Throughout geological history and on both continents and seafloor there have periodically been erupted enormous amounts of basalt in very short periods of time, often apparently thousands of cubic Km in a matter of a few hundred years, or even days. These may cover areas of as much as 100,000 sq. km or more as both flood basalts and as intruded sills, the magmas of each major pulse often being remarkably homogenous. At the same time, different units in the same region can differ widely especially in their Ti, P, K, Zr, in fact in all residual elements. They are in many respects similar to but are seldom as basic as ORBs usually having 52 - 56% SO2, have high and variable Ti, Fe as have both ORBs and OIBs (but usually showing marked Fe-Mg fractionation) and have a typical continental signatures indistinguishable from the andesites. Late stage alkali-basalts, melilotites and carbonatites are sometimes associated. Small volumes of residual iron-rich granophyres are fairly common. Usually tholeiitic, the typical cumulate is an orthpyroxenite, but sometimes picrite, rarely anorthosite.

Gondwana Flood Basalts and Sills

When the Gondwana supercontinent fragmented into Europe, Asia, the Americas, Africa, Australia and Antarctica in the Jurassic era (about 165Myr bp), the fractures resulted in enormous outpourings of basalt, not only along the spreading axes now seen as mid-oceanic ridges but along transverse faults extending into the continents where we now see the continental flood basalts. These are now termed the Ferrar Dolerites and Volcanics in Antarctica, the Tasmanian Dolerites in Australia, the Deccan Traps in India, (mainly basalts), the Parana Basalts in southern Brazil, the Karoo Dolerites in South Africa, the Hebridean Province of the British Tertiary in western Scotland and northern Ireland, the Palisades Sill and related intrusions found in eastern USA., and the Flood basalts or east and West Greenland. There are older Flood basalts, mainly found in the Archaean and Proterozoic, for example the Coppermine Basalts of the north-western Canadian Shield, and the Keewenawan Basalts of Southern Canada-Northern Minnesota. The Siberian Traps are another extensive series of flood-basalts but of Permo-Triassic age. The only younger members are the 15 myr Columbia River basalts of the state of Oregon, Western USA.

Variation diagram for all available Jurassic flood basalts of Antarctica. Some sills included, seen especially in those samples of more than 10% MgO which include orthopyroxenites and with some higher fractionates, esp ferro-basalts. Note wide range in TiO2 in different magmas, the Kirwan basalts being the lowest.

Originating in the enriched sub-continental mantle, continental Flood Basalts have a very different fingerprint to the Oceanic Ridge basalts. No depleted members are known, the least enriched being close to EMORB composition, but usually with even more elevated Cs, Rb, Ba, U, Th, K but depleted Nb-Ta. Some form the most LILE-enriched quartz saturated basalts known,in fact they have a typical continental signature, very similar to Andean andesites, though lower in silica and alumina and much more iron-rich in fractionated members.

Because they were extruded through very thick continental plates of lower specific gravity, in most areas covered with a up to 5-6000ft of detrital sandstone-arkose in the old Gondwana terranes, the basalts often intruded as sills, many of 1000-1200ft in thickness. The volumes of sill in Antarctica are of at least 200,000 cubic km, though possibly as much more is buried under ice.

The Gondwana basalts are mainly quartz saturated rocks with plagioclase, clinopyroxene, orthopyroxene, pigeonite and, rarely, olivine except in South Africa where olivine-bearing rocks are more common. There may be half a dozen distinct magma types in one area usually but not always obviously related by orthopyroxene fractionation. Sills may form in the basement granites, at the peneplained surface on which rest the sandstones (at least in Antarctica and Tasmania) and at intervals of a few thousand feet up through the flat-lying sandstones above. Usually the heavier, orthopyroxene or olivine-rich sills are found low in the sequence, higher up may be ferrobasalts with up to as much as 56% silica.

The sills are quite differentiated whenever the thickness exceeds a few hundred feet with the most magnesian compositions being found in the lower third (often with crystal cumulates of orthopyroxene) with more feldspathic rock in the upper third and lenses rich in quartz and potash-group elements near the top.

High in the sequence and resting on the old Jurassic landsurface are remnants of lavas, tuffs and laharic mudflows. Some flows and tuffs near the upper Mawson Glacier (South Victoria Land) have granophyric glass altered to zeolites especially mesolite, scolecite, heulandite, okenite, stilbite etc, with quartz geodes while in the Parana basalts of South America tropical weathering and percolation of warm acid water causes silica solution at the surface resulting in cavities deeper in the flows being partly filled with gorgeous secondary amethyst quartz.

The Finger Mountain Sill of Jurassic Ferrar Dolerite, Upper Taylor Glacier, Antarctica is about 600ft thick and is intruded into Beacon Sandstone. The younger inclined sheet on the skyline cuts across it diagonally. Finger Mtn is an easy climb from higher up the glacier, not so easy straight up! For futher information on the Beacon Sandstones, visit www.RossSea.info (Photo B.Gunn 1959)

Formation of Dolerite Sills

A basaltic sill of perhaps 1000ft thickness being intruded at depth and being well insulated may take up to 10,000 years to solidify. In this time a good deal of differentiation may take place. Even a small lava lake such as 1959 Kilauea Iki in Hawaii took more than 10 years to solidify and as heavy olivine was present, it showed quite large variations in Cr, Mg, Ni with depth when later drilled, (Helz, 1987, Geochem. Soc.Sp.Pub., 241-258). Virtually all sills more than a few hundred feet in thickness will show chemical variations due to two reasons, both of which may be sometimes found in a single sill.

