD fluid to the water maximum composition are smaller (5.7 g O2 and 1.0 g C per 100 g original EMOD fluid), because of the compositional proximity of the two fluids.

mole % C in C-saturated fluid

10

a

The location of the carbon saturation curve in the CHO ternary moves with changing pressure and temperature (as shown earlier by Huizenga et al., 2012). It is difficult to show these changes at the scale of the whole ternary diagram; a more useful approach is to zoom in by plotting the concentration of carbon in the fluid directly. The concentration of carbon in a carbon-saturated fluid is a strong function of bulk composition, which can be expressed as molar O/(O + H), equivalent to the O–H base of the CHO ternary (Fig. 14). Changing temperature has the largest influence on composition near the water maximum (Fig. 15a). At the water maximum, the solubility of carbon (as CO2 and CH4) increases most strongly with increasing temperature, almost doubling in amount with a 200 °C increase in temperature. At higher and lower O/(O + H) than those shown in Fig. 15a, the changes in C content of the fluid are negligible. The effect of pressure is greatest at the water maximum as well, with the content of C decreasing with increasing pressure. In reducing fluids, however, the C content increases with increasing pressure. Because both pressure and temperature increase with depth along a geotherm, the opposing pressure and temperature effects on the solubility of C in fluids near the water maximum means that we cannot predict a priori how the C-saturation surface near the water maximum will change. Calculating the compositions of the fluids with changing O/(O + H) at two depths along a 40 mW/m2 geotherm, however, shows that the carbon content decreases with decreasing depth in both the water maximum fluid and reduced fluids (Fig. 15b). The effect drops off rapidly as the fluids become more oxidized than the water maximum, such that the two curves are coincident by EMOD-like conditions (O/(O + H) ~ 0.35). The decrease in maximum C content with ascent in reduced fluids reflects a pressure effect. Another way to view the change in composition of the fluid along the geotherm is to look specifically at the water-maximum fluid, and to see how the solubility of carbon species in that fluid changes. This fluid has O/(O + H) = 0.333, but the C content in the fluid increases dramatically, from 0.32 mol% at 3 GPa to 0.91 mol% at 6 GPa. This change in the C content of the fluid essentially results from the oxygenconservative reaction CO2 + CH4 → C + 2H2O. As well as composition, the speciation in the fluid will change with depth along the geotherm (Fig. 16). Comparing the curves for X(H2O) and X(CH4) in the fluid with the global xenolith dataset, most of the samples plot to the reduced side of the water maximum, and would coexist with water-rich H2O–CH4 fluids.

20

b

8 15 6 10 4

2

0 0.20

5

1556°C 1456°C 1356°C

0.25

192 km 120 km

0.30

0.35

O/(O + H) (molar)

0.40

0.1

0.2

0.3

0.4

O/(O + H) (molar)

Fig. 15. (a) Change in the carbon content of diamond-saturated fluid with temperature at 192 km depth as a function of the O/(O + H) of the fluid. The black curve shows the composition of the fluid at the temperature appropriate for a 40 mW/m2 geotherm at that depth. The other two curves illustrate the increasing solubility of C with increasing temperature. (b) Change in the carbon content of C-saturated fluid with depth along a 40 mW/m2 geotherm as a function of the O/(O + H) of the fluid. For reference, the O/(O + H) for a C-saturated fluid with fO2 equal to IW is 0.10 at 192 km and 0.03 at 120 km depth. The O/(O + H) for a C-saturated fluid with fO2 equal to EMOD is 0.35 at 192 km and 0.34 at 120 km depth. Calculated with GFluid (Zhang and Duan, 2010).

0 .0 5

0 .0 1

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

IW

3

100 120

5

160 180

-4

220 0.2

0.8

0.8

max

0.2

0.4

0.6

0.2

7 -6

200

EMOD

0.4

6

Depth [km]

140

Gr Dia

0.6

P [GPa]

0.8

4

-2

0

Δlog fO2 (FMQ) Fig. 16. Pressure-Δlog fO2 (FMQ) diagram calculated along a 40 mW/m2 geotherm illustrating how the speciation of CHO fluids changes along the geotherm. Solid gray lines show the mole fraction of H2O, long dashed gray lines show the mole fraction of CH4, and short dashed black lines are the mole fraction of CO2 for 1 and 5 mol% CO2; at more oxidized conditions, the fluid lacks CH4 and is essentially H2O + CO2, so the X(CO2) = 0.2 curve coincides with the curve for X(H2O) = 0.8. Calculated with GFluid (Zhang and Duan, 2010). For reference, the global xenolith database and reaction boundaries from Fig. 9 are shown as well.

5.5. Ascent along lithospheric PT–fO2 path By comparing the change in fO2 of an isochemical mantle composition and the fluid speciation along the geotherm (Figs. 11 and 16), it is apparent that a fluid in equilibrium with an isochemical mantle would become more oxidized at shallower depths. This effect can be illustrated quantitatively for two values of bulk Fe3+/ΣFe of the primitive mantle (0.02 and 0.03). In both cases, the fluids coexisting with the mantle become more oxidized and lower in carbon content with decreasing depth as they shift to higher O/(O + H) and move along the Csaturation surface towards the water maximum (Fig. 17).

mole % C in C-saturated fluid

14 12

200 150

10 8 6 200

4 100

2 0 0.10

Fe3+/ΣFe = 0.02 Fe3+/ΣFe = 0.03

0.15

0.20

0.25

150

213

Because the fluid has to change composition to maintain equilibrium with a constant-composition mantle as it ascends along the geotherm, a source of oxygen (or sink for hydrogen) would be required, in addition to the precipitation of carbon that takes place. If the surrounding mantle were to act as this source or sink, it follows that the “isochemical mantle” assumption would no longer hold, and the mantle would have to become more reduced with decreasing depth. The ability of the mantle to do so will be addressed presently, but we will first consider the case of a fluid cooling isobarically or ascending along the geotherm without exchanging O (or H) with the surrounding mantle. 5.6. “Isochemical” ascent or isobaric cooling of fluids What happens to a C-saturated CHO fluid that ascends along the geotherm “isochemically” — that is, without exchanging either oxygen or hydrogen with the surrounding mantle? Given the expansion of the C-saturation curve with ascent along a geotherm (Fig. 15b), an ascent in which the O/(O + H) of the fluid remains constant would oversaturate a C-saturated fluid and cause precipitation of diamond. This effect can be seen from the relative locations of the two C-saturation surfaces in Fig. 15b — a fluid composition that is C-saturated at 192 km depth would contain more C than would be stable in the fluid with the same O/(O + H) at 120 km depth (at least for O/(O + H) ≲ 0.35; that is, for fluids with fO2 of EMOD or lower). Along with precipitating diamond, CHO fluids ascending along the geotherm at constant O/(O + H) will shift to more oxidized values relative to FMQ (Fig. 18a). Fig. 18a also shows the amount of carbon that would precipitate from four distinct fluids as a percentage of the amount of carbon in the original fluid. Arguably, expressing the carbon precipitation in this fashion fails to account for the effect of the decreasing carbon content in the fluid with increasing O/(O + H). In terms of the mass of carbon precipitated, more reduced fluids precipitate more carbon (Fig. 18b), but at a lower percentage simply because they are more carbon-rich. The change in carbon content of a diamond saturated fluid with temperature only (i.e., isobaric cooling) suggests an alternative mode of diamond formation. Fig. 15a shows that the decrease in maximum carbon content with temperature is particularly prominent for fluids near the water maximum. This implies that even fluids that initially were not saturated in carbon may begin to precipitate diamond during cooling. A possible scenario for the infiltration of CHO fluids hotter than ambient mantle is the release of fluid from crystallizing magmas or mantle plumes. As shown in Section 3.4, diamond formation generally takes place below the wet solidus of harzburgite but above the wet solidi of lherzolite and eclogite. This effectively limits diamond precipitation from cooling water-rich CHO fluids to harzburgitic substrates as fluid dilution through a melt component would occur in lherzolitic and eclogitic lithologies (except along conductive model geotherms b40 mW/m2 at pressures b5 GPa; see Fig. 1). This may explain the observed close relationship between harzburgitic (“G10”) garnets and diamond (Gurney, 1984) employed as the principal tool of indicator mineral based diamond exploration. More reducing fluids may also penetrate lherzolitic and eclogitic lithologies due to the higher solidus temperature associated with such fluids, but as seen in Fig. 15a, the change in maximum carbon content of such fluids with temperature is negligible.

100

5.7. Diamond precipitation from ascending or isobarically cooling melts

0.30

0.35

O/(O + H) (molar) Fig. 17. Change in carbon content of a C-saturated fluid in equilibrium with primitive mantle compositions with constant bulk Fe3+/ΣFe of 0.02 and 0.03 (compare Fig. 11). Numbers by symbols are depth in km. In both cases, the fluids become more oxidized with decreasing depth and consequently liberate elemental carbon upon ascent. Calculated with GFluid (Zhang and Duan, 2010).

