A rare occurrence of a crater-filling clastogenic extrusive coherent kimberlite, Victor Northwest (Ontario, Canada)

Transcription

1 DOI /s RESEARCH ARTICLE A rare occurrence of a crater-filling clastogenic extrusive coherent kimberlite, Victor Northwest (Ontario, Canada) Bram I. van Straaten & M. G. Kopylova & J. K. Russell & B. H. Scott Smith Received: 10 March 2010 /Accepted: 20 May 2011 # Springer-Verlag 2011 Abstract Kimberlite pipes can contain significant proportions of dark and dense kimberlite that have mostly been interpreted as intrusive coherent (hypabyssal) in origin. This study reports a well-documented occurrence of a fresh intra-crater clastogenic extrusive coherent kimberlite that is concluded to have formed as a result of lava fountaining. This paper focuses on a dark, dense, competent, generally crystal-rich, massive kimberlite unit within the Victor Northwest kimberlite pipe (Ontario, Canada). Using a comprehensive volcanological and petrographic analysis of all available drill cores, it is shown that this unit has a fresh well-crystallised coherent groundmass and is extrusive and pyroclastic in origin. The proposed clastogenic coherent extrusive origin is based on deposit morphology, gradational contacts to enveloping pyroclastic units, as well as the presence of remnant pyroclast outlines and angular broken olivines. This paper, and an increasing number of other studies, suggest that fragmental extrusive coherent kimberlite in intra-crater settings may be more common than previously Editorial responsibility: R.A.F. Cas This paper constitutes part of a special issue: Cas RAF, Russell JK, Sparks RSJ (eds) Advances in Kimberlite Volcanology and Geology. B. I. van Straaten (*) : M. G. Kopylova : J. K. Russell : B. H. Scott Smith Department of Earth and Ocean Sciences, University of British Columbia, 6339 Stores Road, Vancouver, BC, Canada V6T 1Z4 bvanstraaten@gmail.com B. H. Scott Smith Scott-Smith Petrology Inc, 2555 Edgemont Blvd., North Vancouver, BC, Canada V7R 2 M9 thought. The emplacement history and volcanology of these pipes need to be reconsidered based on the emerging importance of this particular kimberlite facies. Keywords Kimberlite. Volcanic pipe. Extrusive coherent kimberlite. Clastogenic deposit. Coalescence. Lava fountaining Introduction Coherent kimberlite, defined here as rocks formed from the cooling and solidification of kimberlite magma or lava, commonly occurs as intrusive sheets and dykes, and is also found to fill or partly fill pipes and diatreme root zones (e.g. Field and Scott Smith 1999; Skinner and Marsh 2004). For the most part, these deposits have a well-crystallised igneous groundmass and have been interpreted as being intrusive coherent, i.e. hypabyssal, in origin. This paper describes a rarely reported occurrence of extrusive coherent kimberlite in a pipe-filling geometry at Victor Northwest, Northern Ontario, Canada. We propose that the extrusive coherent rocks formed by coalescence of (mostly) liquid pyroclasts ejected by lava fountaining. This study provides several novel insights into the eruption mechanisms of kimberlites, specifically regarding the formation of reconstituted extrusive coherent deposits. Coherent extrusive rocks can form by a variety of processes. True effusive lava flows originate from magma flowing out of fissures and vents; however, it has been suggested that most lava flows are formed by lava fountaining and preserve little to no evidence for their original fragmental origin (Sumner et al. 2005). In addition, welded fall deposits can form in vent-proximal environments given high enough eruption temperatures and accumulation rates (Walker et al. 1984; Carey et al. 2008). More explosive eruptions can also

2 produce pyroclastic flow deposits that are re-formed into coherent or lava-like rocks by dense welding (Ekren et al. 1984; Branney and Kokelaar 1992), generally aided by the overburden pressure of the overlying deposit. In some cases, coalescence, agglutination or welding can produce rocks that are coherent and may be indistinguishable from lavas on the hand sample and microscopic scale (Sumner and Branney 2002, and references therein). These lavalike pyroclastic deposits are generally believed to have formed by syn-depositional reconstitution of mostly molten magma droplets. From these examples, it is clear that there is a complete continuum between effusive lava flows, lava fountain-fed clastogenic lavas, lava-like pyroclastic deposits to non-welded pyroclastic deposits. Reconstitution of hot pyroclastic material has only recently been suggested as a possible mechanism for forming coherent extrusive kimberlite deposits (Webb et al. 2004; Russell and Moss 2006; Sparks et al. 2006; Brown et al. 2008; Nowicki et al. 2008; Crawford et al. 2009; van Straaten et al. 2009). In this paper, we identify coherent kimberlite as a rock having an igneous, well-crystallised groundmass, irrespective of any potential earlier fragmental history. In the case of Victor Northwest, abundant carbonate laths are generally present and their euhedral and interlocking texture provides strong evidence for a coherent origin. We use reconstituted coherent rock as an umbrella term to describe magmas or lavas inferred to have undergone pyroclastic fragmentation but where the particles were subsequently re-formed into coherent rocks, or made to appear coherent, by syn- to immediately post-depositional processes ranging from (1) welding, sticking together at point contacts and compaction of particles under the influence of the overburden pressure in a still-hot deposit (terminology after Sumner et al. 2005), (2) agglutination, the flattening, sticking together and gradual cooling of hot, fluid pyroclasts as they impact on an accumulation surface, with particle outlines generally partly retained (Sumner et al. 2005), to (3) coalescence, a process where fluidal magma droplets form a homogeneous liquid in which the original particle outlines are obliterated (Sumner et al. 2005). In contrast, the term clastogenic extrusive coherent rock only applies to deposits that were formed by coalescence, and are generally formed in lava fountaining environments (definition following Wolff and Sumner 2000). The range of processes that can create coherent rocks via particle reconstitution can be seen as a continuum from welding to agglutination and coalescence, with the latter favoured by a decrease in viscosity, load and/or an increase in temperature. Volcaniclastic kimberlite (VK) refers to all rocks or unconsolidated deposits that consist of volcanic fragments, irrespective of the mode of fragmentation or final deposition (Cas et al. 2008a). Hence, these deposits encompass both primary pyroclastic (pyroclastic kimberlite, PK) and resedimented volcaniclastic kimberlite (RVK) deposits. This study is based on a comprehensive volcanological and petrographic analysis of all available drill cores from Victor Northwest. This paper focuses on the extensive dark and competent kimberlite unit and shows that it is coherent and then demonstrates that it has an extrusive origin. The potential effects of alteration processes, which are commonly quoted as