Hydrothermal alteration of kimberlite by convective flows of external water
- First Online:
- 1.1k Downloads
Kimberlite volcanism involves the emplacement of olivine-rich volcaniclastic deposits into volcanic vents or pipes. Kimberlite deposits are typically pervasively serpentinised as a result of the reaction of olivine and water within a temperature range of 130–400 °C or less. We present a model for the influx of ground water into hot kimberlite deposits coupled with progressive cooling and serpentisation. Large-pressure gradients cause influx and heating of water within the pipe with horizontal convergent flow in the host rock and along pipe margins, and upward flow within the pipe centre. Complete serpentisation is predicted for wide ranges of permeability of the host rocks and kimberlite deposits. For typical pipe dimensions, cooling times are centuries to a few millennia. Excess volume of serpentine results in filling of pore spaces, eventually inhibiting fluid flow. Fresh olivine is preserved in lithofacies with initial low porosity, and at the base of the pipe where deeper-level host rocks have low permeability, and the pipe is narrower leading to faster cooling. These predictions are consistent with fresh olivine and serpentine distribution in the Diavik A418 kimberlite pipe, (NWT, Canada) and with features of kimberlites of the Yakutian province in Russia affected by influx of ground water brines. Fast reactions and increases in the volume of solid products compared to the reactants result in self-sealing and low water–rock ratios (estimated at <0.2). Such low water–rock ratios result in only small changes in stable isotope compositions; for example, δO18 is predicted only to change slightly from mantle values. The model supports alteration of kimberlites predominantly by interactions with external non-magmatic fluids.
KeywordsKimberlites Hydrothermal alteration Serpentine
Many volcaniclastic kimberlite deposits are strongly altered with the original igneous minerals typically altered to hydrous minerals and carbonate. Serpentine is ubiquitously a major component of the alteration assemblage as a consequence of the high abundance of olivine in kimberlite. Fresh olivine is commonly not preserved (Mitchell 1986, 2013; Sparks 2013), and features such as serpentine pseudomorphing olivine, and serpentine replacing groundmass phases and precipitating in pore space, are observed. The origin of the water is controversial. Mitchell et al. (2009, 2013) have attributed the serpentine-dominated hydrous mineral assemblages to alteration related to deuteric (magmatic) fluids and regarded serpentine as a primary mineral. In contrast, others (Stripp et al. 2006; Hayman et al. 2009; Buse et al. 2011; Porritt et al. 2012; Giuliani et al. 2014) have argued that alteration is the result of circulation by external water through hot olivine-rich pyroclastic deposits. This controversy is addressed here through numerical modelling of post-emplacement external water circulation through kimberlite pipes. The focus here is on serpentinisation as the dominant reaction. The model couples fluid flow, heat transfer, and serpentinisation reactions with the evolution of porosity and permeability in the pipe’s interior. The model shows that strong influx of external water accompanied by hydrothermal metamorphism, represented in the model by serpentine formation, is inevitable. The distribution of kimberlite rocks that preserve fresh olivine is predicted and compared with observations from Diavik A418 kimberlite pipe (NWT, Canada) and also with features of kimberlites of the Yakutian province in Russia. External water is identified as the main cause of alteration on the basis of the models and other evidence. This study concerns volcaniclastic kimberlite deposits in diatremes and not hypabyssal kimberlites.
Geological constraints and context
Kimberlite pipes are the conduits of monogenetic diatreme volcanoes (Lorenz 1975; Sparks et al. 2006; Scott Smith 2008). They are typically 1–3 km deep, steep-sided (~80o) and up to a few 100 m across. Idealised geological models developed in southern Africa (e.g. Hawthrone 1975) recognise a shallow crater zone, a downward tapering pipe, and a root zone, which can be wider than the overlying pipe. Subsequent studies have recognised a wider variety of kimberlite pipes (Field and Scott Smith 1999; Scott Smith 2008), including those that have wide craters and lack diatreme and root zones, and those that are infilled with reworked volcaniclastic deposits.
The geology of kimberlite pipes, summarised by Sparks (2013), indicates that the dominant lithologies infilling kimberlite pipes are volcaniclastic deposits. They typically display a wide diversity of different geological units and lithofacies, reflecting complex multiphase eruptions and a diversity of primary volcanic and secondary reworking processes. Layered and massive volcaniclastic rocks vary widely in grain size, sorting, and componentry; olivine (typically altered to serpentine) is a major constituent of most kimberlites. Contacts between different geological units range from steep near vertical and cross-cutting, to layer cake style. Layering is commonly steep with dips at or well above the angle of repose of loose volcaniclastic deposits, typically oriented towards the pipe centre. Important varieties of kimberlite pipe infill are dark coherent rocks, which have been described as hypabyssal, but more recently have been interpreted as high-temperature welded clastogenic rocks (Brown et al. 2008a). Kimberlite dykes and sills, which are volumetrically insignificant in pipes, are coherent rocks sensu stricto and more commonly preserve fresh olivine. Lithofacies analysis indicates some kimberlite eruptions occurred in a shallow submarine environment with extensive reworking (Pittari et al. 2008), whereas others occurred in sub-aerial settings. We discuss below the role of emplacement temperature in explaining the preservation of fresh olivine.