In the presence of methane-rich fluids, the high solidus temperature of peridotite and to a lesser degree eclogite largely precludes percolation of strongly reducing melts through the lithospheric mantle. Therefore, for melt associated diamond precipitation, likely to dominate in eclogitic and lherzolitic substrates (see Section 3.4), only carbonatebearing melts need to be considered. With decreasing carbonate content such melts become stable to fO2 conditions far below the EMOD

49%

3

140

4

Gr Dia

5

0

100

140

Gr Dia

5

160 180

180

EMOD

6 200

-2

0

Δlog fO2 (FMQ) 2.5

7 -6

b

220

max

-4

200

EMOD

max

160

6

g C precipitated per 100 g original fluid

0 .5

IW

120

P [GPa]

IW

-4

-2

0

Δlog fO2 (FMQ)

2.0

Fig. 19. Pressure-Δlog fO2 (FMQ) diagram calculated along a 40 mW/m2 geotherm. Light blue curves show the maximum-fO2 stability of carbonate-bearing melts according to the model of Stagno and Frost (2010). Numbers at the top of the curves indicate the molar fraction of CO2 (Xmelt(CO2)) in the melt (pure carbonate melt has Xmelt(CO2) = 0.5). Unlabeled curves (from left to right) are for Xmelt(CO2) = 0.005, 0.05, 0.2, 0.3, and 0.4.

1.5

1.0

0.5

0.0

0 .1

1

0 .0

120

Depth [km]

4

P [GPa]

2%

EMOG

6%

Depth [km]

4%

01

100

a

0

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

0 .0

214

0.1

0.2

0.3

0.4

O/(O + H) (molar) Fig. 18. (a) Pressure-Δlog fO2 (FMQ) diagram calculated along a 40 mW/m2 geotherm. Yellow-orange lines show the trajectory of four CHO fluid compositions ascending along the geotherm at constant O/(O + H). The numbers at the top of each arrow provide the percentage of carbon in the original fluid that precipitates during the ascent of each fluid from 192 to 120 km. The trajectory for the most oxidized fluid, as it is entirely above the fO2 of EMOD, could only occur in an olivine-free lithology. Curves calculated with GFluid (Zhang and Duan, 2010). For reference, the global xenolith database (gray symbols) and reaction boundaries from Fig. 9 are shown as well. (b) Amount of carbon precipitated from fluids during ascent along a 40 mW/m2 geotherm, as a function of O/(O + H). The amount of carbon is expressed as the mass of carbon precipitated from 100 g of the fluid that was present at 192 km. Calculated with GFluid (Zhang and Duan, 2010).

buffer (Fig. 19) and consequently, precipitation of diamond from the melt is possible. The precipitation of diamond directly from carbonatebearing melts would involve an internal redistribution of oxygen and thus have an oxidizing effect on the residual melt, which ultimately may stall further diamond precipitation. Percolating carbonate-bearing melts are in equilibrium with the surrounding peridotitic mantle (e.g., Hiraga and Kohlstedt, 2009; Watson et al., 1990) and, consequently, have the full peridotitic mineral assemblage as liquidus phases. In the absence of adequate thermodynamic data for carbonate-bearing melts we cannot evaluate if they would experience a similar decrease in maximum carbon content as CHO fluids during cooling or ascent along a geotherm. The possible appearance of diamond on the liquidus of cooling and evolving carbonate-bearing melts at fO2 conditions below the EMOD buffer has not been investigated experimentally and, therefore, is speculative. The absence of a clear association between diamond and lherzolitic garnets, however, suggests that this either is not the case or only rarely occurs.

The high abundance of eclogitic suite diamonds (1/3 of all inclusionbearing diamonds) derived from a volumetrically very minor component of lithospheric mantle (b1% to 5%; Dawson and Stephens, 1975; Schulze, 1989; McLean et al., 2007) implies that a fundamental difference exists between diamond precipitation in lherzolite and eclogite. The extension of diamond stability in eclogite, compared to peridotite, by at least one log unit to more oxidizing conditions (shift from EMOD to the DCDD [dolomite + coesite = diopside + diamond] equilibrium; Luth, 1993) potentially establishes such a key difference. Relative to pure diopside, the omphacitic nature of eclogitic clinopyroxene shifts the DCDD equilibrium even further to more oxidizing conditions (see Fig. 2 in Luth, 1993). This expansion of diamond stability to higher fO2 values may strongly enhance diamond precipitation directly from cooling or upward migrating carbonate-bearing melts. Alternatively, the redox buffering capacity of eclogite, which is much more Fe-rich than peridotite, far exceeds that of cratonic peridotites and, consequently, the possible extent of diamond precipitation during carbonate reduction is greatly enhanced. Olivinefree garnet pyroxenite layers with original diamond contents of up to 15 vol.% (preserved as graphite pseudomorphs) embedded in nondiamondiferous peridotite at Beni Bousera (Pearson et al., 1989) strongly support localized diamond precipitation from melts more oxidizing than EMOD in olivine-free lithologies. Extreme diamond contents, such as in the Beni Bousera websterites but also in diamondiferous eclogite xenoliths (e.g., up to 20% diamond in eclogite xenoliths from the Jericho kimberlite; Smart et al., 2009) cannot conceivably result from wall rock buffered redox reactions and, thus, support an “isochemical” mode of diamond precipitation from cooling or ascending carbonate-bearing melts. The redistribution of oxygen within the melt and consequent increase in fO 2 associated with this “isochemical” mode of diamond precipitation may be alleviated when continuous flow of percolating melt into eclogite/pyroxenite involving incremental diamond growth is considered. When such melts cross from ubiquitous peridotitic lithologies into eclogite, they will also be out of equilibrium with their new eclogitic wall rocks, which at fO 2 conditions below the DCDD equilibrium may enhance diamond precipitation. In any case, that diamond is so much more abundant in eclogite compared to lherzolite must relate to the more oxidizing conditions permissive of diamond precipitation in olivine-free lithologies (Luth, 1993).

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

5.8. The redox buffering capacity of cratonic peridotite As outlined above, redox reactions such as the oxidation of methane or the reduction of carbonate are thought to be essential aspects of diamond formation in Earth's upper mantle. Expressing these two mechanisms by the simple reactions CH4 + O2 = C + 2 H2O and CO23 − = C + O2 − + O2 demonstrates the fundamental role of O2 in mass-balancing this process as either a reactant or product. That is not to say that molecular oxygen per se is involved, but there has to be another coupled reaction to either produce or consume oxygen to allow redox precipitation of diamond to occur. A logical candidate for such a coupled reaction is the iron redox couple, in which reduced ferrous iron reacts with oxygen to produce oxidized ferric iron. This reaction, expressed as 4FeO + O2 = 2Fe2O3, could proceed in this form in iron-bearing melts, but in the assemblage of solid silicates that make up peridotitic mantle, charge-balance and crystal–chemical constraints require more complex reactions to provide or consume oxygen. These reactions are actually the same reactions previously discussed as the basis for oxybarometry in mantle peridotites (Eqs. (1)–(5) above). A key issue, therefore, is the ability of these reactions in the lithospheric mantle to act as the necessary source or sink of oxygen to allow redox precipitation of diamond to occur. In order to address this question, it is necessary to quantify both the amount of oxygen required for the redox precipitation of diamond, and the capacity of the mantle to act as a source or sink for oxygen. The former is reasonably straightforward, in that the oxygen required can be scaled to the amount of diamond precipitated, assuming that the redox reactions involve either methane or carbonate. In either case, the molar ratio of O2 to C is 1:1, which translates into a mass ratio of 2.67 g of O2 per gram of C. As seen from the previous discussion of the study of Luth and Stachel (2014), the capacity of mantle peridotite to act as an effective source or sink of O2 is limited by the sensitivity of its redox state to small changes in O2 content, at least relative to the much larger amount of O2 required to change the oxidation state of a CHO fluid. As shown above, b50 ppm O2 is needed to move the oxidation state of a strongly depleted peridotite from IW to EMOD (i.e., b5 mg O2 per 100 g rock) compared to the 119 g O2 that must be added to 100 g IW fluid to move it to the water maximum composition. It is also worth noting that the amount of diamond that would precipitate as a result of the shift of the oxidation state of even primitive mantle from IW to EMOD is relatively small (~150 ppm — see Fig. 13). For comparison, diamond-bearing peridotite xenoliths can have 5500 ppm diamond (Viljoen et al., 2004). Luth and Stachel (2014) modeled various scenarios of oxidized fluids interacting with reduced peridotite and vice versa; their conclusion was that b50 ppm fluid would be required to reset the oxidation state of depleted peridotite to that of the fluid. Thus, relative to a CHO fluid, the oxygen buffering capacity of cratonic (depleted) peridotites is very small, and the ability of such mantle to act as the sink or source for oxygen necessary for redox precipitation of diamond is quite low. To allow for the growth of common commercial sized diamonds through wall rock buffered redox reactions, diffusive transport of oxygen would need to occur over large distances (10s of cm); diffuse exchange over such length scales during metasomatic events is inconsistent with common evidence for mineral compositional and isotopic heterogeneity observed on the hand-sample scale in cratonic peridotite xenoliths (e.g., Burgess and Harte, 1999; Schmidberger et al., 2003 and references therein). From the degree of secondary LREE enrichment in cratonic garnet peridotites, Luth and Stachel (2014) estimated bulk addition of metasomatic fluid/melt approximately in the range of 0.1–5 wt.% — that is 20 to 1000 times the maximum amount of fluid required to reset the oxidation state of these rocks. This implies that redox profiles through subcratonic lithospheric mantle (Fig. 9) have no bearing on Archean mantle fO2 but merely are a reflection of the redox state of the last metasomatic fluid/melt passing through a section of lithosphere.