masking original textures in kimberlites (e.g. Stripp et al. 2006; Hayman et al. 2009), were considered. It is suggested that alteration did not have a major influence on this unit. In addition, we show that the coherent unit has a fragmental origin, and discuss which processes among non-explosive effusive activity, lava fountaining and welding may have been involved in its emplacement. Finally, we show that this unit represents a clastogenic extrusive coherent deposit most likely formed by lava fountaining. Geology of the Victor kimberlite The Victor kimberlite complex is located in Northern Ontario, Canada and is part of the Middle Late Jurassic Attawapiskat kimberlite province (Kong et al. 1999). De Beers Canada Inc. is currently mining the Victor kimberlite pipes for diamonds. The Victor diamond deposit comprises several cross-cutting and adjacent steep-sided kimberlite pipes (Webb et al. 2004; van Straaten et al. 2008, 2009), emplaced within a 275 m thick Ordovician to Silurian sedimentary sequence which overlies Precambrian granitoid basement (Fig. 1). We studied the volcanic stratigraphy and petrography of the earliest pipe in the Victor complex, the Victor Northwest pipe, using detailed investigations of drill cores, polished drill core samples and thin sections. The Victor Northwest pipe is crosscut by the later Victor Main pipe (Fig. 1; van Straaten et al. 2009). The internal geology of the Victor Northwest (VNW) kimberlite pipe is complex, and comprises numerous smallvolume stratigraphic units (Fig. 1). The intra-pipe stratigraphy was the focus of a previous study (van Straaten et al. 2009) and is summarised here. The internal geology of the VNW kimberlite pipe has been reconstructed using a total of 22 drill cores (Fig. 2; van Straaten et al. 2009) with a total cumulative length of 4.2 km. The average collar elevation of drill holes in VNW is 83.3 m above sea level. This is considered to be the average present surface elevation, and all depths are given relative to this level (e.g. 100 metres below surface is 100 m below the present surface). The 3-D distribution of the different rock types within the VNW pipe was reconstructed by extrapolating the data between the drill holes shown in Fig. 2. The stratigraphic units can be subdivided in two distinct packages, upper (U) and lower (L), each of which can be shown to correspond to a separate

3 Geology of the VNW upper strata Fig. 1 Cross-section through the Victor Northwest (VNW) pipe showing the detailed stratigraphy in the upper (U) part of the pipe. Location of cross-section is shown in Fig. 2. Blue colours represent dark and competent kimberlite (DCK; note that this rock unit forms a funnel shape), green colours are sedimentary country rock breccia (CRB), red colours are volcaniclastic kimberlite (VK), and red white striped colours are resedimented volcaniclastic kimberlite (RVK). Numbers indicate sub-units of each rock type. Pale green indicates the combined lower (L) VNW kimberlite pipe stratigraphy. Magenta shows location of the later cross-cutting Victor Main pipe. Pale yellow and pink represent the country rock succession. Depths are in metres below the surface. Lines are drill holes, tick marks on drill holes are rock type subdivisions based on logging (see Fig. 3 for details). After van Straaten et al. (2009) where further details are provided. For the colour figure, the reader is referred to the web version of this article eruptive cycle (Fig. 1; van Straaten et al. 2009). Each eruptive cycle comprised an early explosive phase that produced pyroclastic deposits containing abundant angular broken olivine grains. In this context, it is important to note that sub-round olivine grains are the dominant crystal component in pre-eruption kimberlite magmas (Mitchell 1986). The early pyroclastic deposits are overlain by dark and competent kimberlite (DCK), which is massive and characterised by mostly unbroken olivines. The last deposits of each eruptive cycle comprise (resedimented) volcaniclastic kimberlite and a thick blanket of sedimentary country rock breccias. Each eruptive cycle consisted of an early explosive eruption followed by less explosive eruptive activity, with subsequent post-eruption pipe wall collapse and production of country rock breccias (van Straaten et al. 2009). The dark and competent kimberlite within the upper VNW pipe (DCK-U) is the focus of this paper (Fig. 1). The DCK- U is overlain by pyroclastic kimberlite, volcaniclastic kimberlite, resedimented volcaniclastic kimberlite and sedimentary country rock breccias (VK-U 2, VK-U 3, RVK- U, CRB-U, respectively; see Figs. 1 and 3); and underlain by pyroclastic kimberlite (VK-U 1 ). The pertinent features of each kimberlite unit in the upper VNW pipe are summarised in Table 1. Rock units were discriminated by their difference in country rock fragment abundance, proportion of angular broken olivine grains, and degree of olivine alteration (illustrated in Fig. 4 for VK-U 1, DCK-U and VK-U 3 ). The relatively small-volume unit VK-U 2 is located in the pipe centre (Figs. 1 and 3; van Straaten et al. 2009) and is characterised by low country rock fragment abundances ( 6 vol.%), altered olivines and mostly intact olivine macrocrysts. In addition to the features shown in Fig. 4, the juvenile pyroclasts can also be used to differentiate among units within VNW. The juvenile pyroclast shape, size and constituents within the pyroclasts show distinct textural patterns between VK-U 1, VK-U 2 and VK-U 3 (Table 1). The VK-U 1 is most distinct from other units, and shows predominantly small, generally uncored, irregularly shaped, variably vesicular juvenile pyroclasts and rarer rim-type accretionary pyroclasts (sensu Schumacher and Schmincke 1991). In contrast, VK-U 2 and VK-U 3 show mostly nonvesicular juvenile pyroclasts cored by olivine macrocrysts; VK-U 3 also contains rare >5 cm uncored poorly vesicular juvenile pyroclasts. The CRB-U comprises 10 cm to at least 1 m sized sedimentary country rock fragments of limestones, dolostones, mudstones and siltstones, and is interpreted to have formed by pipe wall collapse following the eruption (van Straaten et al. 2009). The RVK-U is only intersected in one drill core (V , Fig. 3) and comprises a thin intersection of well-sorted country rock fragment-poor massive to layered RVK containing abundant foreign (quartz, mud) material. This unit shows clear evidence for addition of non-kimberlitic detrital material and resedimentation by water (van Straaten et al. 2009). The three major volcaniclastic kimberlite units have been interpreted as pyroclastic (VK-U 1, VK-U 2 ) and volcaniclastic (VK-U 3 ) in origin (van Straaten et al. 2009). The pyroclastic nature of VK-U 1 and VK-U 2 is inferred from the presence of fragile clastic aggregate pyroclasts (VK-U 1 only), the absence of significant sorting and bedding that is typical of reworked deposits, and the absence of detrital non-volcanic material. In contrast, the final mode of deposition of VK-U 3 is suspected to have been reworking of tephra-ring material based on the gradual transition between RVK-U and the overlying pipe wall collapse CRB-U (van Straaten et al. 2009).