1. Gravitational Fractionation
   Early formed minerals especially chromite, olivine, orthopyroxene, clinopyroxene and calcic plagioclase tend to settle into the lower half of the sill. The residual liquid, lower in MgO, Cr, Ni, Co etc but richer in Si, Al, Na, Sr, solidifies in the upper half of the sill. The results in an "S" shaped distribution with height for Mg, Cr, Ni, Ca and a reverse "S" for Na, Al, Sr, Na. The last few percent of the basalt to solidfy is greatly enriched in the elements found in low temperature minerals. These include include K, Ba, Rb, Cs, U, Th, Si, Fe, Ti, P, Y, V, REE all of which percolate upward to form lenses of residual pegmatite a few inches to feet thick at levels a few feet to perhaps as much as 100ft below the top of the sill. Graphic intergrowths of quartz and a potash feldspar are typical.
   Variation diagrams of such sills are usually elliptical or form a figure "8". In a plot of say Na against Mg, the Na content at 7% Mg on the lower limb of the "S" will be much less than the Na at 7% Mg on the middle limb of the "S". Smooth variation diagrams are therefore impossible in sills, a fact that baffled me for years.
   The end product of sill fractionation is an Fe-rich granophyre which may be very coarse-grained and look almost granitic. There are a few cases of an iron poor rhyo-dacitic rock being formed.
2. Multiple Injection.
   Lava rising through the continental crust forms a sill when the "Hydrostatic" (magmastatic??) head of pressure exceeds the confining pressure of the country rock. For basalt of sg 3.5 or more rising through quartzites and arkose of sg 2.4 or less, this point is reached well below the surface, as much as 8000ft below, possibly even more.
   As it may take several hundreds to several thousands of years to solidify there is a high probability that a second pulse of magma will occur before final solidification. This new injection follows the path of least resistance into the already-formed sill and makes it thicker. If the sill has solidified the new pulse will be injected above it, in some cases actually between it and the overlying sandstone, sometimes a few hundred feet higher. If the new pulse has a different composition the variation with height in a doubly injected sill may look very odd. In some cases what follows as a second injection is a thick mush of olivine (Palisades Sill, NY) or orthopyroxene (Vanda Sill, Basement Sill, Solitary Rocks Sill (Antarctica), or the Ben Lomond Sill, Tasmania) which has settled out in some vast magma chamber below. Perhaps the added weight of the sill above renders it unstable and it is set in motion upwards to pour into the partly formed sill above. Some times the contact between mush and new sill is gradational, but if the sill is 90% solidified it may be quite abrupt. In the Solitary Rocks Sill in the Taylor Valley, (Gunn, 1966) a sill of about 7-8% MgO within a few inches grades into a crystal cumulate with an average of 18-20% MgO. This mush in turn may gravitationally settle so the cumulate of 25% MgO may decline to only 10-12% a few hundred feet above but becoming richer in bytownitic plagioclase. Near the top is the granophyre which came out of the original sill, orthopyroxene cumulate having little or no granophyric residue.
   The result of the cumulate injection is to produce an average bulk compostion with perhaps 150% or more Cr, Ni, Mg etc than the average found in the upper and lower chilled margins. Few sills anywhere have an average bulk composition the same as that of the margins. so repeated injection into sills must be common. The Palisades Sill has much higher bulk Ca-Al for example.
   Fingerprints of sills of variable orthopyroxene content will, like lavas of variable olivine content, show a parallel pattern. What will be the effect on light REE vs heavy REE in an REE diagram?? What ought to be the effect on REE in sills like the New Mtn Sill (Ferrar Dolerites) where only Cpx has accumulated? I do not believe anyone has yet analysed REE from a sill known to be clinopyroxene-fractionated rather than olivine or orthopyroxene-fractionated.

Ferrar Dolerites (Antarctica)

	Variation in metals for all Ferrar dolerites lavas and some sills.
	Alkaline earths / Zr. Note that Y far exceeds Nb (and Ta) as it does in all NMORBs. Rb is variable but usually also in excess of Nb, while Ba which is enriched steeply with smaller degrees of melt and also with fractionation is much greater than Y (unlike NMORB) and very variable. The fact that few of these samples are chilled or glasses increaces the Ba scatter especially.
	The REE relative dispersion is typical of flood basalts. being more LREE enriched than standard EMORB, with Ce/La = 2 and Nd = La. The Kirwan Basalts from the Kirwan Mts, on the South African side of the continent are relatively the lowest in both LILE and LREE as well as TiO2.
	Type Ferrar Dolerite sills, Ferrar Glacier region, South Victoria Land, (Gunn, 1962, 1963, 1965, 1966) Note low but regular TiO2, moderately high K, regular CaO but erratic Al2O3 due to sill effect, moderated in the case of CaO by clinopyroxene compensating for decrease in calcic plagioclase. Cumulates are all orthopyroxene, but with some enrichment in Cpx + pigeonite at intermediate MgO levels. Some low MgO pegmatites are iron enriched.
	Composite diagram of hypersthene dolerite type from Portal Peak, near Beardmore Glacier, Hergt et al, (1989). These are akin to the Lake Vanda Sill type of the Wright Valley near McMurdo Sound. Included also are "high Ti" lavas from Scarab Peak, in Northern Victoria Land (Fleming et al, 1992). These are "pigeonite dolerite" type (actually lavas), fractionated from the Vanda type by removal of orthopyroxene-plagioclase. Sc is not depleted so no clinopyroxene is involved. The La/Lu remains the same so the difference is due to fractionation only.
	Two variants of Ferrar Dolerite from Coates Land on the Weddell Sea coast.
	Vestfjella lavas and dolerites, Dronning Maud Land, Indian Ocean coast. Typical Ferrar type but low Th-U. Luttinen et al (2001)
	Lavas of Ferrar Volcanics, Prince Albert Mts, North Victoria Land. Again, a typical Ferrar signature. (Antonini et al, 1999)
	Kirwan basalts from Dronning Maud Land, on South African side of continent. Note similarity to low LILE Flood Basalts of Picture Gorge, Columbia River, the Low-Ti basalts of the Deccan and the Lesotho Basalts of the Karoo.

The Ferrar Dolerites show low HFSE and high LILE and are of the kind sometimes called "Low Ti-type". However as we shall see in the Parana and Columbia River, a range of type exists between "High Ti" and "Low Ti" The "Pigeonite Dolerites" found in the Ferrar seem to be crystal fractionates and have high silica (54-55%) elevated LILE and Fe, but low Mg (<3%).

Tasmanian Dolerites (Australia)

The Tasmanian Dolerites have a similar high LILE to the Ferrar Dolerite type but are of fairly constant composition in chilled marginal rocks. Orthopyroxene cumulates are common. Drill cores of thick intrusions reveal thick black glassy granophyres in the upper sections of, eg, the Great Lakes Sill. The Mt Wellington Sill, above Hobart, the capital city, is the most studied, and appears to be a multiple-injection type sill with excess orthopyroxene. Ian McDougal of A.N.U. Canberra, argues that the sill dips west, and sampling up a sloping road has falsified the apparent heights in the sill. I do not believe this controversy has been settled as yet.