Recent investigations have highlighted that the emplacement temperatures of volcaniclastic kimberlites vary greatly and include both low-temperature and high-temperature varieties. Evidence for temperatures of up to several 100 °C includes the following: estimates from the thermal maturity of organic materials (Stasiuk et al. 1999); estimates from typical hydrothermal metamorphic mineral assemblages (Stripp et al. 2006; Buse et al. 2011); estimates of lithic clast temperatures from TRM studies (Fontana et al. 2011); and textural evidence of welding and agglutination of primary magma clasts (Brown et al. 2008a, 2009; van Straaten et al. 2011; Gernon et al. 2012). Low-temperature volcaniclastic deposits, likely related to quenching during phreatomagmatic styles of eruption, have also been recognised (Brown et al. 2008b; Kurszlaukis and Lorenz 2008; Pittari et al. 2008; Porritt et al. 2012; van Straaten and Kopylova 2013).
Host rocks to kimberlite vary greatly, but a fairly common setting is an Archaean basement overlain by younger sedimentary and volcanic strata; this arrangement is typical of many southern African and Canadian kimberlites.
Serpentine in kimberlites includes replacement of primary igneous minerals, principally olivine and components of the igneous groundmass, and precipitation of serpentine in pore space. Textural relationships show that there can be a multiple generations of serpentine (Mitchell 2013), and serpentine can be co-precipitated with other minerals such as diopside and chlorite (Stripp et al. 2006; Mitchell et al. 2009; Buse et al. 2011). Pervasive fracture networks, as commonly observed in serpentinised peridotite (Jamtveit and Austrheim 2010), are not observed in kimberlites. The large volumes changes of the serpentisation reaction are accommodated by filling of pore space, replacement of earlier minerals, and possibly open system loss of components rather than deformation.
The model assumes that the host rocks are saturated with water. The emplacement of the kimberlite is assumed to be very short compared to typical cooling times of the kimberlite pipe and so is simplified to being instantaneous. This assumption will be justified a posteriori. The temperature is high with pores occupied by low-density volcanic volatiles. The initial pore pressure inside the kimberlite is assumed to be very low and approximately atmospheric. Another way of describing the last assumption is that the kimberlite has sufficiently high porosity and high permeability that it is well connected to the atmosphere above. The volcaniclastic deposits that infill kimberlite pipes are inferred to have similar high permeabilities to other kinds of moderately to poorly sorted pyroclastic deposits (Wilson 1984; Miller 1990). If accumulation rates are high, then the deposits might develop transient pore pressures approaching lithostatic values. These pore pressures, however, dissipate rapidly. Scaling analysis (Sparks et al. 1999) allows calculation of an approximate time scale τ ≈ H/(Kpz/μg), where K is permeability, pz is the pressure gradient across the layer, and μg is the gas viscosity. For H = 1 km (a typical pipe depth), μg = 10−5 Pa s, K = 10−12 m−2 (a representative permeability), and pz = 2 × 104 Pa m−1 (a representative lithostatic gradient), and the time scale τ is approximately 1 month to be compared with a typical cooling time in the order of 1,000 years.
With these assumptions, a large horizontal pressure gradient is initially present between the hydrostatic pressure in the host rock and the low pressure in the kimberlite, resulting in water being drawn into the pipe interior along horizontal flow paths. Here, the water is heated and mostly flows upwards in the pipe interior driven by large vertical pressure gradients. The water reacts with the olivine to form serpentine. Besides olivine, other primary igneous minerals, possible glass, and incorporated accidental lithic clasts are reactive and other alteration minerals may form (e.g. talc and chlorite). However, the serpentisation reaction is quite well understood (see below), and olivine is a major constituent, so first-order features of the alteration process can be captured by a model in which only the serpentisation is considered.