4 V Bull Volcanol V c 0 A V e > m V V c V V m VNW 95 V V c >72 V c V V c 93 V e >43 <123 V c 118 V c > m V c V < m <136 <208 A V c V VM V V V c VM N Fig. 2 Plan view of the Victor Northwest (VNW) kimberlite pipe showing contours for the base of the VNW kimberlite unit DCK-U (contour interval 20 m, depth in metres below the surface). VM indicates the later Victor Main pipe. Thick solid line outlines the VNW and VM kimberlite bodies, dashed line represents the internal contact between the two pipes. Locations of 22 drill holes that intersect VNW are shown. Open circles are drill hole collar locations (multiplication symbol inside collar location circle indicates drill cores that do not intersect the upper VNW strata); straight black lines are surface traces of angled drill holes. Lithologies in the 18 drill holes intersecting the VNW upper strata are shown in Fig. 3. Black star and number indicate the depth of the basal contact of the DCK-U, grey star and number indicate that the contact is not intersected but likely occurs below (>) or above (<) the indicated depth. Cross-section A-A is shown in Fig. 1. For the colour figure, the reader is referred to the web version of this article Despite the abundant textural differences among the different kimberlite units in the upper VNW pipe (VK-U 3, VK-U 2, DCK-U 2, DCK-U 1 and VK-U 1 ), there are a number of striking similarities. Firstly, the absolute abundance, relative proportions and grain sizes of mantle-derived indicator minerals (MIMs) are very similar among units (Fig. 5). Changes in absolute abundance can be explained by variations in the proportion of country rock fragments (higher in the samples of VK with respect to DCK-U). In addition, variations in the absolute abundance of MIMs reflect changes in the total macrocryst (mantle-derived coarse crystal) content by magmatic (filter pressing, flow) and volcanic (fragmentation, deposition) processes. Secondly, the five kimberlite units of VNW also are very similar in their mineralogical and petrological characteristics. The groundmass of juvenile pyroclast in the volcaniclastic units, as well as the well-crystallised interlocking groundmass in the dark and competent units (see Geology of the DCK-U section) is very similar in terms of mineralogy, textural relationships among minerals, individual crystal morphology, individual mineral modal abundance and individual mineral size ( Geology of the DCK-U section and van Straaten et al. 2009). All units comprise, apart from macrocrysts and olivine phenocrysts, a groundmass of (in order of modal abundance, see Geology of the DCK-U section): carbonate laths, spinel, monticellite, perovskite, minor phlogopite and micro- to cryptocrystalline carbonate±serpentine. Geology of the DCK-U The upper dark and competent kimberlite in the VNW pipe (DCK-U in Fig. 1 and van Straaten et al. 2009) was investigated using detailed megascopic, macroscopic and microscopic observations on drill cores, polished drill core samples and thin sections prepared from 77 representative drill core samples (see Fig. 3 for sample locations). In this study, 16 of 17 available drill cores containing this rock unit were logged (Figs. 2 and 3; the total intersected true core length of this rock unit is 805 m). The results of this study are summarised in Figs. 1, 2, 3, 4 and 5 and the representative samples are illustrated in Figs. 6, 7, 8 and 9. Geometry and stratigraphic framework The 3-D distribution of the DCK-U is shown in Figs. 1, 2 and 3. Figure 2 presents the reconstructed contours for the base of the DCK-U. The data presented in Figs. 1, 2 and 3 shows

5 Fig. 3 Distribution of lithologies in all 18 drill cores that intersect the Victor Northwest (VNW) upper strata, grouped into different areas of VNW. The funnel-shaped dark and competent kimberlite (DCK-U 1 and -U 2 ) is underlain by volcaniclastic kimberlite (VK-U 1 ) and overlain by volcaniclastic kimberlites (VK-U 2, VK-U 3 ), resedimented volcaniclastic kimberlite (RVK-U) and country rock breccia (CRB-U). Note that the Victor Main pipe (VM-VK) cross-cuts the units within the earlier VNW pipe. OVB is unconsolidated overburden, CRS is in situ sedimentary country rock, mbs is metres below surface. See Fig. 2 for drill hole locations. Tick marks on side of drill holes are locations of representative petrographic samples, open circle tick marks are samples shown in Figs. 6 and/or 7, solid squares are samples illustrated in Fig. 8, open square is sample shown in Fig. 9. For the colour figure, the reader is referred to the web version of this article the pipe-wide extent of the DCK-U, the funnel-shaped geometry, and the fact that this unit is the volumetrically dominant unit in the upper VNW pipe fill. The average vertical thickness of the unit varies from 45 m to at least 95 m near the reconstructed pipe centre. Twenty-three out of 25 internal contacts between adjacent kimberlite units (VK-U 3 to DCK-U 2, VK-U 2 to DCK-U 2 and DCK-U to VK-U 1 ) are not sharp (Fig. 3). Generally, the contacts between the DCK-U and the enveloping extrusive and fragmental deposits (see Geology of the VNW upper strata section) are characterised by a gradational change in componentry over tens of metres including country rock fragment type and abundance, pyroclast abundance and type and broken olivine content (Figs. 3 and 4). Features such as the olivine alteration change over 3 10 m from within the DCK-U across the contact to surrounding units (Fig. 4b). These contact relationships contrast markedly with other contacts elsewhere in Victor that show knife-sharp erosional or cross-cutting contacts with abrupt changes in componentry within 1 10 mm (e.g. Victor Main to VNW pipe contact, see Fig. 3 and van Straaten et al. 2009). Macroscopic and microscopic features Detailed petrographic observations on drill core and representative samples show that the DCK-U unit has a dark, dense, competent, generally matrix-supported and massive appearance and macroscopic texture. This unit contains