215

Consequently, diamond stable lithospheric mantle sampled, e.g., by Mesoproterozoic to Cretaceous kimberlites on the Kaapvaal Craton last interacted with fluids that invariably were more reducing than the EMOD buffer (Fig. 9c). In contrast, peridotite xenoliths in the Eocene A154 kimberlite at Diavik (Central Slave Craton) show metasomatic overprint by fluids/melts that are more variable in character, ranging from highly reducing to oxidizing (Fig. 9d). 6. Diamond forming processes based on co-variations in δ13C–N and on fluid inclusions The isotopic fractionation factor of carbon (ΔCdiamond-fluid) for precipitation of diamond from reduced (methane-bearing) and oxidized (carbonate- or CO2-bearing) fluids or melts has opposite signs and, on this basis, it should be relatively straightforward to derive the mode(s) of natural diamond formation based on population density plots of diamond carbon isotopic compositions (Deines, 1980) and through evaluation of co-variations between δ13C and the compatible trace element nitrogen (Stachel et al., 2009; Thomassot et al., 2007). Diamonds in a typical kimberlite deposit, however, originate over a depth range that in some cases encompasses the entire diamond stable lithospheric mantle and reflect multiple growth episodes (Harte et al., 1999a; Palot et al., 2013; Wiggers de Vries et al., 2013a, 2013b) rather than single diamond forming events. Variation in the bulk distribution coefficients for carbon and nitrogen (e.g., through co-precipitation of other phases) combined with large differences in the concentration of carbon (major element) and nitrogen (trace element) in the diamond forming fluids/melts introduce further complexity and may lead to an apparently decoupled behavior of δ13C and nitrogen concentration. On this background, discernible correlations between carbon isotopic composition and nitrogen content, indicative of a particular mode of formation, are unlikely to be observed based on bulk diamond analyses even on the level of individual deposits. The character of diamond forming fluids/melts, therefore, is difficult to constrain. Thomassot et al. (2007) presented evidence for diamond formation from a reduced, methane-bearing fluid, based on a suite of diamonds recovered from a single garnet–lherzolite xenolith from the Cullinan (Premier) mine. Nitrogen concentration and aggregation state characteristics suggested that the diamonds in this xenolith are cogenetic; a linear positive correlation between log N and δ13C for these diamonds, therefore, appeared to indicate operation of a single diamond forming process that could be modeled as precipitation from a methanebearing fluid or melt. Thomassot et al. (2007) also presented nitrogen isotopic data (δ15N) for this suite of diamonds and documented a positive linear correlation with δ13C, implying (based on a methane precipitation model) a positive nitrogen isotope fractionation factor. The subsequent determination of a negative sign of ΔNdiamond-fluid (Petts et al., 2014, submitted for publication) is incompatible with the model of Thomassot et al. (2007) and indicates that a more complex growth event (e.g., fluid mixing or co-precipitation of nitrogen free phases) occurred. Consequently, to date there is no compelling case of diamond precipitation from a reducing fluid documented in the literature that is built on the stable isotope and nitrogen characteristics of diamond itself. Indirect evidence for diamond precipitation from highly reducing fluids is, however, provided by reduced inclusions in diamond (e.g., native iron; see Section 5.2). Ion microprobes – for high precision analyses equipped with a large magnetic sector and a multi-collection system – enable the study of possible co-variations between δ13C and nitrogen content on the level of individual diamond growth zones. On this spatially highly resolved level, strongly correlated variations in δ13C and N-content are occasionally observed. To date, the best documented example is diamond 25-S1 recovered from a low-Mg eclogite xenolith from the Jericho kimberlite (Smart et al., 2011): in its outer core zone, characterized by homogenous blue CL response, this diamond shows an outward increase in δ13C (and δ15N; Petts et al., 2014; Petts et al., submitted for publication) coupled

216

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

with a decrease in nitrogen content. Such a relationship can only be modeled based on precipitation from an oxidized (carbonate- or CO2-bearing) fluid or melt (Smart et al., 2011). This conclusion is supported by the occurrence of (rare) carbonate inclusions in smoothsurfaced monocrystalline diamond (see Section 5.2). For non-gem fibrous diamonds, a clear link to carbonatitic melts (“high density fluids”) has been established through the study of finely dispersed, micrometer to nanometer-sized original melt inclusions (e.g., Klein-BenDavid et al., 2009; Kopylova et al., 2010; Navon et al., 1988). Similar carbonatitic melt inclusions have now also been documented within smoothsurfaced gem-type diamonds (Weiss et al., 2014), providing strong evidence that carbonate-bearing melts are an important diamond forming medium. With decreasing carbonate content, such melts may represent a spectrum of fO2 conditions from oxidizing (carbonatites) to reducing (possible for carbonated silicate melts; Fig. 19). In view of the currently very limited number of in situ studies documenting internal variations in stable isotope composition and nitrogen content during gem-diamond growth, the current absence of clear evidence for methane-driven diamond precipitation can, however, not be used to discount its operation in nature. Given the limited buffering capacity of cratonic peridotites (see Section 5.8), their overall reduced character (Fig. 9a) clearly implies that they last interacted with a reduced fluid or melt, documenting upward migration of chiefly reduced fluids through diamond stable lithospheric mantle. Whilst in the case of CHO fluids this implies methane as an important component, in melts the carbon speciation is less straightforward to predict and even at the reducing conditions of typical diamond stable mantle may still involve carbonate (see Fig. 19). Based on these results, natural diamond formation involving methane- and carbonate-bearing fluids/melts are both possible although actual diamond based evidence so far points to a strong predominance of carbonate-bearing diamond forming media. 7. Conclusions Inclusion based geothermobarometry indicates that peridotitic suite diamonds typically (1 sigma range about the average) originate from a 140–190 km depth at temperatures between 1040–1250 °C. Regression of the entire dataset results in a “diamond geotherm” that is slightly oblique to the model geotherms of Hasterok and Chapman (2011), increasing from 36 mW/m2 at the graphite–diamond transition to 39 mW/m2 at the intersection with the mantle adiabat (base of the lithosphere). The resulting “diamond window” (diamond stable region of the lithospheric mantle) extends from 110–205 km depth. In the absence of a reliable barometer for garnet–clinopyroxene assemblages, projection of thermometric data for diamonds hosted by eclogite on the peridotitic suite “diamond geotherm” indicates derivation mainly (1 sigma range about the average) from 155–200 km depth and 1060–1340 °C. Projection on the 40 mW/m2 Precambrian shield geotherm of Hasterok and Chapman (2011) results in a 1 sigma range of 135–190 km depth at temperatures of 1040–1330 °C. Compared to the peridotitic suite, typical eclogitic suite diamonds thus show a larger spread in temperatures extending to higher values. A detailed comparison of temperature estimates based on nontouching (diamond formation) and touching (final residence temperature) inclusion pairs of peridotitic and eclogitic paragenesis indicates that diamond formation in some cases is associated with elevated temperatures (by about 100–180 °C), whilst in other cases temperatures of formation and final residence are comparable or final residence may even occur at slightly elevated (40 °C) temperatures. For peridotitic suite diamonds from the Panda kimberlite (Central Slave Craton), comparison with nitrogen-in-diamond based mantle residence temperature estimates indicates that cooling after diamond formation was slow (100 s of millions to billion year time scale), suggesting lithospherescale thermal relaxation (e.g., after impact of a plume) rather than cooling following a local advective heating event.