	Variation diagram for Tasmanian Dolerites. Many years ago Dr Janet Hergt sent me her PhD data on Tasmanian Dolerites. As this is not on the GEOROC database it may not have been published, but together with Dr Ian MacDougals' PhD data, it makes up a useful file. This diagram throws new light on all the CFB's. The cluster of points between 6 and 7% MgO are the chilled margins or primary rocks and the whole series are quite linear until the separation of the granophyres and pegmatites at MgO <2%. Obviously both orthopyroxne and clinopyroxene (mainly latestage pigeonite) have settled, alumina is expelled but CaO increases with MgO. Sc also shows that CPX goes along with OPX. There is obviously only one magma type the small differences in chilled margins being related by difference in orthopyroxene, as in the different chilled margins seen in South Victoria Land Ferrar Dolerites, and if there is any late stage autointrusion of orthopyroxenite it is directly related. The Ben Lomond sill in NE Tasmania appeared in thin section to have unusually pure Opx concentrate which in fact shows at the high MgO end of the diagram. Ni-Mg distribution (see below ) is unusually flat and quite linear from 10% MgO to 2% as is Zn, Cu, etc. This series gives us definite fractionation paths for CFB's in high level sills. We can envision pools of several hundred thousand cubic km of magma trapped beneath the continental mass possibly for times measured in millions of years, slowly separating out orthopyroxene, (and in early stages , olivine) concentrates, until fracturing of the crust and earth movment allow sudden outpouring, which in regions of dense crust, may actually mainly reach the surface. The crustal signature appears to be inherited from from the subcrustal mantle previously enriched by subduction. The Columbia River Basalts seem to indicate sometimes separate olivine, OPX and CPX trends. Neither seem at this point to offer an explanation of the very different Ti, P levels seen in other CFB's, which, while never greater than NMORB levels, would demand unacceptable levels of crustal contamination to reduce the TiO2 to the 0.75% levels seen. P2O5 is also remarkably low, compared with P2O5 in the Umatilla Basalts of the Columbia River Group where it may exceed 1%. However, the same scale of variation is seen in all arc andesites.
	Mg vs Cr, Co, Ni, Cu, Zn, V for Tasmanian Dolerites. The unusually flat Ni trend points toward orthopyroxene. Zn declines with increacing MgO more steeply than in olivine controlled series such as seen in Kilauea. Cu is unusually regular. Is the Cr concentrated in Cpx or in sulfides? We do not know at this point, one does get kicks on the SEM when traversing OPX grains, probably Cr spinels are present.
	The Tasmanian Dolerites.
	Two types of magma from Tasmania. Pigeonite type from Red Hill and the more magnesian Ben Lomond and Mt Wellington type both associated with orthopyroxene-rich cores to sills, approximating to Peneplain Sill type of Antarctica. The great similarity to Ferrar type is obvious.(Brauns & Hergt, 2000)
	Na-Mg-K diagram for the Tasmanian dolerites. As these are all sills, it shows that within sill fractionation is no different to the trends seen elsewhere for lavas.

Deccan Traps (India)

Mambai, India

By Hetu Sheth

Karoo Dolerites (South Africa)

The Karoo Series.
Dr Tony Erlank of Rondebosch University, now deceased, about ten years ago put together a file of about 950 analyses of Karoo rocks. While now somewhat outdated these show, perhaps 2/3 of the samples to be tholeiitic basalts with numerous picrites, with, after a considerable andesite gap, numerous potassic rhyolites, with some alkaline rocks, basanites and even carbonatites. Karoo sediments intruded by scattered sills and plugs extend from near Capetown to about the Limpopo River. Flood basalts are found at Lesotho, west of Durban and along the Lebombo Mts west of Maputo, near Kruger National Park. A third area occurs in the west in the Angolan desert at Etendeka.

	Lesotho Fm. Basalts. These are the familiar "Low Ti" flood basalt similar to the Kirwan Basalts of Antarctica, or the Picture Gorge type of the Columbia River. The Kraai River Basalts and the Brosterlei Basalts are similar. The data for the Lebombo rocks, is not complete but they appear to be considerably more enriched, in fact the bulk of samples are very potassic, but not enough information is available at this point to delineate potassic areas.
	Etendeka basalts- Quartz latites. (Erlank File). These occur in the western Angolan desert and range from potassic basalts with again an andesitic gap to what are termed "latites". They appear related to the basalts and appear to be potassic granophyres. The La/Yb ratio changes continuously and there seem to be a series of lineages. With up to 5% K2O these latites are probably the most potassic rocks associated with Flood Basalts.
	Etendeka picrites. Thompson et al, 2001 shows similar, probably in part the same rocks, without the "latite" but including picrites up to 25% MgO. These are similarly highly potassic.
	Etendeka ferro-picrites. Gibson et al, (2000) show a series of HFSE depleted ferro-picrite from the same region with elevated titanium and only a minor negative Nb anomaly.
	Na-Mg-K diagram for the Karoo, Tony Erlank's file. The Karoo is more variable than perhaps any other CRB province and in includes the Etendeka dacites. Later data which includes carbonatites is even more variable.

No data is yet available on the Dicker Villem carbonatites which occurs in the Etendeka area.

The Gramado type Parana Basalts on the opposite shore of the Atlantic are similarly potassic.

Parana Flood Basalt types (Southern Brazil)

The Parana Basalts extend through southern Brazil, Paraguay and Uraguay and are reputed to cover over a million sq km. At Iguazu Falls on the Parana R,, the river cascades over a series of massive flat-lying basalts flows in a series of the most spectacular falls in the world. The different magma types range from high Ti to low Ti (about the same as E-Type MORB), but show only very slight fractionation effects.