6 Table 1 Summary of significant characteristics of all units in the VNW upper stratigraphy (+ present, ± sparsely present, not present) Unit CRF Textures Pyroclasts CRF (vol.%) Massive Layered Clast supported Matrix supported Broken olivine Faint Juvenile (cored) Juvenile (uncored) Clastic CRB-U RVK-U 8 ± + + VK-U , rare 100 ± + + ± + ± VK-U ± + ± + DCK-U ± DCK-U ± ± + ± ± VK-U ± ± + ± abundant mm sub-round mostly unbroken olivine macrocysts (Figs. 4c, 6 and 9) and minor macrocrystic phlogopite. The olivine macrocrysts are generally partly fresh, which contrasts distinctly with the complete replacement of olivine by carbonate±serpentine in the overlying and underlying volcaniclastic kimberlite units (Fig. 4b). Mantle-derived indicator minerals are rare with an average of approximately three grains per 40 cm 2 of core surface (Fig. 5). The MIMs are dominated by ilmenite, garnet (most commonly purple) and minor Crdiopside (mainly occurring as intergrowths with some of the larger olivine macrocrysts). The DCK-U typically has low abundances of country rock fragments (Fig. 4a); approximately 1 2 vol.% basement clasts (granitoid and dolerite) plus 1 8 vol.% sedimentary clasts (white grey limestone and green mud/siltstone). The sedimentary country rock fragment abundance increases towards the upper and lower contacts (Fig. 4a). There are some differences in preferred orientation, abundance, packing density and distribution of olivine macrocrysts and/or country rock fragments throughout the DCK-U. Based on these criteria the unit can be divided into two subunits, a lower DCK-U 1 and volumetrically dominant upper DCK- U 2 (Figs. 1 and 3; DCK-U 2 and DCK-U 1 sections). The area between the macrocrysts in the DCK-U is generally composed of the following euhedral subhedral minerals: olivine phenocrysts set in a groundmass of evenly distributed oxides (spinel), perovskite, carbonate laths, phlogopite laths, and μm equant six-sided serpentine pseudomorphed monticellite (Fig. 7). The carbonate laths are often abundant ( 15%), and commonly form an interlocking texture. Some of the laths have been pseudomorphed by a later carbonate, as laths occurring in clusters extinguish together (Fig. 7b). In addition to the various minerals listed above, variable proportions (0 15%) of anhedral micro- to cryptocrystalline carbonate and/or serpentine of uncertain origin are present throughout the groundmass. The latter minerals also locally occur as distinct mm sized pool- or segregation-shaped areas. The DCK-U groundmass can be subdivided into three types (types 1 3; Fig. 3) based mainly on carbonate textures observed in thin section (Fig. 7). The groundmass types indicated in Fig. 3 between petrographic samples are either extrapolated (up to 5 m on either side of the sample), or are based on visual observations of drill core during logging using a binocular microscope (in rare cases, the carbonate laths were visible with the naked eye). Type 1 groundmass dominates the DCK-U ( 50%, see Fig. 3) and comprises well-formed (>100 μm) interlocking crystals, most commonly euhedral carbonate laths (Fig. 7a, b). Types 2 and 3 groundmass (Fig. 7c, d) contain more abundant ( 5 15% and 15 30%, respectively) interstitial micro- to cryptocrystalline carbonate and/or serpentine, and lack readily discernible (type 2) or completely lack (type 3) >100 μm interlocking crystals typical of type 1. Types 2 and 3 groundmass occur in approximately 35% and 3% of the DCK-U intersections, respectively. It should be emphasised that the distribution of all minerals, except carbonate and serpentine, is similar for the types 1, 2 and 3 groundmasses. All three types of groundmass occur in both subunits of DCK-U described below ( DCK-U 2 and DCK-U 1 sections), and there is no relationship between groundmass type and macroscopic textural appearance. In approximately 13% of the intersected DCK-U, the groundmass type could not be established, generally because of a lack of available samples. At the macroscopic scale, a gradational change in textures and componentry across the contacts of DCK-U with the adjacent VK units is observed over tens of metres of drill core ( Geometry and stratigraphic framework section, Fig. 4). These gradual changes cannot be verified at the microscale because of the limited spatial distribution of samples. However, based on gradually changing macroscopic features, a sharp change in microscopic texture from

7 0 CRF abundance (%) Olivine alteration index Angular olivine (%) 0 60 a b c VK-U 3 11 Normalised depth (m) Fresh Altered Typical range for intrusive dykes DCK-U 65 VK-U 1 86 Error bar Fig. 4 Graphs of three petrographic properties plotted by depth that distinguish the dark competent (DCK-U) from the adjacent volcaniclastic kimberlite units. a Modal abundance of country rock fragments, b amount of fresh olivine, and c proportion of olivine macrocrysts that have an angular shape. A total of six to 13 drill holes are shown by different data points or lines at the same normalised depth (average thickness of VK-U 3 =11 m, DCK-U=54 m, VK-U 1 =21 m). Dashed line is boundary between rock units. a Country rock fragment (CRF) abundance is based on a visual estimate in drill core, b amount of fresh olivine is based on a visual estimate of the proportion of fresh remnants within olivine macrocrysts and phenocrysts as observed in thin section, and c proportion of angular olivine was established by counting the number of olivines with one or more sharp angle (<80 ) out of at least 25 olivine macrocrysts in thin section and is interpreted to represent the proportion of angular broken grains. The alteration index in b ranges from 0 to 6 and is defined as follows: 0, completely fresh olivines of all sizes; 1 only 75% of the smallest (<200 μm) olivine phenocrysts are altered, no altered olivine macrocrysts; 2 75% of olivine phenocrysts altered, no altered olivine macrocryst cores; 3 all olivine phenocrysts and 25% of olivine macrocryst cores are altered; 4 all olivine phenocrysts (<500 μm) altered, 75% of olivine macrocryst cores altered; 5 all olivines