Comparison of the geothermobarometric data with hydrous (-carbonated) solidi indicates that diamond formation in lherzolitic and eclogitic substrates typically occurs in the presence of a melt, whilst harzburgitic paragenesis diamonds generally form under sub-solidus conditions. The increase in solidus temperature associated with reduced fluids is permissive of melt-absent diamond formation in lherzolite and eclogite under such conditions. Consequently, for the dominant harzburgitic paragenesis, diamond formation can only relate to CHO fluids (without geothermobarometric constraints on fO2, i.e., within the range from H2O–carbonate to H2O–CH4–H2) whilst for diamonds hosted by lherzolite and eclogite both melts (carbonatites or silicate melts with dissolved OH− and CO23 −; e.g., Green and Falloon, 1998) and reducing fluids need to be considered. Diamond formation is generally believed to relate to redox reactions involving either oxidized (carbonate- or CO2-bearing) fluids/melts interacting with reduced wall rocks or reduced (methane-bearing) fluids/melts in contact with more oxidized wall rocks. Modeling the interaction of fluids with depleted cratonic peridotites indicates, however, that only less than 50 ppm fluid are required to completely reset the oxidation state of depleted peridotite to that of the fluid. This extremely limited buffering capacity of cratonic peridotites implies that redox reactions between fluid and wall rock are not an efficient way to precipitate diamond and extremely unlikely to produce large macro-diamonds. An additional implication of the limited buffering capacity of cratonic peridotites is that fO2 studies on garnet peridotite xenoliths generally only determine the redox state of the last metasomatic fluid/melt that interacted with these rocks. The observation that the bulk of peridotite xenoliths derived from the diamond stability field yields fO2 conditions more reducing than the EMOD buffer, consequently, implies that the last metasomatic event in most sections of deep cratonic lithosphere was reducing in character. We propose that a much more efficient mode of diamond precipitation is “isochemical” (not involving oxygen exchange with the wall rock) cooling of CHO fluids or their “isochemical” ascent along a geothermal gradient. In particular for fluids with compositions close to the water maximum both scenarios are associated with a distinct decrease in the solubility of carbon species (CH4, CO2 or CO2− 3 ), leading to diamond precipitation in peridotite from fluids more reducing than the EMOD buffer and in eclogite from fluids more reducing than the DCDD equilibrium. Incorporating the derived relationships between diamond forming conditions and hydrous solidi, this mode of diamond formation will be largely restricted to harzburgitic substrates and thus may explain the strong association between diamond and harzburgitic garnets observed worldwide. For melt-associated diamond precipitation, likely to dominate in eclogitic and lherzolitic substrates, the effect of cooling or ascent along a geotherm cannot be assessed in the absence of adequate thermodynamic data. The extreme overabundance of eclogitic suite diamonds (1/3 of all inclusion bearing diamonds belong to the eclogitic suite but eclogite constitutes only a very minor portion of lithospheric mantle) indicates, however, that the extension of diamond stability by at least one log unit to more oxidizing conditions (shift from EMOD to DCDD for olivine-free lithologies) plays a critical role. This expansion of diamond stability in fO2 space may strongly enhance diamond precipitation either during redox reactions with such far more Fe-rich bulk rock compositions or directly from “isochemically” cooling or upward migrating carbonate-bearing melts. The occurrence of originally highly diamondiferous garnet pyroxenite layers surrounded by carbon-free orogenic peridotites (Pearson et al., 1989) strongly supports preferential carbon capture in olivine-free pyroxenite or eclogite, specifically in settings more oxidized than EMOD. Extreme (15–20 vol.%) diamond contents in these garnet websterite layers and in some kimberlite born eclogite xenoliths, and the occurrence of very large (cm-sized) diamonds hosted by eclogite appear to be incompatible with formation during wall rock buffered redox reactions and again favor an “isochemical” mode of precipitation directly from a fluid or melt more reducing than DCDD.

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

Stable isotope and nitrogen content studies on diamond itself so far have not provided a conclusive answer as to what the redox state of the dominant diamond forming fluid/melt may be. Combining the currently available data from in situ (ion microprobe) studies with the recent finding of broadly carbonatitic high density fluids (melt inclusions) in gem diamonds, then diamond formation from carbonate-bearing fluids/ melts can be considered as being well established, but given the current paucity of conclusive datasets, diamond formation from reducing, methane-bearing fluids cannot be discounted. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.lithos.2015.01.028. Acknowledgments Marco Scambeluri is thanked for inviting this review paper and showing both persistence and patience over the two years of writing. The new concepts presented here matured during fruitful discussions in particular with Herman Grütter, Jeff Harris, Gerhard Brey and Tom Chacko. Herman Grütter and Fanus Viljoen are also thanked for their detailed input during formal reviews. George Read is thanked for showing TS a diamond vein cutting through an eclogite xenolith from Fort a la Corne, changing TS' thinking about diamond forming processes. TS and RL both acknowledge funding of their research programs through the Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery grants. TS receives additional funding through the Canada Research Chairs program. References Aulbach, S., Stachel, T., Creaser, R.A., Heaman, L.M., Shirey, S.B., Muehlenbachs, K., Eichenberg, D., Harris, J.W., 2009. Sulphide survival and diamond genesis during formation and evolution of Archaean subcontinental lithosphere: a comparison between the Slave and Kaapvaal cratons. Lithos 112 (Suppl. 2), 747–757. Ballhaus, C., Frost, B.R., 1994. The generation of oxidized CO2-bearing basaltic melts from reduced CH4-bearing upper mantle sources. Geochimica et Cosmochimica Acta 58, 4931–4940. Ballhaus, C., Berry, R.F., Green, D.H., 1991. High-pressure experimental calibration of the olivine–orthopyroxene–spinel oxygen geobarometer — implications for the oxidation state of the upper mantle. Contributions to Mineralogy and Petrology 107, 27–40. Bonney, T., 1899. The parent-rock of the diamond in South Africa. Proceedings of the Royal Society of London 65, 223–236. Boyd, F.R., Gurney, J.J., 1986. Diamonds and the African lithosphere. Science 232 (4749), 472–477. Boyd, F.R., Mertzman, S.A., 1987. Composition and structure of the Kaapvaal lithosphere, Southern Africa. In: Mysen, B.O. (Ed.), Magmatic Processes: Physicochemical Principles. Geochemical Society, University Park, PA, USA, pp. 13–24. Boyd, S.R., Mattey, D.P., Pillinger, C.T., Milledge, H.J., Mendelssohn, M., Seal, M., 1987. Multiple growth events during diamond genesis: an integrated study of carbon and nitrogen isotopes and nitrogen aggregation state in coated stones. Earth and Planetary Science Letters 86 (2–4), 341–353. Boyd, S.R., Pillinger, C.T., Milledge, H.J., Mendelssohn, M.J., Seal, M., 1992. C-isotopic and N-isotopic composition and the infrared-absorption spectra of coated diamonds — evidence for the regional uniformity of CO2–H2O rich fluids in lithospheric mantle. Earth and Planetary Science Letters 109 (3–4), 633–644. Brey, G.P., Köhler, T., 1990. Geothermobarometry in four-phase lherzolites II. New thermobarometers, and practical assessment of existing thermobarometers. Journal of Petrology 31, 1353–1378. Brey, G., Brice, W.R., Ellis, D.J., Green, D.H., Harris, K.L., Ryabchikov, I.D., 1983. Pyroxene–carbonate reactions in the upper mantle. Earth and Planetary Science Letters 62, 63–74. Bulanova, G.P., 1995. The formation of diamond. Journal of Geochemical Exploration 53 (1–3), 1–23. Burgess, S.R., Harte, B., 1999. Tracing lithospheric evolution trough the analysis of heterogeneous G9/G10 garnets in peridotite xenoliths, I: Major element chemistry. In: Gurney, J.J., Gurney, J.L., Pascoe, M.D., Richardson, S.H. (Eds.), The J.B. Dawson Volume, Proceedings of the VIIth International Kimberlite Conference. Red Roof Design, Cape Town, pp. 66–80. Burgess, R., Layzelle, E., Turner, G., Harris, J.W., 2002. Constraints on the age and halogen composition of mantle fluids in Siberian coated diamonds. Earth and Planetary Science Letters 197 (3–4), 193–203. Canil, D., O'Neill, H.S.C., 1996. Distribution of ferric iron in some upper-mantle assemblages. Journal of Petrology 37 (3), 609–635. Chinn, I., 1995. A study of Unusual diamonds From the George Creek K1 Kimberlite Dyke, Colorado. (PhD thesis), University of Cape Town, RSA, p. 95 (146 pp.). Connell, S.H., Sellschop, J.P.F., Butler, J.E., Maclear, R.D., Doyle, B.P., Machi, I.Z., 1998. A study of the mobility and trapping of minor hydrogen concentrations in diamond