Lower TiO2		Higher TiO2
	Variation diagram for the Esmeralda and Gramado low Ti Parana magmas. Fractionation is more obvious in this group, but note how, apart from the high and variable K, the similarity to MORBs, showing the same spread of degrees of partial melt. In this series the P levels are quite low.		Parana variation diagram with 980 analyses including the Pitanga type, the Paranapamena, Ribeira, and Urubici flows with the Ponta Grossa Dykes but no sills. Thanks are cordially extended to D. Peate of Open University who sent us his entire database. The Parana series have a mode between 4 - 4.5% MgO. A poorly defined break can be seen between the high and low TiO2 members. The Parana show the characteristics of most Flood Basalts, low alumina averaging 14%, higher than usual Fe and Ti but without any peak which would be seen had fractionated sills been included, though a few samples with > 7% MgO suggests some fractionation in the Ponta Grossa Dykes.
	Variation in metallic trace elements in the Gramado and Esmeralda types with MgO. Notice the erratically high Co, Cu and Cr which suggests the inclusion of sulfides, yet the Zn remains constant!		Trace metals for High Ti Parana CFB variants. The V follows entirely the ORB fractionation - partial melt constraints as does Cu and Zn. Cu has occasional straying values, Cr and Ni again seems to have rather more variance than they should but the plane of partial melt is known to be at a broad angle to the fractionation direction but is not yet well defined.
	The alkaline earth elements for the low-Ti Gramado and Esmeralda types of the Parana Basalts. Note the domination of fractionation effects, the low Nb vs Y as in all CFBs and the minimal scatter in Zr.		The alkaline earths. In this more enriched, higher Ti group the Nb is almost = to Y and to Rb. Ba & Sr lie mainly within the fractionation - partial melt limits but about 10% are undoubtedly outside these limits. It may be that these late stage elements tend to puddle in the flows. We can see the absence of high magnesian members and that CFBs such as these are far from being all high-degree melts.
	Low Ti Parana basalt		Paraparanema Magma Type, Parana Basalts
	Esmeralda Type		Pitanga Type.
	Gramado Type.		Urubici type Parana basalt
			Parana Rhyolites.

---

P.Hooper, (2000, G-Cubed, Vol.1, June), gives a total area for the Columbia River basalts of approx 77,000 sq miles (=200,000 sq km).

Columbia River Basalts (USA)

Thanks mainly to the work of Dr Peter Hooper at WSU and his associates these are perhaps the best known of all the flood basalts. Hooper claims that the Roza centre twice erupted 700 cubic km within a few days 15,000,000 years ago, forming a flow with a length of 190 miles (300km), an area of 15,400 sq miles (40,000 sq km), while the Pomona member flowed an even greater distance of 600km to the Pacific Ocean.

Like the Gondwana flood basalts, the CRB's are quite different to MORBs and are divisible into a dozen magma types on their characteristic chemistry. Some of the lava groups, eg the Prineville Basalts and the Umatilla series plot in the Hawaite-Mugearite areas of an alkali-silica diagram but this is due to their enhanced K content. Four of the main groups are shown below, the Prineville, Grande Ronde, Umatilla and Roza series. They do not fall into the "Basalt Triangle" on a Al-Mg-Ca diagram but above it, and they are extremely potassic.

Their signatures vary from similar to a rather mature island arc andesite type (Picture Gorge, Powder River Group, Imnaha Basalt, Wanapum Basalt) to a more enriched form with Rb at 8-10 times E-type basalt, (Prineville, Grande Ronde, Roza) with the extreme being seen in the Umatilla Basalt which may have only 5-12 times Rb but is very enriched in Ba (50-60 times E-type) as well as in P and Zr.

This means that Umatilla has 2700 to 4000 ppm Ba cf 140-430 Ba for the Imnaha rocks while Zr is at 400-500 ppm cf 120-230. Fortunately some new REE data has become available for Umatilla, these being among the most Ba (and Zr, P) enriched basalts known.

In the western USA, this characteristic andesite-like signature is not confined to Columbia River Basalts. Similar rocks may be seen elsewhere in what is termed the Basin and Range province, eg in the Elkhead Vol. Field in SW USA, (Leat, et al, 1988, JGR 102) where the rocks have a typical orogenic signature but have acquired more K (and Ba) than soda and plot in the Hawaiite-mugearite fields, however, Ti as well as Nb remain low. Other continental basalts showing a similar signature include Walton Peak (Thompson, et al, J.Geol.Soc.S.L., 150), the Flattop field, (Gibson et al, 1991, JGR 96). These Basin and Range lavas may also be CFBs.

In the "GEOROC" database, these are included in the Columbia River group, along with the Snake River basalts, the Craters of the Moon and even the Yellowstone-Shoshone Mtn lavas, a somewhat doubtful association.

Odd chemical composition might be expected as in the Washington-Oregon area, where an offshore ocean-spreading ridge (the Gorda Ridge) sends major transcurrent faults under the continent,. This does not however explain the unusually high Ba enrichment or the highly variable P and Zr nor the fact that exactly the same variations are seen in other continental basalts from a wide range of tectonic environment.

	Columbia River basalts; Na2O - MgO - K2O
	Columbia River basalts variation diagram (new)
	Columbia River basalts MgO vs metals (new)
	Columbia River basalts Zr vs Nb (new)
	Columbia River basalts TiO2 vs K2O (new)
	Columbia River basalts MgO vs TiO2 (new)
	Columbia River basalts MgO vs Alkaline elements (new)

Steens Mountain, Oregon

Click to enlarge The block-faulted scarp of Steen's Mts, Southern Or. as seen from the Alvord Desert. The thick stacked flows of continental flood basalt make this the best cross section of an old shield volcano known. Photo courtesy of Nick Jarboe who is beginning new work. He promises close ups of those feldspar cumulates!

Steens Mountain is a centralised Flood basalt subprovince of 16.6M year age closely allied to the Columbia River Basalts and lying in southern Oregon. An excellent cross-section is on view due to the centre being cut by a Basin and Range fault which has created a NNE-trending, east-facing scarp about 3200ft (1000m) high, one of the greatest vertical exposures of Tertiary basalt known. Related flows may have once covered 25,000 to 50,000 km2 but effusion finally centred near the crest of the present Steens Mtn, forming a shield volcano.

In the early 1970's the late N.D.Watkins and I began a study, in the hope that such a large lava pile would reveal mantle depletion effects in late compared to early stages. Watkins sampled by diamond drill some 70 lava flows of a total of perhaps 100 ranging from 1-10 m thick, taking four samples per flow.

Steens forms the surface expression of a presumed large fractionating magma chamber, which periodically expelled highly feldspar-phyric magmas. Plagioclase is the main phenocrysts phase (as in most ORBs) followed by much finer-grained olivine, at least no coarse olivine has been expelled. The plagioclase probably floated and was largely extruded with some flows including up to 50% feldspar phenocrysts sometimes up to 4cm long. The Sc content remains constant showing no clinopyroxene formed. Other flows are quite aphyric. Extensive olivine fractionation is indicated by the trend of increasing Fe/Mg from bottom to top and the strongly decreasing Ni and Cr. No olivine cumulates or picrites were extruded however, the most magnesian rocks in our study having 7% MgO and in a study by J. Johnstone of WSU, 12%. The magma chamber below must therefore now include extensive peridotite. Residual elements not included in either olivine or plagioclase, eg Cs, Rb, Ba, Th, U, Nb, La, K and light REE increase steadily towards the top of the lava pile while feldspars become more sodic and ferro-magnesians more iron-rich, an excellent example of "hidden layering" . Nickel variation is wide, 5 - 345ppm, yet the correlation with MgO is unusually poor which may indicate sulfide mineralisation. Olivines in section are limited to a few grains and normatively to only 8.25% in the most magnesian rock.