altered, no chlorite/vermiform antigorite; 6 all olivines altered and chlorite and/or vermiform antigorite present. In c, solid triangles are samples from the contact between the DCK-U (diamonds) and VK (solid squares); solid diamonds indicate DCK-U 1 and open diamonds represent DCK-U 2 samples. Grey shading adjacent to the left-hand axis summarises the results of point counting two typical macrocrystic kimberlite intrusive coherent dykes (Snap Lake and Diavik lower A154N pipe, NWT, Canada; see Field et al and Moss 2009 respectively, for location) with values for angular olivine ranging from 0% to 9%. For the colour figure, the reader is referred to the web version of this article type 1, 2 or 3 groundmass and clearly volcaniclastic intervals is extremely unlikely. DCK-U 2 The DCK-U 2, which comprises an estimated 80% of the total volume of the DCK-U, is fairly homogenous (Fig. 6), displaying only slight variations in preferred orientation, abundance, packing density and distribution of grains and/ or clasts. About 35 40% of the observed DCK-U 2 contains macroscopically discernible generally darker selvages around some olivine macrocrysts. The selvages are most common within the upper N NE parts and in the central to lower parts of the central areas of the DCK-U 2 unit (Fig. 3). The selvages are 100 μm to 1 mm thick, ranging from complete symmetrical (Fig. 8a) to partial and asymmetrical (Fig. 8b, c). The selvages are different from the surrounding groundmass, as they contain notably finer-grained olivine phenocrysts (maximum dimensions of μm compared to 1 2 mm outside the selvages) and carbonate laths (up to compared to μm outside the selvages). The boundary between the selvage and surrounding groundmass is generally diffuse (i.e. not sharp). In numerous cases, the surrounding coarser groundmass contains well-formed interlocking crystals (type 1 groundmass in Figs. 3 and 8b). The selvages are similar in texture to juvenile pyroclasts in the overlying pyroclastic and volcaniclastic units (VK-U 2, VK-U 3 ; see Geology of the VNW upper strata section, and van Straaten et al. 2009). Other

8 VK-U 3 VK-U 2 DCK-U 2 DCK-U 1 VK-U MIM abundance normalised to 40 cm 2 slab ~2 1-4 significant features within the DCK-U 2 include minor areas that show patchy crystal size variations, as well as the presence of a generally subhorizontal preferred orientation of elongate grains (e.g. Fig. 6b, c). One drill ~1 ~3 1-4 one grain: Ilm Purple Grt Red Grt Orange Grt Cr-Di Fig. 5 Bar graph showing similar abundances for mantle-derived indicator minerals (MIMs) for all five kimberlite units within the upper VNW pipe. The graph shows the abundance of ilmenite (Ilm), purple/ pink peridotitic garnet (purple Grt), red garnet (red Grt), orange/yellow eclogitic garnet (orange Grt), and Cr-diopside (Cr-Di) normalised to a surface area of 40 cm 2 of polished slab. Average abundances vary from 1.9 to 3.2 total MIMs per 40 cm 2 slab. The inset number in each bar refers to the typical grain size (in mm) of the ilmenite, garnet and Crdiopside. For each unit, the MIMs 0.5 mm in size were counted in seven polished core samples with the aid of a binocular microscope. The results were normalised to an area of 40 cm 2. The total area counted for each unit was cm 2 (total of 3,300 cm 2 for all units combined); and in all five units combined, a total of 200 grains were identified. Note that the red garnets have a very low abundance; only two grains per cm 2 were counted in the upper two units, one grain per 700 cm 2 was counted in the DCK-U units, and no grains were found in the 600 cm 2 of the lower unit (approximate bar width representing one counted grain is show on top right of graph). The data are not corrected for variable olivine macrocryst abundance reflecting, for example, sorting processes and/or variable dilution by country rock fragments. The latter complicating factor was minimised by selecting samples with low dilution ( 0 15% CRF) ~2 ~1 core contains a vesicular interval associated with type 1 groundmass (V ; Fig. 3). DCK-U 1 The DCK-U 1 forms an estimated 20% of the total volume of the DCK-U and occurs below DCK-U 2 in 11 out of 14 drill holes that intersect the lower DCK-U contact (Fig. 3). Where present, the DCK-U 1 has a (projected) vertical thickness of 5 30 m. This unit contains a higher proportion of angular broken olivine crystals than does DCK-U 2 (Fig. 4c). DCK-U 1 includes three slightly different textural types: types a, b and c, which are shown with different cross-hatching patterns in Fig. 3. Type a generally contains slightly finer-grained olivine macrocrysts compared to DCK-U 2. Type b, which also contains finer grained olivines (similar to type a), commonly displays a heterogeneous and patchy distribution of grains. Type c has domains supported by clasts (>0.5 mm), or by matrix (grains<0.5 mm), possibly representing graded bedding (Fig. 9). It is important to note that the textures typical of DCK-U 1 types a to b to c become increasingly different from the overlying DCK-U 2 and more comparable to the underlying VK-U 1 suggesting a gradational change in rock type. The DCK-U 1 tends to be absent near the centre of the VNW pipe, and DCK-U 1 types b and c are generally more common near the outer portions of the pipe (Fig. 3). Interpretation and discussion VNW upper strata: deposits from a single batch of magma Similarities in mantle indicator mineral abundance (Fig. 5) as well as groundmass mineralogy ( Geology of Fig. 6 Macroscopic characteristics of DCK-U 2 illustrated using photographs of representative polished core samples from different parts of the pipe (see Fig. 3 for sample locations). Scale bar is 1 cm. DCK-U 2 contains abundant intact olivine macrocysts (Ol), minor country rock fragments (CRF, predominantly white limestone) and rare garnet macrocrysts (Grt). The overall uniform macroscopic constituents are set in a dark coloured fine-grained groundmass shown in thin section in Fig. 7. Note sub-parallel fabric of elongate grains in samples b and c