Serpentinisation involves a large volume increase, and here, it is assumed that this volume is locally divided between replacing olivine and infilling pore space. With the exception of the addition of water, we model the process as chemically closed with no constituents derived from the original unaltered kimberlite being transported out of the system. Mg and Si are regarded as immobile on a large scale, but locally can be mobile on the scale of pores to facilitate serpentisation reactions. As water flows through and reacts with the kimberlite, heat is generated due to the exothermic character of the reaction. The porosity and permeability of the kimberlite decrease with time as pore space are filled with serpentine.
The process of serpentinisation of ultramafic rocks has mostly been studied in peridotite bodies. The serpentinisation reactions and the conditions in kimberlites share some attributes with those of serpentinisation of other ultramafic bodies, but key differences exist. Kimberlites are rich in olivine and other low-silica primary minerals, and possible glass or melt may be present at emplacement when temperatures are high. The dynamics of emplacement results in the admixture of the primary kimberlitic constituents with xenoliths of other crustal rock types, which are richer in silica. The volcaniclastic deposits of the kimberlite pipes can have an initial porosity of 20–30 % (e.g. Stripp et al. 2006). Fracturing of the country rock that accompanies the emplacement of the cross-cutting kimberlite pipe will likely intersect aquifers, which can deliver large amounts of groundwater to the hot kimberlite pipe and initiate serpentine formation from olivine in addition to phreatic activity.
In order to simulate cooling of a kimberlite body by external water influx, we have used a modified version of the filtration code MUFITS (Afanasyev 2013a, b). The code is developed for simulation of multiphase multicomponent flows in porous media in a wide range of pressures and temperatures, including sub-critical and supercritical conditions. It solves mass conservation laws for individual components (water and CO2) together with energy equation for the system as a whole including the solid rock matrix, and Darcy transport equations for different phases. The equation of state is specified in a table form of spline coefficients for the thermodynamic potential (entropy) that allows calculation of all properties of the mixture in terms of pressure, enthalpy, and total composition. Two modifications of the code were implemented: Serpentinisation of the olivine leads to a decrease in the density of the rock matrix and filling pore spaces resulting in significant decrease in porosity and permeability; latent heat of serpentinisation is accounted for in the energy equation.
Kimberlite and host rock properties
Manning and Ingebritsen (1999) summarised estimates of the bulk permeability of continental crust. They show that in the uppermost 5 km of the crust, the permeability varies between 10−12 and 10−16 m2. We thus have chosen a range of host rock permeabilities between 10−15 and 10−16 m2. Volcaniclastic kimberlites have similar sorting and grain size to many pyroclastic deposits. They are expected to have similar permeabilites to unconsolidated to weakly indurated pyroclastic flow deposits, which have measured permeabilities in the range 3 × 10−13–3 × 10−15 m2 (Wilson 1984; Miller 1990; Sruoga et al. 2004). We use an illustrative ratio of initial kimberlite to host rock permeability of 300 with initial kimberlite permeability of 3 × 10−14 m2. Values for particular calculations are given in the figures displaying the results.
We assume that initially pipe-filling kimberlite deposit consists of two types of material: non-reactive (inert) fragments and olivine. The non-reactive material includes igneous minerals other than olivine and country rock xenoliths, which are treated as non-reactive in this simplified model, although we recognise that in nature, some of these materials are likely to also hydrate. Our goal here though is to develop an alteration model based on the conversion of olivine to serpentine that occurs in the presence of water. The amount of bound water in the rock matrix determines the degree of serpentinisation. While other reactive constituents could be modelled by similar algorithms, their inclusion is not expected to produce significantly different results because they also require similar amounts of water and involve comparable volume changes. Thus, at first order, they can be considered as being represented by the serpentinisation of olivine.
Parameters of kimberlitic rock
Bounded water density ρ*
Bounded water heat capacity C*
3 kJ/(kg K)
Heat of reaction qe
Maximal bonded water content xe
Olivine density ρol
Olivine heat capacity Col
1 kJ/(kg K)
Inert component density ρin
Inert component heat capacity Cin
0.84 kJ/(kg K)
Minimal porosity mmin
Serpentinisation of olivine
Alteration of the petrophysical parameters
At low water pressures, serpentinisation occurs over an intermediate range of temperatures T ∊ [Tα, Tβ] (Fig. 3), where we assume Tα = 130 °C, Tβ = 430 °C. If T > Tβ, then the equilibrium value xeq is equal to zero leading to zero rate of the serpentinisation. This temperature range is arbitrary, but lies within the stability field of serpentine minerals. The Tα used here is slightly above 100 °C, which MacDonald and Fyfe (1985) give as the lower limit of fairly rapid serpentinisation. We assume that in a narrow range of temperatures, the equilibrium bonded water fraction xeq increases up to its maximum value with decreasing temperature and stays constant at lower temperatures.