A little below the centre of the series there occurs an odd "mafic zone" of 15 flows of higher MgO, averaging about 8% compared with 7% near the base and 2.3% at the top of the mountain. In our original paper (Gunn & Watkins, 1970, GSAB 81) this is termed the "E" group. This horizon appears in other sections, eg between 7200ft and 7600ft in Johnstone, (1999).

Calculated CIPW Norms show the lower lavas to have an An51 plagioclase compared with An57 in the mafic zone and An27 at the top. Ferro-magnesian ratios for minerals are 78.5 in the mafic zone and 61 and 50. The most plagioclase-rich sample recorded came from above the mafic layer and had An57.5 with a total plagioclase content of 76%. However, with 1-inch drill cores and 4 cm plagioclases, such a mode has little meaning.
The REE show a divergence toward the light REE end. La/Lu ratio varies from 28.5 in the mafic zone to 41.7 at the base to 63 at the top. Johnstone (1999) therefore erred when it was stated "Rare Earth and trace element patterns throughout the sequence are similar"!

Though not shown in our original paper, a Ni/Ba diagram shows Ni decreasing from 260 ppm to 5 while Ba increases from a low of 300 ppm to almost 1000 ppm in some upper flows. If we note that the plagioclase is considerably more sodic and the ferromagnesians more Fe-rich than is seen in layered flood basalts of the British Tertiary and in Skaergaard, we might be tempted to guess that a considerable volume of more calcic and magnesian magma was not erupted. Such a basic gabbroic intrusion might well be heavily Cu-Ni mineralised.

	MORB- normalised fingerprint for Steens lavas. Data from Johnstone, (1999). Compare with Imnaha and Grand Ronde.
	Variation in FM ratio and Na/(Na+Ca) for the Steens lava pile. Data from Gunn and Watkins (1971) Both increase upwards, and in general at the secondary variation level show some similarity, but it is difficult to explain the apparent variability. The magma chamber may have been periodically refilled, or magmatic overturn might have brought variable compositions to the top. Are the first part of each flow to be extruded the most feldspar rich? Why is the feldspar content often cyclic (Watkins & Gunn, 1969, Nature, 224,(5217), 360-361). Steens is still somewhat enigmatic.
	Variantion Diagram for a cross section of Steens Mtn in Southern Oregon, all available data. See the high Alumina plagioclase cumulate and the low alumina, high Ti,Fe prag depleted samples. As all tholeiitic basalts of mantle origin crystallise in the order olivine - ol+plag - ol + plag + cpx it is perhaps surprising we do not see plagioclase cumulates more frequently.
	Metals for Steens Mtn. There appears to be some mineralisation of Cu and Ni. V appear to be unusually irregular, reason not known.

Further Work

Steens has perhaps been already subject to undue attention, but as it has the greatest number of exposed flood basalt lavas known together with the fact it is perhaps the only exposed "layered extrusion" known and has the same "allivalite" fractionation patterns (Ol + Plag) seen in the British Tertiary, still more could be learnt. What is the range in composition of single flow members both vertically and in length? What an unparalleled opportunity to study feldspar compositions and their effects on fractionation! Why are some flows described as "highly altered"? This usually suggests different chemistry. How deep down does the now solidified peridotitic magma chamber lie?
Have other members of the CRBs undergone similar fractionation?
There are many questions yet to be answered.

The Snake River Basalts

The Snake River Plains lie in a boomerang-shaped fault depression extending east from the eastern edge of the  Columbia River Basalts  for 400 miles to the edge of the rhyolitic caldera of the Yellowstone thermal area. The rocks are mainly very flat-lying basalts with occasional associated rhyolites while the very youngest cinder cones at the “Craters of the Moon” national park fall in the  hawaiite-mugearite fields on an alkali/silica diagram.  The ages range from 15myr at the west, approximately the same as the youngest CRBs) and reportedly young to the east with the 2 ka Craters of the  Moon lie in the middle of the eastern sector.

Map of the Snake River Plains Basalts (USGS).

Drilling and seismic work show, according to the USGS, the maximum thickness of the older flows to be in excess of  4000ft, so that very roughly, about 60,000 cub.km.of lava must be present. Flows have mainly pahoehoe  surfaces though the glass “skin” has been etched away by wind-blown dust. Fissures lie open, one of the youngest shows a levee enclosing an area of a square mile from which most of the lava has drained back.  Down in one of the fissures at a depth of 200ft ice lies in the hottest summer.

The Snake River Plains, satellite view. Basin and Range structure is seen both to the north and South.

Lava which flowed into pre-existing stream and lake beds formed porous pillow lava and tuff which now act as channels by which aquifers drain into the Snake River Canyon.  We had a rather amusing experience, too long a story to relate in detail, when the Idaho Nuclear Research Station (located about 20 miles from Pocatello) chief asked me to give a talk to some of the technicians on how to locate aquifers for which they were drilling, without a very good success rate.  It turned out they saw an aquifer as a good way to dispose of unwanted radioactive nuclear waste. They were not geologists and it had not occurred to them that water at a depth of several hundred feet could ever come to the surface. It had not occurred to them either that the water-falls cascading from between the flows into the Snake were  those same streams and they were proposing to pour radioactive waste into the drinking-water supply for Lewiston, Clarkston, Portland, Sanfrancisco and in fact a good proportion of the western US.

Challenged to suggest an alternative I could only suggest the opposite, a desert area in a tectonic depression which was seismically stable such as Death Valley where there could be no fear of concrete enclosed waste ever getting below a water table.  The waste disposal unit for the USA is in fact now located on the edge of Death Valley, so I hope it pans out.

Lava Compositions of the Snake River Basalts

As yet there is surprisingly little data.  Leeman and Vitaliano (1976, BGSA 87, 1777-92) show 16 ME analyses of a compact group of high K basalts  some showing marked Fe enrichment. Tilley and Thompson (1970) and George Stone of the USGS (unpublished data, circa 1967) showed a wider range varying from 44-48% SiO₂ and I analysed some hawaiites from Craters of the Moon (unpublished).