217

in three dimensions using quantitative ERDA microscopy. Diamond and Related Materials 7 (11–12), 1714–1718. Creighton, S., Stachel, T., Matveev, S., Höfer, H., McCammon, C., Luth, R.W., 2009. Oxidation of the Kaapvaal lithospheric mantle driven by metasomatism. Contributions to Mineralogy and Petrology 157, 491–504. Creighton, S., Stachel, T., Eichenberg, D., Luth, R.W., 2010. Oxidation state of the lithospheric mantle beneath Diavik diamond mine, central Slave craton, NWT, Canada. Contributions to Mineralogy and Petrology 159, 645–657. Daniels, L.R.M., Gurney, J.J., 1991. Oxygen fugacity constraints on the southern African lithosphere. Contributions to Mineralogy and Petrology 108 (1–2), 154–161. Dasgupta, R., Hirschmann, M.M., Withers, A.C., 2004. Deep global cycling of carbon constrained by the solidus of anhydrous, carbonated eclogite under upper mantle conditions. Earth and Planetary Science Letters 227 (1–2), 73–85. Dawson, J.B., Stephens, W.E., 1975. Statistical classification of garnets from kimberlite and associated xenoliths. Journal of Geology 83 (5), 589–607. Day, H.W., 2012. A revised diamond–graphite transition curve. American Mineralogist 97 (1), 52–62. De Vries, R.C., 1975. Plastic deformation and “work-hardening” of diamond. Materials Research Bulletin 10, 1193–1200. Deines, P., 1980. The carbon isotopic composition of diamonds: relationship to diamond shape, color, occurrence and vapor composition. Geochimica et Cosmochimica Acta 44, 943–961. Di Pierro, S., Gnos, E., Grobety, B.H., Armbruster, T., Bernasconi, S.M., Ulmer, P., 2003. Rock-forming moissanite (natural alpha-silicon carbide). American Mineralogist 88, 1817–1821. Eggler, D.H., Baker, D.R., 1982. Reduced volatiles in the system C–O–H: implications for mantle melting, fluid formation and diamond genesis. In: Akimoto, S., Manghnani, M.H. (Eds.), High pressure research in geophysics. Center for Academic Publications, Tokyo, pp. 237–250. Eugster, H.P., 1957. Heterogeneous reactions involving oxidation and reduction at high pressures and temperatures. Journal of Chemical Physics 26, 1760–1761. Foley, S.F., 2011. A reappraisal of redox melting in the Earth's mantle as a function of tectonic setting and time. Journal of Petrology 52 (7–8), 1363–1391. Foley, S.F., Yaxley, G.M., Rosenthal, A., Buhre, S., Kiseeva, E.S., Rapp, R.P., Jacob, D.E., 2009. The composition of near-solidus melts of peridotite in the presence of CO2 and H2O between 40 and 60 kbar. Lithos 112, 274–283. French, B.M., 1966. Some geological implications of equilibrium between graphite and a C–H–O gas phase at high temperatures and pressures. Reviews of Geophysics 4 (2), 223–253. Frey, F.A., Green, D.H., 1974. The mineralogy, geochemistry and origin of lherzolite inclusions in Victorian basanites. Geochimica et Cosmochimica Acta 38, 1023–1059. Frost, D.J., McCammon, C.A., 2008. The redox state of Earth's mantle. Annual Review of Earth and Planetary Sciences 36, 389–420. Goncharov, A.G., Ionov, D.A., 2012. Redox state of deep off-craton lithospheric mantle: new data from garnet and spinel peridotites from Vitim, southern Siberia. Contributions to Mineralogy and Petrology 164, 731–745. Goncharov, A.G., Ionov, D.A., Doucet, L.S., Pokhilenko, L.N., 2012. Thermal state, oxygen fugacity and C–O–H fluid speciation in cratonic lithospheric mantle: new data on peridotite xenoliths from the Udachnaya kimberlite, Siberia. Earth and Planetary Science Letters 357, 99–110. Green, D.H., 1990. The role of oxidation–reduction and C–H–O fluids in determining melting conditions and magma compositions in the upper mantle. Proceedings of the Indian Academy of Sciences — Earth and Planetary Sciences 99 (1), 153–165. Green, D.H., Falloon, T.J., 1998. Pyrolite: a Ringwood concept and its current expression. In: Jackson, I. (Ed.), The Earth's mantle. Cambridge University Press, pp. 311–378. Griffin, W.L., Ryan, C.G., 1995. Trace-elements in indicator minerals — area selection and target evaluation in diamond exploration. Journal of Geochemical Exploration 53 (1–3), 311–337. Griffin, W.L., Jaques, A.L., Sie, S.H., Ryan, C.G., Cousens, D.R., Suter, G.F., 1988. Conditions of diamond growth — a proton microprobe study of inclusions in west Australian diamonds. Contributions to Mineralogy and Petrology 99 (2), 143–158. Griffin, W.L., O'Reilly, S.Y., Natapov, L.M., Ryan, C.G., 2003. The evolution of lithospheric mantle beneath the Kalahari Craton and its margins. Lithos 71 (2–4), 215–241. Grütter, H.S., 2009. Pyroxene xenocryst geotherms: techniques and application. Lithos 112 (S2), 1167–1178. Grütter, H.S., Gurney, J.J., Menzies, A.H., Winter, F., 2004. An updated classification scheme for mantle-derived garnet, for use by diamond explorers. Lithos 77 (1–4), 841–857. Gudmundsson, G., Wood, B.J., 1995. Experimental tests of garnet peridotite oxygen barometry. Contributions to Mineralogy and Petrology 119, 56–67. Gurney, J.J., 1984. A correlation between garnets and diamonds in kimberlites. In: Glover, J.E., Harris, P.G. (Eds.), Kimberlite Occurrence and Origin: A Basis for Conceptual Models in Exploration. Publs Geol. Dept. & Univ. Extension, Univ. West. Aust., Perth, pp. 143–166. Gurney, J.J., 1989. Diamonds. In: Ross, J., et al. (Eds.), Kimberlites and Related Rocks. GSA Spec Publ 14. Blackwell, Carlton, pp. 935–965. Gurney, J.J., Hildebrand, P.R., Carlson, J.A., Fedortchouk, Y., Dyck, D.R., 2004. The morphological characteristics of diamonds from the Ekati property, Northwest Territories, Canada. Lithos 77 (1–4), 21–38. Gurney, J.J., Helmstaedt, H.H., Richardson, S.H., Shirey, S.B., 2010. Diamonds through time. Economic Geology 105 (3), 689–712. Haggerty, S.E., 1986. Diamond genesis in a multiply-constrained model. Nature 320 (6057), 34–37. Harley, S.L., 1984. An experimental study of the partitioning of iron and magnesium between garnet and orthopyroxene. Contributions to Mineralogy and Petrology 86, 359–373.