The characteristic time of reaction k−1(T) at T = Tβ is approximately 20 days (Evans et al. 2013). As the temperature decreases, the reaction time tends to infinity and the reaction rate tends to zero (Fig. 3). This kinetics is oversimplified in many senses. However, as the timescale of kimberlitic pipe cooling is much longer than the reaction timescale, the system will stay close to equilibrium until the temperature decreases significantly, leading to an increase in the kinetic timescale.
We use the model for binary mixture flows in porous media (Afanasyev 2013a) for modelling the hydrodynamic processes associated with meteoric water invasion. The pipe-filling kimberlite deposits can be saturated by a binary mixture of water and a passive fluid component which does not participate in serpentinisation (i.e. air). Because the model of Afanasyev (2013a) is developed for water–CO2 mixtures instead of air, we use CO2 as the passive fluid component. This assumption is consistent with the widespread view that CO2 is major component of the volatile component associated with kimberlite magma (reviewed in Sparks 2013).
The uppermost part of the kimberlite cools down quickly due to heat loss to the atmosphere, and the temperature rapidly drops to below Tα = 130 °C suppressing serpentinisation. This leads to the formation of a high-permeability channel at the top of the pipe, enabling intense water influx in this region (see Fig. 4c). The water front in the deep part of the kimberlite is sharp because the temperatures are high enough that most of the water immediately reacts to form serpentine. A region filled with CO2 is isolated in the centre of the pipe suppressing serpentinisation (e.g. Fig. 4c, 1,370 years). Competition of conductive cooling and water inflow into the pipe leads to a complicated pattern of serpentine distribution because the reaction occurs over a relatively narrow range of temperatures in water-saturated conditions. Serpentinisation ceases after 1,860 years when the temperature within the entire pipe drops below Tα = 130 °C as shown in Fig. 3.
The process of serpentinisation is strongly controlled by the evolution of the porosity and permeability (Eq. 12). Serpentinisation causes a large increase in the volume of the matrix of the kimberlite, reducing water flux through the serpentinised part of the pipe. Reduction of permeability along the pipe margins in the early stages of the cooling results in decrease in water flux into the system and leads to predominantly conductive cooling of the pipe. Only the upper part of the kimberlite, which cools rapidly by heat exchange with the atmosphere, remains mostly unaltered preserving its initial permeability, and this then acts as a water pathway into the pipe.
In order to understand the competing influences of both the heat of the reaction and the cooling due to water influx, we performed a numerical experiment for purely conductive cooling of the pipe with no water influx and no serpentinisation. The simulation shows that for conductive cooling, the temperature drops below Tα after 1,260 years, which is actually faster than for the case of water influx. This apparent contradiction is explained by the release of the exothermic heat of reaction of serpentinisation that leads to slower cooling of the kimberlite and a prolonged interval of serpentinisation. We will discuss the relative importance of different processes on the energy budget later.
Duration of serpentinization of the kimberlitic pipe after emplacement (years)
Host rock permeability, KB and KC (×10−15 m2)
Kimberlite permeability KA (×10−15 m2)
The numerical simulations demonstrate that ground water flows easily and rapidly into a pipe infilled with hot volcaniclastic deposits. External fluids are drawn in, are heated, and ascend upwards in the centre of the pipe. The models abstract the alteration process by considering serpentisation of olivine as the dominant reaction. The reactions lead to a decrease in porosity and permeability. Because the serpentinisation reaction is typically fast, in comparison with the cooling time scale of the pipe infill, there is a sharp reaction front, dividing into regions with olivine and regions with abundant serpentine at any one time (Fig. 4).
Cooling time scales and water–rock ratios
Cooling time scales of centuries to a few millennia are calculated for the olivine fraction y = 0.6 (Table 2). Serpentisation reduces porosity and permeability leading to suppression of fluid flow. Thus, the flow regime is not like a typical hydrothermal system, where large amounts of fluid circulate freely and greatly reduce cooling times compared to pure conduction. Using our model with constant porosity and no hydration reactions, the pipe cools rapidly (~200 years to reach Tα) through conductive and convective cooling. With the hydration reactions, however, cooling time increases by a factor of 10 (see Fig. 5 caption for details) and can exceed the cooling time by pure conduction (1,310 years) due to the counteracting effect of heat generated by the reaction.