Lum and Leeman (1989, JGR, B94, 7871-84) and Honjo and Leeman (1987, C.Min.Pet. 96, 163-167) added trace elements and some REE.   The data is now rather old and there is a good deal of scatter.  In general they show the high K and Ba with REE equal to or greater than EMORB levels as have most CFB’s, but except possibly for the most basic, no negative Nb-Ta anomaly is seen. As we have seen, this is an invariable characteristic of CFBs, there being no exceptions. Two more detailed though old analyses by Thompson etal, 1983, (Continental Flood Basalts” also give a low Zr/Nb of 11. However the presumably related Steens Mountain and Columbia River Basalts are notoriously variable and at least one characteristic, the very high P seen especially in  the  Umatilla series  is also seen in  Snake River where it reaches 2%.

Variation diagram for some Snake River Basalts. Note the high Fe trend. (Data up to mid-2003)

Geist et al, (2001, Geol.Soc.Amer.Sp. Pap. 353) present data from drill cores from the north-eastern extremity of the group and are also puzzled by the high P2O5 content and suggest the possibility of crustal contamination, though their highest P2O5 is only 1%. Their MORB normalised diagrams again show a more  OIB-type fingerprint, not a typical CFB one.

Iron is high at 16% and TiO2  unusually so even for CFBs at 4% with some indication of two levels being present. Again this matches the Umatilla rocks of the CRB. Rhyolites are very potassic (6% K2O) but seem to be related to the basalts and intermediate rocks though the extensive ignimbrites of the Yellowstone region are almost certainly of refused crustal keel material and there is a mineralogical suggestion of hybridisation.

A group from the University of Utah headed by Dr John Shervais are making a new study of the Snake River and we may hope to see new data by early 2004. The Idaho Geological Survey also have 400 new analyses in the publication process.

Other groups from  Idaho State University as well are also investigating the series. Dr Scott Hughes has recently (Dec/03) forwarded a data file of the SRB with a total of nearly 1000 new and old analyses from the eastern Snake River plains. Two fairly distinct series are present, the tholeiitic flood basalts which, though lacking Nb, appear to be CFB type and a series of alkali basalt, hawaiite, mugearite, trachyte, commendite, rhyolite which seem to be confined to the monogenetic superficial scoria and lava cones as seen at Craters of the Moon. This section will be updated as soon as these new data have been properly evaluated.

In general, the high iron enrichment  of the tholeiitic type and the high Ba and P suggests a CFB affiliation  in the older rocks. The reversion to an alkaline type in the closing stages seems similar to that seen in Mauna Kea (Hawaii) where after the cessation of the shield-building tholeiitic stage and a considerable time hiatus, superficial monogenetic alkaline rocks are extruded. No fingerprints are as yet available for the  olivine tholeiite-type rocks of the 9 – 7 ma period (Bonnishchen et al, see NET page).

The Siberian Traps (Russia)

Lightfoot et al (1990, 1993) show the Siberian Traps to be similar to the CRB's. Unfortunately Ba was not determined, and the wide range of Rb may suggest some alteration. Six magma types are present ranging from tholeiitic to alkali-basalt. Carbonatites are also found associated with these rocks. At minimum the Siberian Traps of the Norils'k area cover 1,000,000 sq km but related areas have either been buried or removed by erosion. These rocks are now a major source of nickel and also of PGE's.

The Siberian Traps.

The Palisades Sill (USA)

The Palisades form an extensive 1000ft-thick dolerite sill lying along the western side of the Hudson River, west and north of New York. Being so prominent it has been the focus of a good deal of primary study. It is of typical Flood Basalt composition being rather high in K, Ba (x2 - x4 times MORB) and what data are available suggest it is of similar composition to the geochemical standard rock W-1, the Centerville Diabase from near Washington DC which is I believe of the same age,
Walker (1940, BGSA 51) of Walker & Poldervaart fame, decided that the variations in the sill, including the well-known olivine ledge near the base, was due to olivine-clinopyroxene fractionation with the heavy olivines settling to the bottom and attributed the lack of any smooth compositional variation diagrams to the fact that clinopyroxene had about the same silica content as the average rock. Pearce (1969, J.Pet 11) ratioed the major oxides to Al₂O₃ which then showed fair agreement in diagrammatic slope with olivine and clinopyroxene. In the same era or a little earlier (1955-65) I was also somewhat baffled by the erratic compositional nature of the Antarctic Ferrar Dolerite sills, but found that element variation diagrams described ellipses, not straight lines or simple curves, the reason being that two points of similar Mg contents in the lower and upper parts of a sill, will have different LILE/HFSE ratios. Plotting several overlapping ellipses from different locations does indeed produce a profusion of apparently scattered points Plotting element concentration against height also showed that all the larger Ferrar Dolerite sills were multiple injections. I tested the same theory on the Palisades sill and found that the bulk average composition of the sill was much higher in Ca, Al etc that the average of the upper and lower chilled margins, the magnesian olivine ledge being only ten feet thick. Like the Ferrar Sills it must have been multiply injected by magmas of different composition, on top of which was superposed fractionation effects, the lower parts of the sills always being more magnesian that the upper parts.
As the Palisades Sill was by this time regarded as the classic example of insitu fractionation, I had considerable trouble defending my dissertation! In about 1970 I spoke to Dr Ken Walker who, using an electron microprobe, found two distinct generations of olivines, which indicated multiple injection again.
In about 1990, Mathew Gorring of Columbia University, did his doctoral thesis on the sill, and sent me his results which included some trace elements, (Ni, Co, Cr, Ba, Cu, Zr etc)
Mathew did two sections, but only of the lower 100ft across the olivine layer, at Fort Lee and at Alpine, N.J. The latter is somewhat more magnesian but the trends are the same. An oddity lies in the fact that which Ni/MgO and Co/MgO diagrams show the usual gentle curve, but Cr/MgO shows an inverted "V", the highest magnesian rocks having about half the Cr of the intermediate rocks(450 ppm vs 900), Clinopyroxene of course contains much higher Cr than olivine, but this is usually masked by the inclusion of Cr spinels. It seems any chrome spinel must have been dropped out before the intrusion of the olivine ledge.
Another PhD student, Shirley by name, sent me data in 1987 of a large number of ferro-basalts and granophyres terminating in highly fractionated rocks of FM#96, though only 62% SiO₂. Unfortunately he did not send any location data, but we now have a compositional range from 25% MgO down to about 0.2% and can produce quite repectable variation diagrams. As the only granophyre previously reported was a single thin late-stage dike, one would love to know where he found them.
Mathew Gorring also analysed some olivines, all very iron rich at Fo72 to Fo55, but the picrite trend in the sill does not 'point' to these olivines but to a more magnesian but still unusually iron-rich one of about Fo80. They must be the most iron rich 'picrites' known. One can only assume MG did not find the most magnesian olivines.
So it seems there are still some mysteries to be cleared up about the Palisades Sill. I have never seen the data of Gorring or Shirley in print or referred to, but presumably it rests in some archive at Columbia U.