218

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

Harris, J.W., Stachel, T., Leost, I., Brey, G.P., 2004. Peridotitic diamonds from Namibia: constraints on the composition and evolution of their mantle source. Lithos 77 (1–4), 209–223. Harte, B., Harris, J.W., 1994. Lower mantle associations preserved in diamonds. Mineralogical Magazine 58A, 384–385. Harte, B., Fitzsimons, I.C.W., Harris, J.W., Otter, M.L., 1999a. Carbon isotope ratios and nitrogen abundances in relation to cathodoluminescence characteristics for some diamonds from the Kaapvaal Province, S-Africa. Mineralogical Magazine 63 (6), 829–856. Harte, B., Harris, J.W., Hutchison, M.T., Watt, G.R., Wilding, M.C., 1999b. Lower mantle mineral associations in diamonds from Sao Luiz, Brazil. In: Fei, Y., Bertka, C.M., Mysen, B.O. (Eds.), Mantle Petrology: Field Observations and High Pressure Experimentation: A Tribute to Francis R. (Joe) Boyd. Special Publication. The Geochemical Society, Houston, pp. 125–153. Hasterok, D., Chapman, D.S., 2011. Heat production and geotherms for the continental lithosphere. Earth and Planetary Science Letters 307 (1–2), 59–70. Hiraga, T., Kohlstedt, D.L., 2009. Systematic distribution of incompatible elements in mantle peridotite: importance of intra- and inter-granular melt-like components. Contributions to Mineralogy and Petrology 158 (2), 149–167. Holland, T.J.B., Powell, R., 2011. An improved and extended internally consistent thermodynamic dataset for phases of petrological interest, involving a new equation of state for solids. Journal of Metamorphic Geology 29, 333–383. Huizenga, J.M., Crossingham, A., Viljoen, F., 2012. Diamond precipitation from ascending reduced fluids in the Kaapvaal lithosphere: thermodynamic constraints. Comptes Rendus Geoscience 344 (2), 67–76. Ickert, R.B., Stachel, T., Stern, R.A., Harris, J.W., 2013. Diamond from recycled crustal carbon documented by coupled δ18O — δ13C measurements of diamonds and their inclusions. Earth and Planetary Science Letters 364, 85–97. Irifune, T., Kurio, A., Sakamoto, S., Inoue, T., Sumiya, H., Funakoshi, K., 2004. Formation of pure polycrystalline diamond by direct conversion of graphite at high pressure and high temperature. Physics of the Earth and Planetary Interiors 143–44, 593–600. Jaques, A.L., Hall, A.E., Sheraton, J.W., Smith, C.B., Sun, S.-S., Drew, R.M., Foudoulis, C., Ellingsen, K., 1989. Composition of crystalline inclusions and C-isotopic composition of Argyle and Ellendale diamonds. In: Ross, J., et al. (Eds.), Kimberlites and Related Rocks. GSA Spec Publ 14. Blackwell, Carlton, pp. 966–989. Keller, R., Taylor, L., Snyder, G., Sobolev, V., Carlson, W., Sobolev, N., Pokhilenko, N., 1998. 3-D petrography of diamondiferous eclogite from Udachnaya, Siberia. Seventh International Kimberlite Conference, Extended Abstracts, pp. 405–407. Kessel, R., Ulmer, P., Pettke, T., Schmidt, M.W., Thompson, A.B., 2005. The water–basalt system at 4 to 6 GPa: Phase relations and second critical endpoint in a K-free eclogite at 700 to 1400 °C. Earth and Planetary Science Letters 237 (3–4), 873–892. Klein-BenDavid, O., Logvinova, A.M., Schrauder, M., Spetius, Z.V., Weiss, Y., Hauri, E.H., Kaminsky, F.V., Sobolev, N.V., Navon, O., 2009. High-Mg carbonatitic microinclusions in some Yakutian diamonds — a new type of diamond-forming fluid. Lithos 112 (Suppl. 2), 648–659. Knoche, R., Sweeney, R.J., Luth, R.W., 1999. Carbonation and decarbonation of eclogites: the role of garnet. Contributions to Mineralogy and Petrology 135 (4), 332–339. Kopylova, M.G., Gurney, J.J., Daniels, L.R.M., 1997. Mineral inclusions in diamonds from the River Ranch kimberlite, Zimbabwe. Contributions to Mineralogy and Petrology 129 (4), 366–384. Kopylova, M., Navon, O., Dubrovinsky, L., Khachatryan, G., 2010. Carbonatitic mineralogy of natural diamond-forming fluids. Earth and Planetary Science Letters 291 (1–4), 126–137. Kramers, J.D., 1979. Lead, uranium, strontium, potassium and rubidium in inclusionbearing diamonds and mantle-derived xenoliths from southern Africa. Earth and Planetary Science Letters 42 (1), 58–70. Krogh, E., 1988. The garnet–clinopyroxene iron–magnesium geothermometer — a reinterpretation of existing experimental data. Contributions to Mineralogy and Petrology 99, 44–48. Kushiro, I., Syono, Y., Akimoto, S.I., 1968. Melting of a peridotite nodule at high pressures and high water pressures. Journal of Geophysical Research 73 (18), 6023–6029. Lazarov, M., Woodland, A.B., Brey, G.P., 2009. Thermal state and redox conditions of the Kaapvaal mantle: a study of xenoliths from the Finsch mine, South Africa. Lithos 112, 913–923. Leahy, K., Taylor, W.R., 1997. The influence of the Glennie domain deep structure on the diamonds in Saskatchewan kimberlites. Russian Geology and Geophysics 38 (2), 451–460. Leost, I., Stachel, T., Brey, G.P., Harris, J.W., Ryabchikov, I.D., 2003. Diamond formation and source carbonation: mineral associations in diamonds from Namibia. Contributions to Mineralogy and Petrology 145 (1), 15–24. Lewis, H.C., 1887. On a diamondiferous peridotite, and the genesis of the diamond. Geological Magazine 4, 22–24. Litasov, K.D., Shatskiy, A., Ohtani, E., 2014. Melting and subsolidus phase relations in peridotite and eclogite systems with reduced COH fluid at 3–16 GPa. Earth and Planetary Science Letters 391, 87–99. Luth, R.W., 1993. Diamonds, eclogites, and the oxidation state of the Earth's mantle. Science 261 (5117), 66–68. Luth, R.W., 2014. 3.9 — Volatiles in Earth's mantle. In: Holland, H.D., Turekian, K.K. (Eds.), Treatise on Geochemistry, Second edition Elsevier, Oxford, pp. 355–391. Luth, R.W., Stachel, T., 2014. The buffering capacity of lithospheric mantle: implications for diamond formation. Contributions to Mineralogy and Petrology 168. http://dx. doi.org/10.1007/s00410-014-1083-6. Luth, R.W., Virgo, D., Boyd, F.R., Wood, B.J., 1990. Ferric iron in mantle-derived garnets — implications for thermobarometry and for the oxidation state of the mantle. Contributions to Mineralogy and Petrology 104, 56–72.

Malaspina, N., Poli, S., Fumagalli, P., 2009. The oxidation state of metasomatized mantle wedge: insights from C–O–H-bearing garnet peridotite. Journal of Petrology 50, 1533–1552. Malaspina, N., Scambelluri, M., Poli, S., Van Roermund, H.L.M., Langenhorst, F., 2010. The oxidation state of mantle wedge majoritic garnet websterites metasomatised by Cbearing subduction fluids. Earth and Planetary Science Letters 298, 417–426. McCammon, C., Kopylova, M.G., 2004. A redox profile of the Slave mantle and oxygen fugacity control in the cratonic mantle. Contributions to Mineralogy and Petrology 148, 55–68. McDonough, W.F., Sun, S.S., 1995. The composition of the Earth. Chemical Geology 120, 223–253. McLean, H., Banas, A., Creighton, S., Whiteford, S., Luth, R.W., Stachel, T., 2007. Garnet xenocrysts from the diavik mine, NWT, Canada: composition, colour, and paragenesis. Canadian Mineralogist 45, 1131–1145. Meyer, H.O.A., 1968. Chrome pyrope — an inclusion in natural diamond. Science 160 (3835), 1446–1447. Meyer, H.O.A., 1987. Inclusions in diamond. In: Nixon, P.H. (Ed.), Mantle Xenoliths. John Wiley & Sons Ltd., Chichester, pp. 501–522. Meyer, H.O.A., Boyd, F.R., 1972. Composition and origin of crystalline inclusions in natural diamonds. Geochimica et Cosmochimica Acta 36 (11), 1255–1273. Meyer, H.O.A., McCallum, M.E., 1986. Mineral inclusions in diamonds from the Sloan Kimberlites, Colorado. Journal of Geology 94 (4), 600–612. Meyer, H.O.A., Tsai, H.M., 1976. Mineral inclusions in diamond — temperature and pressure of equilibration. Science 191 (4229), 849–851. Moore, R.O., Gurney, J.J., 1985. Pyroxene solid solution in garnets included in diamonds. Nature 318 (6046), 553–555. Moore, R.O., Gurney, J.J., 1989. Mineral inclusions in diamond from Monastery kimberlite, South Africa. In: Ross, J., et al. (Eds.), Kimberlites and Related Rocks. GSA Spec Publ 14. Blackwell, Carlton, pp. 1029–1041. Navon, O., Hutcheon, I.D., Rossman, G.R., Wasserburg, G.J., 1988. Mantle-derived fluids in diamond micro-inclusions. Nature 335 (6193), 784–789. Newton, R.C., Sharp, W.E., 1975. Stability of forsterite + CO2 and its bearing on the role of CO2 in the mantle. Earth and Planetary Science Letters 26, 239–244. Nimis, P., 2002. The pressures and temperatures of formation of diamond based on thermobarometry of chromian diopside inclusions. Canadian Mineralogist 40 (Part 3), 871–884. Nimis, P., Grütter, H., 2010. Internally consistent geothermometers for garnet peridotites and pyroxenites. Contributions to Mineralogy and Petrology 159 (3), 411–427. Nimis, P., Taylor, W.R., 2000. Single clinopyroxene thermobarometry for garnet peridotites. Part I. Calibration and testing of a Cr-in-Cpx barometer and an enstatite-in-Cpx thermometer. Contributions to Mineralogy and Petrology 139 (5), 541–554. O'Neill, H.S.C., 1980. An experimental study of the iron–magnesium partitioning between garnet and olivine and its calibration as a geothermometer: corrections. Contributions to Mineralogy and Petrology 72, 337. O'Neill, H.S.C., Wall, V.J., 1987. The olivine–orthopyroxene–spinel oxygen-geobarometer, the nickel precipitation curve and the oxygen fugacity of the Earth's upper mantle. Journal of Petrology 28, 1169–1191. O'Neill, H.S.C., Wood, B.J., 1979. An experimental study of the Fe–Mg partitioning between garnet and olivine and its calibration as a geothermometer. Contributions to Mineralogy and Petrology 70, 59–70. Otter, M.L., Gurney, J.J., 1989. Mineral inclusions in diamond from the Sloan diatremes, Colorado–Wyoming State Line kimberlite district, North America. In: Ross, J., et al. (Eds.), Kimberlites and Related Rocks. GSA Spec. Publ. 14. Blackwell, Carlton, pp. 1042–1053. Palot, M., Pearson, D.G., Stern, R.A., Stachel, T., Harris, J.W., 2013. Multiple growth events, processes and fluid sources involved in diamond genesis: a micro-analytical study of sulphide-bearing diamonds from Finsch mine, RSA. Geochimica et Cosmochimica Acta 106, 51–70. Pearson, D.G., Shirey, S.B., 1999. Isotopic dating of diamonds. In: Ruiz, J., Lambert, D.D. (Eds.), Applications of Radiogenic Isotopes to Ore Deposit Research. Economic Geology special publication: SEG Reviews in Economic Geology, pp. 143–171. Pearson, D.G., Davies, G.R., Nixon, P.H., Milledge, H.J., 1989. Graphitized diamonds from a peridotite massif in Morocco and implications for anomalous diamond occurrences. Nature 338 (6210), 60–62. Pearson, D.G., Shirey, S.B., Harris, J.W., Carlson, R.W., 1998. Sulphide inclusions in diamonds from the Koffiefontein kimberlite, S Africa: constraints on diamond ages and mantle ReOs systematics. Earth and Planetary Science Letters 160 (3–4), 311–326. Peats, J., Stachel, T., Stern, R.A., Muehlenbachs, K., Armstrong, J., 2012. Aviat diamonds: a window into the deep lithospheric mantle beneath the northern Churchill Province, Melville Peninsula, Canada. Canadian Mineralogist 50 (3), 611–624. Petts, D.C., Stern, R.A., Stachel, T., Chacko, T., Heaman, L., 2014. A nitrogen isotope fractionation factor between diamond and fluid derived from detailed SIMS analysis of an eclogitic diamond. V. M. Goldschmidt Conference, Sacramento, June 2014, Abstract #1953. Petts, D.C., Stern, R.A., Chacko, T., Stachel, T., Heaman, L.M., 2015. A Nitrogen Isotope Fractionation Factor Between Diamond and Its Parental Fluid Derived From Detailed SIMS Analysis of an Eclogitic Diamond (submitted for publication). Phillips, D., Harris, J.W., 1995. Geothermobarometry of diamond inclusions from the De Beers Pool Mines, Kimberley. Sixth International Kimberlite Conference, Novosibirsk, Extended Abstracts, pp. 441–443. Phillips, D., Harris, J.W., Viljoen, K.S., 2004. Mineral chemistry and thermobarometry of inclusions from De Beers Pool diamonds, Kimberley, South Africa. Lithos 77 (1–4), 155–179. Prinz, M., Manson, D.V., Hlava, P.F., Keil, K., 1975. Inclusions in diamonds: garnet lherzolite and eclogite assemblages. Physics and Chemistry of the Earth 9, 797–815.