We have calculated the total amount of water drawn into and through the pipe. For the parameters in our representative case (Fig. 4), we integrated the flux of water over the side surface of the pipe from time zero to final cooling time to Ta. The total mass of water then was divided by the total mass of rock in the pipe at t = 0 (later the total mass of the rock changes because olivine gains some water during serpentinisation). The water-to-rock mass ratio is 0.098, and water/olivine ratio is 0.164. Since 73 % of this water is consumed in the serpentisation reaction, there are rather modest amounts of water which flow completely though the system. In contrast, very high water rock ratios are characteristic of major geothermal convective systems that are driven by magmatic heat from large magma reservoirs.
Kimberlite geology reflects complex multiphase eruptions and depositional events. A wide variety of lithofacies have been recognised, including massive and layered volcaniclastic deposits with diverse genesis and characteristics (Scott Smith 2008; Sparks 2013). Our models show that porosity, permeability, temperature, and olivine content are the main controls on serpentisation. Our results indicate that olivine content, porosity, and permeability of the kimberlite are, in general, more important than the hydraulic properties of the host rock to the extent of serpentinisation, due to the large volume increase of the products of the reaction. Temperature governs the kinetics of the reaction, and our model is specifically for kimberlite deposits emplaced at sufficiently high temperatures for fast reactions. The preservation of fresh olivine in some Fort à la Corne volcaniclastic kimberlites (e.g. Pittari et al. 2008; Scott Smith 2008) likely reflects either their low-temperature emplacement (e.g. <100 °C) during phreatomagmatic eruptions and reworking in marine environments or local lack of water access to parts of the volcaniclastic kimberlites.
Common types of pyroclastic deposits identified in kimberlites include pyroclastic flows, surges, and fallout, as well as massive deposits attributed to gas fluidisation. Breccias, lapilli tuffs, and tuffs indicate a wide range of possible porosities and permeabilities within the pipes, noting that young (non-kimberlitic) pyroclastic deposits typically have porosities in the range 0.2–0.5. Fine-grained, poorly sorted deposits are expected to have lower porosity and permeability, while coarser-grained well-sorted deposits have higher porosity and permeability. In addition, deposits will have different emplacement temperatures influenced by eruption style and abundance of included cold country rock lithic fragments. Facies architecture and the contacts between facies can also influence the pathways of migrating fluids and will therefore affect the progression of serpentinisation throughout the pipe. Even small-scale features such as bedding may influence the fluid pathways and cause local fluctuations in alteration intensity. The models suggest that pervasive serpentinisation is predicted over a wide range of these key parameters. These predictions are, therefore, consistent with observations that many kimberlites are strongly altered volcaniclastic deposits that had high initial porosities.
Although there is considerable overlap, MK has the highest modal abundance of olivine macrocrysts of the three lithofacies and has a higher average olivine/olivine + serpentine ratio than the other deposit types (Fig. 10; 0.71 MK; 0.54 FBK; 0.25 MUD). While poor sorting of the deposit and high original olivine content results in slow water infiltration, complete infilling of the pores with serpentine arrests the serpentinisation process and leads to preservation of fresh olivine. The most serpentinised portions of the MK occur near the pipe walls and in proximity to the contact with the FBK, which are the areas of likely fluid influx into the deposit. The FBK deposits show increased abundance of serpentine in the proximity of the pipe wall and MUD contact in the north of the pipe, and along the contact with the MK in the centre of the pipe, which is consistent with fluid migration along internal boundaries as well as external contacts. These observations are consistent with our expectations that higher porosity and permeability rocks are more susceptible to serpentinisation, and that serpentinisation can localise at the pipe (or unit) margins. Olivine-rich rocks are more strongly serpentinised, provided there is enough porosity for the excess volume of solid products to precipitate. Each of the lithofacies shows a wide range of olivine and serpentine (Fig. 10) with a broad anti-correlation. The data are consistent with initial porosities varying between 40 % and near zero.
A decrease in serpentinisation at depth has been noted in several pipes, for example at the Fox pipe in the NWT, Canada (Porritt and Cas 2009), and Gahcho Kué pipe (Hetman et al. 2004). At Fox pipe, the F1, a massive, poorly sorted kimberlite deposit, pseudomorphed olivine grains occur with fresh glassy cores, whereas at higher levels, more intense alteration has completely pseudomorphed the olivine grains. This trend is consistent with less infiltration of water at depth where country rocks are less permeable and cooling is faster due to the pipe being narrower, as predicted in the models (Fig. 7).