Recently we found that Shirley's data did include REE which had been overlooked and though no Zr, Nb or Y, it does include Ta; so using the constant Nb/Ta = 15, we have added some Nb values.
The results show that the Palisades are formed of a typical Flood Basalt, rather like Grande Ronde in the CRB's, but less enriched in LILE than are the Ferrar Dolerites.
It is moreover a virtual match for the standard Centreville Diabase rocks, W-1 and W-2 even to the rather high Cs. Pb is not included in Shirley's data but is oddly high in W-1 accepted levels.

The Palisades Sill, (Shirley, 1987).

Greenland

The Skaergaard Intrusion is composed mainly of cumulates and chilled margins vary within inches. Producing a good chilled margin fingerprint is therefore difficult to impossible. This is probably the best developed and best known layered intrusion world-wide of peridotite, troctoliite, anorthosite, pyroxenite etc.

Inch scale banding in the Skaergaard Intrusion with olivine-clinopyroxene grading upwards into plagioclase-rich rock. Photo courtesy of Kurt Hollocher, (2001)

The Hebridean (British Tertiary) Basalts

From recent data these are certainly Continental Flood basalts and should be included here,

The scene from Sligachan Pub, Isle of Skye. On left is Mt Marsco, an odd high Fe microgranite (Marscoite) associated with the British Tertiary. On right is Ben Gillean, part of the Cuillins layered Gabbro intrsuion. Under the heather in the foreground are plateau basalts. A grand wee place! Through the gap is Loch Scavaig, where peridotite is exposed. There is a nice walking track to it across the heather, "Will ye gang to the Highlands with me?"

British Tertiary lavas at the Giant's Causeway County Antrim, Northern Ireland. (Rather Altered)

---

Layered and/or mineralised Intrusions of CFB type

Thick sills of 1000 - 1500ft such as the Peneplain Sill in the Ferrar Dolerites, show extensive "hidden layering" (i.e., progressive variation in Na/Ca, Fe/Mg, Rb/Ni with depth) together with some phase variation. As we have said above, orthopyroxene is usually not found in the upper half of a sill, while any olivine bearing rock is invariably near the base, but sills are not usually layered.

Layering consists of alternating concentrations of light and dark minerals, usually olivine-plagioclase, orthopyroxene-plagioclase, clinopyroxene-plagioclase, but also chromite-olivine, magnetite ilmenite-feldspar or even biotite or arfedsonite-feldspar have all been described. The bands may be sometimes on a scale of a few inches, and sometimes feet and is usually seen only in intrusions of considerable depth, often several km. The Dufek Intrusion in the Horlick Mountains of Antarctica, while of Ferrar Dolerite type, has a least 3000ft of anorthosite and layered rocks (Opx - plag) exposed, capped by 900ft of granophyre, (Ford, A, 1970, Geol.Soc. S.Af. Sp. Pub.1) . The total depth of this and many other intrusions is quite unknown.

Layered intrusions have been extensively studied and described, especially the Skaergaard intrusion of East Greenland. The books of Wager & Deer, (1939), and Wager & Brown, (1967), are the best known, though Wager himself wrote about 20 books and papers on the subject. Wager & Brown also discuss all layered intrusions world wide and are a great source of information on the Bushveldt, Stillwater and the layered subvolcanic intrusions of the British Tertiary, Mull, The Cuillins, Rhum, Ardnamurchan etc.

Layering is not commonly seen in Oceanic ridge formations, though gabbroic intrusions with phase separations and concentrations occur, the continental environment of lighter weight crustal rocks is much more conducive to the formation of deep intrusions. Alkaline rocks may also demonstrate, inch scale layering is seen in the "leuco-gabbro" of the alkaline Mount Royal Complex in Québec for example.

Layering appears to be always associated with convective overturn of a deep magma body. Cooling and crystallisation takes place mainly at the top, the lower regions are always hotter. The specific gravity of the upper region with a cooler magma, heavily laden with slowly sinking crystals increases until it becomes unstable, and the whole magma body overturns. This brings the crystals close to the base where they will settle out. As olivines are heavy and sink much faster than plagioclase, they will after overturn be some distance above the base, the feldspar settles out, almost pure until the olivine, now sinking from a hotter to a cooler environment, arrives, usually jacketed with a lower temperature, more Fe-rich rim. Large crystals of any type are more stable than tiny ones, and convincing examples are shown of adcumulous growth onto the crystals which have settled out on the more or less solid base of the intrusion until a layer may be almost pure feldspar or olivine. Finally disequilibrium becomes so great that there is a reversal to another phase, so both convective overturn and disequilibrium growth in which crystals continue to grow even though the magma is supersaturated for another phase, both seem to play a part.

Deep intrusions are also seemingly necessary for the occurrence of sulfide mineralisation. Wager and Brown, (1967) found that when the S content had reached about 400ppm immiscible pyrite (FeS) separated and proceeded to scavenge up all the chalcophile elements in the magma. In low silica, high Mg basic intrusions an early result will be the formation of the nickel-iron sulphide, pentlandite, or pure niccolite or millerite. Olivine or orthopyroxene reduces the Ni, Co content of a magma rapidly so one would not expect Ni sulfides to be formed below 5-6% MgO. However Cu builds up with fractionation, as does Zn to a less degree, so chalcopyrite, sphalerite may occur later.
Ni in the East Pacific Rise at 7.5% MgO is only 80ppm (Regelous, et al, 1999) and ORB glasses of 9% MgO have only 150ppm, eg, Danuchevsky, (2000). Cu and Zn rise to 75 and 150ppm at 5% MgO. However, in a deep intrusion, early olivines with 3000 ppm Ni may, on sinking into hotter magmas, be resorbed and do funny things to magma Ni levels, though I do not believe this has been reported.
Chilled basic Flood Basalts show similar levels of Ni, Co, Cu, Zn to ORBs, ie, about 100 ppm Ni, at 9% MgO, 50-60 ppm Co, with Cu - Zn rising to 75 and 150 pp at 5% MgO (Gunn, 1963), not a promising beginning for some of the largest Ni, Co, Cu deposits known.