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220 Purwin, H., Lauterbach, S., Brey, G., Woodland, A., Kleebe, H.-J., 2013. An experimental study of the Fe oxidation states in garnet and clinopyroxene as a function of temperature in the system CaO–FeO–Fe 2 O 3 –MgO–Al2 O 3 –SiO 2 : implications for garnet–clinopyroxene geothermometry. Contributions to Mineralogy and Petrology 165 (4), 623–639. Richardson, S.H., Harris, J.W., 1997. Antiquity of peridotitic diamonds from the Siberian craton. Earth and Planetary Science Letters 151 (3–4), 271–277. Richardson, S.H., Gurney, J.J., Erlank, A.J., Harris, J.W., 1984. Origin of diamonds in old enriched mantle. Nature 310 (5974), 198–202. Rickard, R.S., Harris, J.W., Gurney, J.J., Cardoso, P., 1989. Mineral inclusions from Koffiefontein mine. In: Ross, J., et al. (Eds.), Kimberlites and Related Rocks. GSA Spec. Publ. 14. Blackwell, Carlton, pp. 1054–1062. Rohrbach, A., Ballhaus, C., Golla Schindler, U., Ulmer, P., Kamenetsky, V.S., Kuzmin, D.V., 2007. Metal saturation in the upper mantle. Nature 449, 456–458. Rohrbach, A., Ballhaus, C., Ulmer, P., Golla Schindler, U., Schönbohm, D., 2011. Experimental evidence for a reduced metal-saturated upper mantle. Journal of Petrology 52, 717–731. Rosenhauer, M., Woermann, E., Knecht, B., Ulmer, C.G., 1977. The stability of graphite and diamond as a function of the oxygen fugacity in the mantle. 2nd International Kimberlite Conference, Santa Fe, p. 297. Saguy, C., 2004. Diffusion of light elements in diamond. Defects and Diffusion in Ceramics — an Annual Retrospective VI, 226–228, pp. 31–48. Schmidberger, S.S., Simonetti, A., Francis, D., 2003. Small-scale Sr isotope investigation of clinopyroxenes from peridotite xenoliths by laser ablation MC-ICP-MS — implications for mantle metasomatism. Chemical Geology 199 (3–4), 317–329. Schulze, D.J., 1989. Constraints on the abundance of eclogite in the upper mantle. Journal of Geophysical Research-Solid Earth and Planets 94 (B4), 4205–4212. Scott Smith, B.H., Danchin, R.V., Harris, J.W., Stracke, K.J., 1984. Kimberlites near Orroroo, South Australia. In: Kornprobst, J. (Ed.), Kimberlites I: Kimberlites and Related Rocks. Elsevier, Amsterdam, pp. 121–142. Shirey, S.B., Harris, J.W., Richardson, S.H., Fouch, M.J., James, D.E., Cartigny, P., Deines, P., Viljoen, K.S., 2002. Diamond genesis, seismic structure, and evolution of the Kaapvaal – Zimbabwe craton. Science 297 (5587), 1683–1686. Shirey, S.B., Cartigny, P., Frost, D.J., Keshav, S., Nestola, F., Nimis, P., Pearson, D.G., Sobolev, N.V., Walter, M.J., 2013. Diamonds and the geology of mantle carbon. In: Hazen, R.M., Jones, A.P., Baross, J.A. (Eds.), Carbon in EarthReviews in Mineralogy & Geochemistry 355–421. Shiryaev, A.A., Griffin, W.L., Stoyanov, E., 2011. Moissanite (SiC) from kimberlites: polytypes, trace elements, inclusions and speculations on origin. Lithos 122, 152–164. Smart, K.A., Heaman, L.M., Chacko, T., Simonetti, A., Kopylova, M., Mah, D., Daniels, D., 2009. The origin of high-MgO diamond eclogites from the Jericho Kimberlite, Canada. Earth and Planetary Science Letters 284 (3–4), 527–537. Smart, K.A., Chacko, T., Stachel, T., Muehlenbachs, K., Stern, R.A., Heaman, L.M., 2011. Diamond growth from oxidized carbon sources beneath the Northern Slave Craton, Canada: a δ13C–N study of eclogite-hosted diamonds from the Jericho kimberlite. Geochimica et Cosmochimica Acta 75 (20), 6027–6047. Smit, K.V., Shirey, S.B., Richardson, S.H., le Roex, A.P., Gurney, J.J., 2010. Re–Os isotopic composition of peridotitic sulphide inclusions in diamonds from Ellendale, Australia: age constraints on Kimberley cratonic lithosphere. Geochimica et Cosmochimica Acta 74 (11), 3292–3306. Sobolev, N.V., Lavrent'ev, Y.G., Pospelova, L.N., Sobolev, V.C., 1969. Chrome pyrope from Yakutian diamonds. Doklady Akademii Nauk SSSR 189 (1), 133–136. Sobolev, N.V., Efimova, E.S., Pospelova, L.N., 1981. Native iron in Yakutian diamonds and its paragenesis. Geologiya i Geofizika 22, 25–29 (in Russian). Sobolev, N.V., Logvinova, A.M., Efimova, E.S., 2009. Syngenetic phlogopite inclusions in kimberlite-hosted diamonds: implications for role of volatiles in diamond formation. Russian Geology and Geophysics 50, 1234–1248. Stachel, T., Harris, J.W., 1997. Diamond precipitation and mantle metasomatism — evidence from the trace element chemistry of silicate inclusions in diamonds from Akwatia, Ghana. Contributions to Mineralogy and Petrology 129 (2–3), 143–154. Stachel, T., Harris, J.W., 2008. The origin of cratonic diamonds — constraints from mineral inclusions. Ore Geology Reviews 34 (1–2), 5–32. Stachel, T., Harris, J.W., Brey, G.P., 1998. Rare and unusual mineral inclusions in diamonds from Mwadui, Tanzania. Contributions to Mineralogy and Petrology 132 (1), 34–47. Stachel, T., Brey, G.P., Harris, J.W., 2000a. Kankan diamonds (Guinea) I: from the lithosphere down to the transition zone. Contributions to Mineralogy and Petrology 140, 1–15. Stachel, T., Harris, J.W., Brey, G.P., Joswig, W., 2000b. Kankan diamonds (Guinea) II: lower mantle inclusion parageneses. Contributions to Mineralogy and Petrology 140 (1), 16–27. Stachel, T., Harris, J.W., Tappert, R., Brey, G.P., 2003. Peridotitic diamonds from the Slave and the Kaapvaal cratons — similarities and differences based on a preliminary data set. Lithos 71 (2–4), 489–503. Stachel, T., Aulbach, S., Brey, G.P., Harris, J.W., Leost, I., Tappert, R., Viljoen, K.S., 2004. The trace element composition of silicate inclusions in diamonds: a review. Lithos 77 (1–4), 1–19. Stachel, T., Harris, J.W., Muehlenbachs, K., 2009. Sources of carbon in inclusion bearing diamonds. Lithos 112 (S2), 625–637. Stachel, T., Harris, J.W., Hunt, L., Muehlenbachs, K., Kobussen, A., 2015. Argyle diamonds — how subduction along the Kimberley Craton edge generated the world's biggest diamond deposit. Economic Geology (in press). Stagno, V., Frost, D.J., 2010. Carbon speciation in the asthenosphere: experimental measurements of the redox conditions at which carbonate-bearing melts coexist with graphite or diamond in peridotite assemblages. Earth and Planetary Science Letters 300, 72–84.