The alteration model results provide insights into two related issues concerning the origin of coherent kimberlitic rocks and transitional rocks between volcaniclastic and these coherent rock types. The latter have widely been interpreted as hypabyssal in origin, but are now increasingly interpreted as densely welded, high-temperature volcaniclastic rocks (Brown et al. 2008a). Coherent kimberlite is particularly common in the root zones of kimberlites, although they can occur at any level. In many pipes, such as the Gahcho Kué pipe, NWT Canada (Hetman et al. 2004), textural variations between volcaniclastic rocks and coherent kimberlite are described and termed transitional. One of the features of the transitional volcaniclastic kimberlite is partially serpentinised olivines, while tuffisitic kimberlite contains completely pseudomorphed olivine and coherent kimberlite commonly contains abundant unaltered olivine (Hetman et al. 2004; Brown et al. 2008a; van Straaten et al. 2011). The lack of serpentisation at depth is readily interpreted as the consequence of rapid densification of high-temperature volcaniclastic kimberlite that reduces porosity and permeability preventing infiltration of water. The common occurrence of coherent or coherent-like kimberlite in root zones is attributed to factors such as late-stage influx of high-temperature kimberlite magma at the base of the pipe (Gernon et al. 2012) or welding of pyroclastic kimberlite (Brown et al. 2008a), both would have low porosities and permeabilities and, hence, result in the sluggish infiltration of external water.
Source of fluids
Another controversy is the origin of the water involved in serpentisation. The opposing viewpoints of a magmatic fluid and external (ground water sources) are discussed by Mitchell (2013), Sparks (2013), and Giuliani et al. (2014) and are not repeated here. Mitchell (2013) presents ion probe data for δ18O in serpentines which have typically positive values in the range >0 to +6.3 with most values exceeding +3. He infers that these data preclude involvement of meteoric water with typically negative δ18O. Part of Mitchell’s argument is that water/serpentine ratios cannot have exceeded 0.8 and “that there was not a significant volume of low temperature water as a cause of serpentisation”. We agree with his conclusions about low water–rock ratios, which are predicted by our model (≪0.8). However, water influx slows and then ceases during serpentinisation due to the reduction in porosity and permeability. Just enough water is used up to convert the olivine to serpentine. Water/serpentine ratios are consequently low and result in small changes of δ18O.
Depending on the reaction, the change is between 10 and 27 % from the initial olivine value towards the water value and depends on whether oxygen in the original olivine is immobile or there is exchange between the reaction products. For example, in reaction 1, using δ18O = −6 for meteoric water and δ18O = +6 for olivine, the serpentine will have δ18O = +4.8 in the immobile case and δ18O = +2.8 with exchange. Furthermore, an initially light δ18O fluid that exchanges with the kimberlite matrix will have its own oxygen isotope composition modified, to become increasingly positive, due to the high proportion of matrix relative to circulating fluid. Accordingly, the shift from mantle δ18O may be even smaller or even negligible depending on the evolution of the fluid. This is consistent with the observations, as the majority of the oxygen in the serpentine originates from the olivine.
There is evidence for external fluids infiltrating into kimberlites in the Yakutian kimberlite province, Russia, where the kimberlites were emplaced into a thick sequence of sedimentary rocks, including carbonates and evaporites, and are associated with brines (Kopylova et al. 2013). The Mir kimberlite is largely serpentinised, but contains groundmass halite and shortite, which may reflect the salty character of the ground waters in the host rock limestones, marls, and dolomites. In the case of the International’naya kimberlite, the kimberlite is elevated in Na and S where there are halite and anhydrite and gypsum beds, respectively, adjacent to the kimberlite. In these regions, the olivine is fresh. The Yakutian kimberlite province includes the unusual Udachnaya kimberlite, where fresh olivines are preserved in volcaniclastic kimberlite with a groundmass containing abundant halides and alkali carbonates (Kamenetsky et al. 2004). This kimberlite has avoided serpentinisation. Kopylova et al. (2013) observed that: “The localization of the highest abundances of Na–K–Cl–S-bearing minerals in the Udachnaya East kimberlite at a depth interval that correlates across three magmatic phases of kimberlites and coincides with the roof of the halite-bearing country rock and an aquifer carrying anomalously Na-rich brines”. Kopylova et al. (2013) also present stable isotope data consistent with interactions between kimberlite and host rocks and brines. The striking feature of these three cases is that the mineralogical characteristics of the matrix of the kimberlites vary in broadly horizontal zones that coincide with wall rocks and different composition brines. In the case of Udachanaya pipe, the matrix mineral assemblages span three geological units emplaced at different times. The localisation of abundant alkali carbonate and halide minerals within the kimberlite matrices adjacent to brine aquifers and carbonate host rocks can be explained by groundwater infiltrating into the kimberlite.