9 Fig. 7 Microscopic characteristics of DCK-U 2 illustrated using photomicrographs of representative thin sections (see Fig. 3 for sample locations). Scale bar is 500 μm. Plane polarised light in a, c and d; crossed polarised light in b. Typical type 1 groundmass (a, b) has abundant evenly distributed olivine phenocrysts (Ol) set in a wellcrystallised groundmass containing the following interlocking finegrained minerals: black oxides (spinel), perovskite, carbonate laths (L), phlogopite laths (Phl) and monticellite pseudomorphs (Mtc). The clustering of carbonate laths is shown in b. Typical type 2 and 3 groundmass with more common interstitial microcrystalline carbonate (M) domains are shown in c and d, respectively the VNW upper strata section) suggest that all units within the upper VNW pipe are derived from a single batch of magma. If so, this has important economic implications because each batch of kimberlite magma typically carries a distinct mantle-derived crystal cargo and, where present, a particular macrodiamond population (e.g. Gurney and Kirkley 1996; Scott Smith and Smith 2009). No data on mantle indicator mineral chemistry is available for the different units in the upper VNW pipe to further test this hypothesis (van Straaten 2010). Any differences in diamond grade and quality between the different kimberlite units in the upper VNW pipe could be caused by volcanic processes that cause dilution by country rock fragments, sorting and crystal breakage, as well as chemical processes such as potentially prolonged diamond resorption (e.g. Fedortchouk et al. 2005) in slowly cooling coherent deposits. DCK-U: textural genetic interpretation Single unit of well-crystallised coherent kimberlite Results presented in the Geology of the DCK-U section show that interpreting the DCK-U units is not straightforward. The units that surround the DCK-U are clearly volcaniclastic, as shown by the clast-supported textures, heterogeneous clast distribution (e.g. graded bedding) and the presence of juvenile pyroclasts and/or clastic aggregate pyroclasts. In contrast, the DCK-U is characterised by a dark and competent appearance, and, based on the dominantly well-formed and well-crystallised interlocking groundmass textures, appears to be coherent in nature. Alteration has been cited as an important process modifying the original igneous texture and mineralogy of kimberlite deposits (Skinner and Clement 1979; Clement Fig. 8 Photomicrographs of thin sections in DCK-U 2 that contain olivine grains with typical diffuse selvages. Scale bar is 1 mm. The selvages range from complete and symmetrical (a) to partial and asymmetrical (b, c). The selvage in a is cross-cut by a late-stage carbonate vein. Note the distinct finer-grained nature of olivine phenocrysts and carbonate laths within the selvages relative to the surrounding main groundmass. Also note the presence of type 1 groundmass with interlocking carbonate laths adjacent to selvage in b. Sample locations shown in Fig. 3

10 Fig. 9 Photograph of typical type c DCK-U 1 polished core sample, showing different domains of coarse-grain-supported and fine-grain matrix-supported kimberlite. Sample location shown in Fig. 3. Scale bar is 1 cm mineral replacement (Fig. 7). The DCK-U types 1 and 2 groundmass textures provide the clearest evidence for the coherent nature of these rocks. Textures in the volumetrically minor type 3 groundmass are probably also coherent, based on the similar macro- and microscopic appearance of these rocks ( Macroscopic and microscopic features section). Also, there are no features such as colloform textures indicative of late inter-clast void infilling in pyroclastic kimberlites (e.g. Mitchell 1997, p.18). Micro- to cryptocrystalline interstitial and segregation domains of carbonate±serpentine presently occupy the area in between some of the fine-grained groundmass minerals, and there is some disagreement in the literature whether these domains are the result of crystallisation from late-stage magmatic fluids (e.g. Mitchell 1984; Skinner and Marsh 2004), alteration overprint (Cas et al. 2008a) and/or represent remnant primary porosity in densely welded rocks (Brown et al. 2008). However, even if these domains are secondary, they generally comprise only a small volume (0 15%) of the rock, and therefore the groundmass is predominantly igneous in origin. Lastly, the texture and mineralogy of the groundmass of the DCK-U kimberlite is typical of worldwide kimberlites (Shee 1984; Mitchell 1986, 1995, 1997; Armstrong et al. 2004; Caro et al. 2004; Fedortchouk and Canil 2004; Roeder and Schulze 2008; van Straaten et al. 2008), and comparable to those produced in experimental studies (Mitchell 1986, p ; Otto and Wyllie 1993; Bellis and Canil 2007; Canil and Bellis 2008; Sparks et al. 2009). From the above discussion, it can be concluded that the DCK-U does not represent an altered volcaniclastic rock, but instead represents a single funnel-shaped unit of wellcrystallised coherent kimberlite formed by magmatic crystallisation of a liquid. Extrusive coherent origin and Reid 1989; Sparks et al. 2006; Stripp et al. 2006; Cas et al. 2008a, b; Porritt 2008; Hayman et al. 2009). Thus, we first consider whether the dark and competent apparently coherent kimberlite at Victor Northwest could represent a pyroclastic rock affected by such processes. The following observations show that the main textures of the DCK-U are not a result of alteration. Firstly, the entire unit contains partly to completely fresh olivine macrocrysts and/or phenocrysts, partly to completely fresh fine-grained minerals (Figs. 4b, 7 and 8), and does not show the pervasive alteration that is characteristic of the Southern African style (sensu Field and Scott Smith 1999) kimberlite pipes quoted in recent papers on alteration (Stripp et al. 2006; Cas et al. 2008a, b; Porritt 2008). Secondly, the fine-grained groundmass minerals in the DCK-U have euhedral subhedral shapes and interlocking textural inter-relationships indicative of igneous crystallisation rather than subsequent The DCK-U unit of the Victor Northwest kimberlite displays many indicators of extrusive origin, as summarised below in point form. 1. The DCK-U deposit has a funnel-shape or crater-like morphology (Figs. 1, 2 and 3). 2. The funnel-shaped basal contact of the DCK-U mirrors the lower contact of the underlying earlier pyroclastic kimberlite (VK-U 1 ; see Fig. 2 as well as Fig. 1c in van Straaten et al. 2009). The well-constrained and coincident location of the vents for the VK-U 1 and DCK-U deposits suggests that the magma forming both these rock units exploited the same feeder system. The similarity in mantle-derived macrocrysts and groundmass mineralogy ( Geology of the VNW upper strata and VNW upper strata: deposits from a single batch of magma sections) shows that the two rock units probably formed from the