219

Stagno, V., Ojwang, D.O., McCammon, C.A., Frost, D.J., 2013. The oxidation state of the mantle and the extraction of carbon from Earth's interior. Nature 493, 84–88. Stosch, H.G., Lugmair, G.W., 1986. Trace element and Sr and Nd isotope geochemistry of peridotite xenoliths from the Eifel (West Germany) and their bearing on the evolution of the subcontinental lithosphere. Earth and Planetary Science Letters 80, 281–298. Sunagawa, I., 1984. Morphology of natural and synthetic diamond crystals. In: Sunagawa, I. (Ed.), Materials Science of the Earth's Interior. Terra Scientific, Tokyo, pp. 303–330. Tappert, R., Stachel, T., Harris, J.W., Muehlenbachs, K., Ludwig, T., Brey, G.P., 2005. Diamonds from Jagersfontein (South Africa): messengers from the sublithospheric mantle. Contributions to Mineralogy and Petrology 150 (5), 505–522. Taylor, W.R., Green, D.H., 1988. Measurement of reduced peridotite-C–O–H solidus and implications for redox melting of the mantle. Nature 332 (6162), 349–352. Taylor, W.R., Green, D.H., 1989. The role of reduced C–O–H fluids in mantle partial melting. In: Ross, J., et al. (Eds.), Kimberlites and Related Rocks: Their Composition, Occurrence, Origin and Emplacement. GSA Spec Publ No 14. Blackwell, Carlton, pp. 592–602. Taylor, L.A., Milledge, H.J., Bulanova, G.P., Snyder, G.A., Keller, R.A., 1998. Metasomatic eclogitic diamond growth: evidence from multiple diamond inclusions. International Geology Review 40 (8), 663–676. Thomassot, E., Cartigny, P., Harris, J.W., Viljoen, K.S., 2007. Methane-related diamond crystallization in the Earth's mantle: stable isotope evidences from a single diamond-bearing xenolith. Earth and Planetary Science Letters 257 (3–4), 362–371. Ulmer, G.C., Grandstaff, D.E., Woermann, E., Gobbels, M., Schonitz, M., Woodland, A.B., 1998. The redox stability of moissanite (SiC) compared with metal–metal oxide buffers at 1773 K and at pressures up to 90 kbar. Neues Jahrbuch Fur MineralogieAbhandlungen 172, 279–307. Viljoen, K.S., Dobbe, R., Smit, B., Thomassot, E., Cartigny, P., 2004. Petrology and geochemistry of a diamondiferous lherzolite from the Premier diamond mine, South Africa. Lithos 77 (1–4), 539–552. Walter, M.J., Kohn, S.C., Araujo, D., Bulanova, G.P., Smith, C.B., Gaillou, E., Wang, J., Steele, A., Shirey, S.B., 2011. Deep mantle cycling of oceanic crust: evidence from diamonds and their mineral inclusions. Science 333 (6052), 54–57. Wang, A., Pasteris, J.D., Meyer, H.O.A., Deleduboi, M.L., 1996. Magnesite-bearing inclusion assemblage in natural diamond. Earth and Planetary Science Letters 141 (1–4), 293–306. Watson, E.B., Brenan, J.M., Baker, D.R., 1990. Distribution of fluids in the continental mantle. In: Menzies, M.A. (Ed.), Continental MantleOxford Monographs on Geology and Geophysics 111–125. Weiss, Y., Kiflawi, I., Davies, N., Navon, O., 2014. High-density fluids and the growth of monocrystalline diamonds. Geochimica et Cosmochimica Acta 141, 145–159. Wiggers de Vries, D.F., Bulanova, G.P., De Corte, K., Pearson, D.G., Craven, J.A., Davies, G.R., 2013a. Micron-scale coupled carbon isotope and nitrogen abundance variations in diamonds: evidence for episodic diamond formation beneath the Siberian Craton. Geochimica et Cosmochimica Acta 100, 176–199. Wiggers de Vries, D.F., Pearson, D.G., Bulanova, G.P., Smelov, A.P., Pavlushin, A.D., Davies, G.R., 2013b. Re–Os dating of sulphide inclusions zonally distributed in single Yakutian diamonds: evidence for multiple episodes of Proterozoic formation and protracted timescales of diamond growth. Geochimica et Cosmochimica Acta 120, 363–394. Williams, A.S., 1932. The Genesis of the Diamond 2 vol. Ernest Benn Limited, London (636 pp.). Wood, B.J., 1993. Carbon in the core. Earth and Planetary Science Letters 117 (3–4), 593–607. Wood, B.J., Bryndzia, L.T., Johnson, K.E., 1990. Mantle oxidation state and its relationship to tectonic environment and fluid speciation. Science 248 (4953), 337–345. Woodland, A.B., 2009. Ferric iron contents of clinopyroxene from cratonic mantle and partitioning behaviour with garnet. Lithos 112 (Suppl. 2), 1143–1149. Woodland, A.B., Koch, M., 2003. Variation in oxygen fugacity with depth in the upper mantle beneath the Kaapvaal craton, Southern Africa. Earth and Planetary Science Letters 214 (1–2), 295–310. 3+ 3 Woodland, A.B., O'Neill, H.S.C., 1993. Synthesis and stability of Fe2+ 3 Fe2 Si O12 garnet and phase relations with Fe3Al2Si3O12–Fe32+Fe3+ 2 Si3O12 solutions. American Mineralogist 78 (9–10), 1002–1015. Woodland, A.B., Peltonen, P., 1999. Ferric iron contents of garnet and clinopyroxene and estimated oxygen fugacities of peridotite xenoliths from the Eastern Finland Kimberlite Province. In: Gurney, J.J., Gurney, J.L., Pascoe, M.D., Richardson, S.H. (Eds.), The P.H. Nixon Volume, Proceedings of the VIIth International Kimberlite Conference. Red Roof Design, Cape Town, pp. 904–977. Wyllie, P.J., 1987. Metasomatism and fluid generation in mantle xenoliths. In: Nixon, P.H. (Ed.), Mantle Xenoliths. John Wiley & Sons Ltd., Chichester, pp. 609–621. Wyllie, P.J., Huang, W.-L., 1976. Carbonation and melting reactions in the system CaO– MgO–SiO2–CO2 at mantle pressures with geophysical and petrological applications. Contributions to Mineralogy and Petrology 54, 79–107. Wyllie, P.J., Ryabchikov, I.D., 2000. Volatile components, magmas, and critical fluids in upwelling mantle. Journal of Petrology 41 (7), 1195–1206. Wyllie, P.J., Huang, W.L., Otto, J., Byrnes, A.P., 1983. Carbonation of peridotites and decarbonation of siliceous dolomites represented in the system CaO–MgO–SiO2–CO2 to 30 kbar. Tectonophysics 100, 359–388. Yasuda, A., Fujii, T., Kurita, K., 1994. Melting phase-relations of an anhydrous midocean ridge basalt from 3 to 20 GPa — implications for the behavior of

220

T. Stachel, R.W. Luth / Lithos 220–223 (2015) 200–220

subducted oceanic-crust in the mantle. Journal of Geophysical Research-Solid Earth 99 (B5), 9401–9414. Yaxley, G.M., Berry, A.J., Kamenetsky, V.S., Woodland, A.B., Golovin, A.V., 2012. An oxygen fugacity profile through the Siberian Craton — Fe K-edge XANES determinations of Fe3+/∑Fe in garnets in peridotite xenoliths from the Udachnaya East kimberlite. Lithos 140–141, 142–151.

Zhang, C., Duan, Z., 2009. A model for C–O–H fluid in the Earth's mantle. Geochimica et Cosmochimica Acta 73, 2089–2102. Zhang, C., Duan, Z., 2010. GFluid an Excel spreadsheet for investigating C–O–H fluid composition under high temperatures and pressures. Computers & Geosciences 36, 569–572.