Brine alteration might be caused by post-emplacement low-temperature infiltration of regional brines, but this hypothesis is not supported by observations of high-temperature olivine inclusions. Kamenetsky et al. (2004) documented the presence of sodalite and inclusions in olivine containing halide-carbonate mineral assemblages, which equilibrated at temperatures of well above 400 °C. They attribute the halides and alkali carbonates to residual melts and magmatic fluids to the kimberlite magma. However, the model of infiltration of brine into hot kimberlitic volcaniclastic deposits can also explain these observations and is analogous with our model of water infiltration discussed above. Kopylova et al. (2013) thought that these temperatures were too high to support alteration. However, brines can infiltrate and react with kimberlite at any temperature at or below the emplacement temperature of the kimberlite. Indeed, at temperatures of 600 °C, it is possible to envisage heated infiltrating brines reacting with kimberlite, the formation of high-temperature melts of natrocarbonatite, and brine forming in pore spaces and then crystallizing out assemblages with igneous textures. Brines saturated in CO2 become acidic and are strongly reactive with calcite (Lamy-Chappuis et al. 2013), and dissolution of original igneous calcite in the kimberlite might be an additional factor with creation of secondary porosity. If these processes occur above 400 °C and infill pore space, then water is prevented from accessing the kimberlite after the temperature declines and, thus, explains preservation of fresh olivine. Although physical incorporation and perhaps assimilation of carbonates and evaporites into kimberlite melts prior to emplacement likely played a role in the characteristics of the Yukatian kimberlites (Kopylova et al. 2013), these processes are not mutually exclusive. High-temperature late fluids exsolved from a magma contaminated with evaporates and carbonates may be difficult to distinguish from fluids generated by infiltration and heating of brines to high temperatures.
Model assumptions and caveats
The model is necessarily a simplification of what might happen in natural systems, so we briefly consider relaxing some of the assumptions. Any model of a geological process has to make simplifying assumptions to be tractable, and simplified models have the benefit of being easier to interpret than more complex models, which attempt to include many possible effects. We briefly comment qualitatively on relaxing some of the assumptions in the model.
Reactions among hydrothermal fluids and kimberlite constituents involve more than olivine and serpentine. Although some reactions, such as the formation of magnetite from the Fe component of olivine, may reduce the solid volume, the majority of products are hydrated minerals, such as talc, chlorite, and smectite. The products of more complex reactions will not change the model results. Multiple generations of different serpentine polymorphs are also a consequence of changing temperature and conditions in the circulation of fluids as the pipe fill cools. Mitchell (2013) draws attention to the sequence of serpentine formation with decreasing temperature in serpentine veins in oceanic peridotites (Andreani et al. 2007) and remarks that this is the same sequence as observed in kimberlites. We concur, but draw a different inference that this similarity is consistent with the same process in both cases, namely progressive alteration from external water. If swelling clays form at low temperature, the kimberlite can become a self-sealing system resistant to further infiltration.
There will be porosity variations within the pipe reflecting geological complexities. Lower porosity, finer-grained volcanic, and sedimentary deposits resides in the uppermost crater zones and forms a low permeability cap. In this case, the internal pressure in the pipe might be expected to become high and so ground water flux rates into the pipe would be slower. However, this should make a major difference. The model predicts that the process is largely the pipe absorbing water from its surrounding and the flow through is modest. In the case where water flows in but cannot flow out, the system will move to a balance of the hydrostatic pressure inside and outside of the pipe. However, this balance can only be achieved for water when the temperatures inside and outside of the pipe are the same. Thus, we anticipate an early stage dominated by inward flow of water and a later convective stage when the main driving force comes from the difference in temperature between water inside and outside of the pipe. This later stage should involve outward flows of hot fluid as well as inward flows of cold water. Country rocks might be altered in the outflow areas.
We have assumed a closed system with only local mobility of Si and Mg to facilitate the serpentinisation reaction. However, high pH fluids emerging from ocean flow hot springs associated with serpentisation (e.g. Kelley et al. 2007) can carry high concentrations of major elements such as Ca2+ and Mg2+. Solute transport will lead to complexities, including re-distribution and quantitative removal of major components. Models that include such complexities are avenues of research, but they should not introduce any fundamental new aspect of the process. Likewise, geological complexities, such as local aquifers and faults, could be included in site-specific modelling studies.
We have only considered conversion of olivine to serpentine. An unmodelled effect, however, is dissolution of minerals. Primary igneous carbonate (typically calcite) has been found in many hypabyssal kimberlites, and models of kimberlite magmas suggest carbonate-rich residual melts (Russell et al. 2012). Textural evidence (Sparks et al. 2009) indicates that calcite and other Ca-rich phases such as apatite, in hypabyssal kimberlite, are dissolved in the presence of serpentinising fluid, an observation consistent with the high Ca content of fluids generated during ocean ridge serpentisation (Kelley et al. 2007). However, primary carbonate is typically absent in pipe-filling volcaniclastic kimberlites, which may be explained by dissolution during serpentisation. Dissolution of calcite and other Ca-rich minerals creates secondary porosity and enhances the prospects for incoming fluids to serpentinise olivine.