11 same batch of magma, but by distinctly different eruptive processes as shown by the textural differences between the VK-U 1 and DCK-U (Table 1, Fig. 4; van Straaten et al. 2009). This suggests that the DCK-U originated from the same central feeder as VK-U 1 and partly infilled a vacant crater, covering earlier extrusive primary pyroclastic deposits. 3. There is no evidence, such as dykes, sills, apophyses, or peperitic margins, for magma intrusion into earlier consolidated or unconsolidated deposits. 4. The vast majority of contacts between DCK-U and enveloping units appear to be gradational over several metres (Fig. 4). Such contacts are much more easily explained by pyroclastic processes than by either intrusive or passive extrusive processes. 5. Mixed fragments of sedimentary country rocks, derived from different stratigraphic levels are present. Wellmixed country rock clasts are difficult to explain in a coherent rock emplaced by either intrusive or extrusive magmatic processes, but relatively easy to explain by pyroclastic processes. The above discussion shows that the DCK-U represents a single batch of coherent magma that was emplaced into an open nested crater within a crater, where it underwent mostly uniform crystallisation. The mode of emplacement of the magma into this nested crater is discussed below. Emplacement mechanisms of extrusive coherent DCK-U Theoretical constraints To consider the emplacement of the extrusive coherent DCK-U, we need to address theoretical constraints on the likelihood that kimberlite magma erupts effusively and/or forms reconstituted (i.e. clastogenic or welded) coherent deposits. This will help in formulating criteria to recognise these types of rock in the field and in thin sections. Possibility of kimberlite lava formation Kimberlite lava has been reported from Angola (Skinner and Marsh 2004; Eley et al. 2008), India (Mainkar et al. 2004) and Tanzania (Igwisi Hills: Reid et al. 1975; Dawson 1994). Detailed evidence that the Angolan and Indian kimberlites are lavas is lacking. The Igwisi Hills occurrence in Tanzania is currently the best candidate for a kimberlite lava, but it is argued by some to be a somewhat atypical kimberlite. The lavas at Igwisi Hills are very calcite rich (comparable to the evolved kimberlite sills at Benfontein and Wesselton described by Dawson and Hawthorne 1973 and Mitchell 1984, respectively) and contain groundmass spinels that could be indicative of significant crustal contamination (Mitchell 1986, p. 31; Mitchell 2008). The high carbonate content would have caused a more depolymerised (lower viscosity, e.g. Brooker et al. 2001) melt and could have promoted effusive rather than explosive behaviour. The rarity or absence of kimberlite lavas in the geological record can be explained by the susceptibility to erosion of the upper parts of ancient kimberlite volcanic systems (e.g. Lorenz 1986), combined with the absence of uneroded modern kimberlite volcanoes. The postulated extremely volatile-rich nature of kimberlite magma generated at depth (e.g. Dawson and Hawthorne 1970; Kopylova et al. 2007; Fedortchouk et al. 2010) does not preclude effusive eruption. No other magma type erupts only explosively, and comparable magma types such as carbonatites and lamproites do form intra-crater lava lakes (Krafft and Keller 1989; Smith and Lorenz 1989; Mitchell and Bergman 1991, p ) and/or extra-crater lava flows (Woolley and Church 2005; Seghedi et al. 2007). High concentrations of carbon dioxide in carbonatites and kimberlites are compensated by its higher solubility as CO 3 2 in the melt phase (Brooker et al. 2001). CO 3 2 in carbonatites is generally not easily exsolved into the vapour phase during ascent and decompression, allowing for non-explosive eruptions to occur; a similar behaviour is expected for carbonate-rich kimberlites. In summary, effusive kimberlite lavas are expected to form but should have a low preservation potential unless protected in a sufficiently deep intra-crater setting or deposited in areas with contemporaneous kimberlite volcanism and sedimentation. Coalescence, welding and hot pressing in kimberlite The extreme composition and resulting physical properties of kimberlite magmas affects volcanological processes, and particularly the process of welding. Kimberlite melt is characterised by a low silica content and high degree of depolymerisation. Estimates for the ratio of non-bridging oxygens (NBO) over cations in the tetrahedral site (T) range from two to six for typical kimberlite melts (Russell et al. 2006; Kopylova et al. 2007). The currently bestavailable viscosity models for silicate melts (Giordano et al. 2008) are based on melts with NBO/T values of and give viscosities down to Pa s. Physical property measurements on vesicle- and crystal-poor carbonatite pahoehoe lava flows (Dawson et al. 1990; Norton and Pinkerton 1997) give viscosities of Pa s. Both viscosity estimates indirectly imply a very low viscosity for kimberlite melt, likely below 1 10 Pa s. The presence of 50 vol.% olivine crystals in kimberlite magma will increase the bulk viscosity by at least one order of magnitude (Krieger and Dougherty 1959; Mueller et al. 2010). In addition, the common sub-spherical shape, and vesicle-poor nature of kimberlite juvenile pyroclasts and common presence of abundant groundmass-free magmaderived crystals as kimberlite pyroclasts is also suggestive