The Sudbury Intrusion

Nothing stated here should be taken as engraved on tablets of stone. We have an on-going project of re-examining the Sudbury Intrusion, one of the world's largest producers of Ni, and a significant source of Co, Cu.
It is probably the most argued over intrusion in the world and the theories propounded range from the serious to the ridiculous. It lies NE of Lake Ontario and NNW of Ottawa, in a mainly Archaean terrane. The intrusion itself has been dated at about 1.7 billion years and is given a thickness now of greater than 3km. It is unlayered but has a basic orthopyroxne-bearing gabbro or norite, a more felsic norite, a micropegmatite zone, and some quartz diorites which appear to be related in most element trends though the diorites have double the Nb of the norites. Above lies 1400m of the Onaping Tuff, a fine black welded? rock, also rather altered as is the whole intrusion. It has a similar fingerprint to the underlying norite.
The norites have up to several percent of pentlandite, niccolite, millerite etc, and we will later append a list of the reported sulfides which is extensive.
Lightfoot et al, 1997, (Ont. Geol. Surv, Open File Report 5959) have assembled good ICPMS data for some of the rocks.
These data show normal amounts of Ni, Cu, Co to which have been added variable amounts of up to 10,000 ppm as sulfides as the excess metals correlate with the S and Te contents. There is little As data and less Se, Te than one might wish for.
One of the more fanciful theories long expounded is that Sudbury is the site of a nickel-iron meteorite impact, which shock melted the country rock into norite and vaporised leaving behind the excess nickel. Quite how this was transposed from nickel-iron metal to sulfide is not revealed, but on a visit once I was shown peripheral shock cones in the country rock, as found in meteorite impact craters. Later theories suggest that a meteorite caused both the formation of shatter cones and the shock-melting of the norite, though the similarity of the latter to Stillwater, the Bushveldt etc demands an impossible series of coincdences

Origins and Associations of the Sudbury Intrusion

	Sudbury basic norites containing variable secondary Ni and Cu sulfides. The negative Zr anomaly appears distinctive.
	Norites of the Creighton Mine, Sudbury, many also containing extra Cu, Ni.
	The Quartz Diorites, including to basic inclusions. There appears to be some mineralised lead, greater in amount in the more basic rocks.

The fingerprints above appear at first glance more potassic calc-alkaline than Flood Basalt. It is a low Ti type but with unusually elevated Cs, Rb, Ba, more so than the Ferrar Dolerites. The relative depletions of Zr and P are very similar to those seen in the island of Salina and others in the Aeolian Arc, though no Zr data is available for the quartz diorites.
However, Flood basalts and orogenic rocks have very similar fingerprints both originating in the enriched sub-continental mantle and a perfect discriminant has yet to be discovered. The more basic rocks of Sudbury have only 49-51% silica and include ferro-gabbros of 14-15% Fe2O3(t), and apart from the Columbia River CFB's, no ferro-basalts are found associated with andesites between the Canadian Border and Tierra del Fuego or indeed in any other island-arc or continental andesite.
We can assume with some confidence that the Sudbury rocks are of CFB origin, along with the British Tertiary, and Skaergaard intrusions, the Stillwater Intrusion, and possibly others such as the Muskox Intrusion of Northern Manitoba and the Dore Lake Layered Intrusion of Québec. The lack of layering in the Sudbury intrusion might be due to shape.
Preliminary calculations suggest the average amount of Cu-Ni present is greater than that in the average unmineralised rock but we do not at this point have a good estimate of the parental magma.
Watch this space!

The Onaping data is remarkable in view of the obvious alteration in hand sample, though each line represents an average.

---

Origins

Shown below is the normalised diagram for a mix of E-type MORB with 10-15% of the standard obsidian NIST 278 from Newberry Crater. The result is somewhat similar to a typical flood basalt with the low Ti, and Nb-Ta and elevated Cs, Rb, Ba etc. Lead was not included for the obsidian (Gladney,1992) but Pb is usually high in rhyolites, 25-30ppm, and would give the usual high Pb seen.

Obsidian/MORB mixtures.

Recently theories on the origin of Flood basalts tend to assume an origin at great depth, near the Earth's metallic core, forming a great mushroom shaped diapir. One might assume that any basaltic rock generated from such depths would be highly alkaline. The salient facts seem to be.

(1) Continental Flood basalts have a continental or crustal type signature identical to that of the andesite series.

(2) All members are highly fluid basalts, not andesites. While some members have high silica (56 to as much as 58%, these are plainly the product of fractionation.

(3) Different members may be related by orthopyroxene or OPX + Plag fractionation.

(4) In sills especially, Fe and Ti fractionation is well developed.

(5) The overlap of Fe with alumina and the high CaO are quite unlike andesites. High iron is a characteristic of Ferrar and Karoo members especially.

(6) All flood basalts are found in a continental environment where the crust has at sometime in it's history been subject to subduction and the sub-crustal mantle to enrichment.

(7) Individual members may vary, but single members of enormous volume (50,000 cub. km are quite homogenous. The post-initial emplacment of huge volumes of orthopyroxeneite shows the magma was held for long periods, very slowly cooling in a subcontinental chamber, probably above a spreading axis.

One can say somewhat tentatively that flood basalts are mantle generated in subcontinental environments where the mantle retains some enrichment from an older subduction phase. Because of the overlying continental roof of low specific gravity, large volumes of magma are generated without penetrating high in the crust. Tensional factures in the crust opening up fissures toward the surface followed by a short compressive phase forcing magma sometimes onto the continental surface would seem to be ideal. An incipient spreading centre developing under continental crust or a spreading centre being forced under a continent seems to be requisite..
Large scale contamination by crustal material is unlikely, The remarkable homogeneity of enormous volumes precludes it. Inspection of contacts once deep in the crust shows that granite and sediment are sealed off by a chilled barrier of basalt glass. Xenoliths are not seen to be even partly fused, though often recrystallised.