Our model considers only high-temperature kimberlite deposits. However, some volcaniclastic kimberlites were likely emplaced cold resulting in a much reduced rate and degree of serpentinisation. Involvement of water in the eruption, either from surface water or ground water from high permeability country rocks, or deposition of the volcaniclastic kimberlite onto the surface outside of the pipe (and subsequent cold resedimentation into the pipe) both result in low-temperature deposits within the pipes. Low temperature may be the reason that olivine is typically well preserved in many of the Fort a la Corne kimberlites, which are attributed to phreatomagmatic eruptions in a marine environment (Pittari et al. 2008).
Our model shows that the timescales of cooling of a kimberlite pipe, from emplacement temperatures through the serpentinisation temperature window to Te, and the timescale of infiltration of external water, which drives the serpentinisation reaction, are comparable. The rate of infiltration of water into the pipe, and consequently the ability of water to react with the kimberlite before it cools below the serpentinisation window, is greatly influenced by the deposit porosity, permeability, and olivine content. Deposits with high initial porosity, permeability, and olivine content are more susceptible to serpentinisation, with olivines being pseudomorphed and the excess volume of serpentine formed during the reaction infilling the pore space. High degrees of serpentinisation along initial fluid pathways, such as pipe margins or internal facies contacts, may occlude porosity preventing further ingress of water, thus preserving fresh olivine in the pipe interiors. Post-emplacement serpentinisation of volcanic deposits within a kimberlite pipe therefore can result from the rapid infiltration of external water and reaction with olivine and other kimberlite components.
A. Afanasyev acknowledges financial support by a grant from the President of the Russian Federation for the Support of the Young Scientists (Project No. SP-2222.2012.5), and O. Melnik support from the Russian Foundation for Basic Research (RFBR-RS 12-01-92601). Sparks and Melnik were supported through funding from the European Research Council (VOLDIES Project). Porritt were supported by a Marie Curie Fellowship. Reviews by Maya Kopylova and an anonymous referee improved the manuscript. The simulations were conducted at the Moscow University Supercomputing Center.
- Andreani M, Mével C, Boullier AM, Escartin J (2007) Dynamic control on serpentine crystallization in veins: constraints on hydration processes in oceanic peridotites. Geochem Geophys Geosyst G3 8:1–24Google Scholar
- Brooks AN, Corey AT (1964) Hydraulic properties of porous media. In: Hydrol. Pap. Colorado State University, Fort CollinsGoogle Scholar
- Field M, Scott Smith BH (1999) Contrasting geology and near-surface emplacement of kimberlite pipes in southern Africa and Canada. In: Proceedings of 7th international kimberlite conference, vol 1, pp 214–37. Cape Town: Red RoofGoogle Scholar
- Fyfe WS, Lonsdale P (1981) Ocean floor hydrothermal activity. The sea. In: Emiliani C (ed) The oceanic lithosphere, vol 7, pp 589–638. Wiley, New YorkGoogle Scholar
- Gernon T, Brown RJ, Tait MJ, Hincks T (2012) The origin of pelletal lapilli in explosive kimberlite eruptions. Nat Commun 3(831):1–7Google Scholar
- Hawthrone JB (1975) Model of a kimberlite pipe. Phys Chem Earth 1:1–16Google Scholar
- Landau LD, Lifshitz EM (1951) Course of theoretical physics, vol. 5: statistical physics: part 1. Butterworth-Heinemann, OxfordGoogle Scholar
- Mitchell RH (2013) Paragenesis and oxygen isotopic studies of serpentine in kimberlite. In: Pearson DG et al (eds) Proceedings of the 10th kimberlite conference, vol 1, special issue of J Geol Soc India, pp 1–12Google Scholar
- Pittari A, Cas RAF, Lefebvre N, Robey J, Kurszlaukis S, Webb K (2008) Eruption processes and facies architecture of the Orion Central kimberlite volcanic complex, Fort a la Corne, Saskatchewan; kimberlite mass flow deposits in a sedimentary basin. J Volcanol Geothermal Res 174:152–170CrossRefGoogle Scholar
- Robie RA, Hemingway BS, Fisher JR (1979) Thermodynamic properties of minerals and related substances at 298.15 K and 1 bar (105 Pascals) pressure and at higher temperatures. US Government Printing Office, Washington, DCGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.