12 of very low melt viscosities. Pyroclasts from low viscosity melt are much more likely to undergo coalescence, agglutination or welding, and create reconstituted coherent rocks (Schmincke 1974; Sumner 1998; Sumner and Branney 2002; Sumner et al. 2005; Sparks et al. 2006; Brown et al. 2008). One of the commonly quoted requirements for welding is the presence of hot (with temperatures above the glass transition temperature, Tg) glassy pyroclasts (Quane and Russell 2005 and references therein). Pyroclasts that crystallise rather than quench to a glass are unlikely to undergo conventional welding. Glassy pyroclasts produced by brittle fragmentation of the magma during eruption can relax, flow viscously and anneal over the longer timescales associated with cooling (e.g. Dingwell 1996; Russell and Quane 2005). In the case of kimberlites, it is unclear whether the melt will quench to a glass or crystallise during an eruption. Kurszlaukis et al. (1998) re-melted a natural occurrence of a coherent kimberlite plug, and produced glassy pyroclasts upon quench fragmentation of the melt. However, the pyroclasts were produced at very high cooling rates due to their relatively small size and fast timescale of fragmentation during magma water interaction. It is unclear whether these conditions are typical of kimberlite eruptions, and there are several indicators that glass might not form during most kimberlite eruptions. Firstly, the gap between the liquidus and glass transition temperature for kimberlite magma is expected to be very high (Russell et al. 2006), making it difficult to quench the melt to a glass. We estimated the liquidus temperature for kimberlite melts as 1,200 C (Fedortchouk and Canil 2004), and the glass transition temperature as 600 C (see Russell et al. 2006; these calculations are based on the Giordano et al. (2008) model applied to several literature estimates of primitive kimberlite melt compositions). Secondly, volatile degassing is unlikely to cause a significant increase in the glass transition temperature of kimberlite. This conclusion follows modelling of Tg s for water-bearing and water-free kimberlite melts using the Giordano et al. (2008) model (Russell et al. 2006), as well a recent study on the effect of moderate amounts of CO 2 on the glass transition temperature (Morizet et al. 2007). The absence of a possible mechanism to change the Tg during an eruption contrasts markedly with other more common (low fragility) silicate melts where degassing is one of the most common quenching mechanisms (Dingwell 1996; Giordano et al. 2008). Thirdly, as a result of the difficulty in forming glass in such low-viscosity melts, crystallisation is likely to occur very fast in kimberlites (Brown et al. 2008). It is only in extremely small melt pyroclasts that we can envisage the cooling rate to be fast enough for glass to form. Fourthly, no kimberlite glass has ever been observed in nature; however, this could also be related to the absence of recent kimberlite deposits combined with the poor preservation potential of these glasses (e.g. Marshall 1961). The above theoretical considerations imply that glass might not form during a typical kimberlite eruption. Instead, the pyroclasts would crystallise with cooling. The pyroclastic deposit, however, could still form an annealed, dark, competent and massive rock via a process referred to as hot pressing in the ceramics industry (Vieira and Brook 1984; Venkatachari and Raj 1986). This process has received only little attention in the geological literature (Russell and Quane 2005), but may be analogous to welding as it can consolidate hot porous crystalline particles at relatively short timescales. Both welded or hot-pressed kimberlite deposits will look very different from common ignimbrites. Welding in kimberlites will not produce a well-developed fabric with elongated fiamme because kimberlite pyroclasts are commonly non-vesicular, crystal-rich and equant in shape. Compaction in kimberlites can proceed only via the reduction of inter-clast pore space. Furthermore, the lack of porous pyroclasts severely limits the total amount of volumetric strain to less than 36% which is a typical value for the inter-clast porosity for random jammed monodisperse spheres (e.g. Scott and Kilgour 1969). For more common rhyolitic ignimbrites, the high porosity of the pyroclasts allows for values of total volumetric strain that are at least double (Quane et al. 2009). Both the lower volumetric strain and extremely equant grain shape population of kimberlites decreases the visibility of the compaction and pore space reduction that accompany welding or hot pressing. In summary, at high emplacement temperatures, high magma flux and/or low cooling rates kimberlite pyroclasts will (mostly) remain liquid and coalesce, forming a clastogenic coherent kimberlite deposit. At intermediate emplacement temperatures, lower magma flux and/or intermediate cooling rates the pyroclasts will likely crystallise and since glass is unlikely to form in all except the smallest pyroclasts, a reconstituted coherent deposit will most likely form by hot pressing. In the latter case, the densification of these deposits containing non- to poorly vesicular equant pyroclasts will likely involve only moderate volumetric strain, hence classical textures such as fiamme or a strong fabric will not be present. At lower emplacement temperatures, magma flux and/or high cooling rates the pyroclasts will simply crystallise and be unable to form a reconstituted deposit. Possible occurrences of reconstituted coherent kimberlite Despite the likelihood of coalescence, agglutination and welding occurring in low-viscosity intra-crater kimberlite deposits, few descriptions of possible reconstituted coherent kimberlite occurrences have been made. Brown et al. (2008, 2009) suggest welded kimberlite deposits are present