398
Earth and Planetary Science Letters, 104 (1991) 398-416 Elsevier Science Publishers B.V., Amsterdam
[CLI
Starting plumes and continental break-up R o b e r t I. Hill Research School of Earth Sciences, The Australian National University, PO Box 4, Canberra, A C T 2601, Australia Received May 14, 1990; revision accepted January 14, 1991
ABSTRACT Laboratory studies of fluid dynamic analogs of mantle plumes have led to important advances in our understanding of the life cycle of hotspots. Melting within the large volume heads of starting plumes gives rise to flood basalt provinces, while uplift of the surface above plumes results in horizontal deviatoric stresses that may precipitate considerable continental extension. However, for modern-day plumes the magnitude of these stresses appears incapable of actually initiating true continental break-up; rather, the rise of a new plume may lead to the local reorganisation of plate-scale motions (including ridge jumping), enhanced propagation of an existing ridge system, or provide sufficient extra force to drive a weak plate-scale system from slow spreading through to continental rifting. Thus, although plumes do not provide the ultimate driving force for continental break-up (these come from the plate-scale motions), the extra gravitational potential that they impose means that they play an important role in determining both where and when continental break-up does occur.
1. Introduction
The notion that hotspots are capable of weakening the lithosphere sufficiently to initiate or even control continental break-up has been around for a long time [1-3]. If correct, this carries the interesting implication that hotspots play an important or even dominant role in the development of oceanic spreading centres, and thus may exert a strong control on the movement of the crustal plates, and on the plate-scale convective motions within the mantle. Laboratory studies of fluid dynamic analogs of mantle plumes [4-10] provide the necessary background for assessment of the role of plumes in continental break-up. The most important observations to come from these experiments are that initiation of a new thermal plume (a "starting plume": see Fig. 1) begins with the development of a large-volume "head", that this head grows by entrainment during ascent, that a conduit (or "tail") carrying hot (and thus buoyant) source material to the top of the rising head may develop and persist for long after the starting plume has flattened and dissipated within the uppermost mantle, and that there may be considerable uplift of the surface above the rising plume head. This 0012-821X/91/$03.50
© 1991 - Elsevier Science Publishers B.V.
work has provided a dynamical framework which has been applied successfully to the origin of continental flood basalts [11,12] and Archean komatiites [13], and is now being applied to problems of continental reworking [14,15]. Here I combine the results of geological observations and the fluid dynamic studies with inferences drawn from two-dimensional modelling of gravitationally-induced extension [16] to consider in more detail the dynamics of the putative link between hotspots, continental flood basalts, and continental break-up. I have concentrated on the opening of the central and northern sections of the Atlantic Ocean for the simple reasons that these are the areas that have been best studied, and because comprehensive summaries of the geological histories are available [17,18]. The purpose of this paper is to attempt to determine what role plumes play in the development of oceanic spreading ridges. 2. Hotspots, continental flood basalts, and continental break-up
The suggested link between hotspots and continental break-up has recently been thoroughly reviewed by White and McKenzie [19], who docu-
STARTING PLUMES AND CONTINENTAL BREAK-UP
399
Before spreading Surface elevation ~
After s~reading
////////////~f
"/~
~.. continued ascent of hot source mantt]uri.~lup axial
1 cooler mate [ within the "head" [ of the starting plume
I
,
,
I • •
[ [
I L o t sourcematenal I
,O00 m
,
I
,
,
l
km
,
I
Fig. 1. Sketch showing the main features of the head of a starting plume before and after near-surface spreading [modified from 10,13]. Darker shading indicates higher temperatures. Upper panel emphasises the horizontal scale over which significant uplift can occur; maximum uplift is predicted to occur when the hot plume material displaces thick, cool continental lithosphere.
ment a third component of a possible trilogy, that of continental flood basalt volcanism. Four cases where continental rifting followed soon after a period of continental flood basalt volcanism are described in detail; these are the rifting of the Seychelles Platform from India, which followed closely after eruption of the basalts of the Deccan Traps; the opening of the North Atlantic (basalts of the Tertiary North Atlantic Province); the opening of the South Atlantic (Parafia-Etendeka); and the beginning of extension in the Red Sea (Ethiopian Traps). Continental flood basalt volcanism in all four areas has been linked to the initiation of new mantle plumes, respectively Reunion, Iceland, Tristan da Cunha and Afar [1,2,11,12,19]. However, not all new mantle plumes lead to the opening of a new ocean basin, with the best example being the plume responsible for the Siberian Traps [11,12]. Even when plume-initiated extension does proceed through to continental rifting
there may be a considerable delay between the arrival of the plume (most obviously signalled by the initiation of basaltic volcanism) and the beginning of spreading within the new ocean basin. For instance, although continental break-up did follow initiation of the Karoo flood basalt province of southern Africa 193 million years (Ma) ago, rifting apparently did not begin until 150-152 Ma ago (anomaly M21 or M22 time), over 40 Ma after arrival of the new plume head [20]. In such a situation the exact nature of the link between the plume and continental break-up is unclear. Even where the rise of a plume can be linked to the initiation of oceanic spreading, it is not clear that the situation is best described as "continental break-up". For instance, I will argue below that for the North Atlantic the spreading Iceland starting plume intersected an existing oceanic ridge system in the Labrador Sea, and that the opening of the North Atlantic is better described as ridge jumping rather than as true continental break-up.
400 By way of contrast, the opening of the South Atlantic appears to have resulted from the rapid propagation of existing spreading ridges into the tensional environment created by the rise of the P a r a h a - E t e n d e k a starting plume. The rifting of the Seychelles Platform from India following the arrival of the Deccan starting plume about 67 Ma ago is also best described as an example of ridge jumping, with the spreading plume head intersecting an existing ridge system in the Indian Ocean. Finally, although the opening of the Central Atlantic came after a prolonged period of continental extension, the rise of a new plume possibly acted to trigger the rifting event. In fact, there appear to be no clear examples over the past 250 Ma where the rise of a new plume has resulted in the rapid initiation of a new episode of continental break-up; instead, plume emplacement has resulted in the local rearrangement or acceleration of a pre-existing spreading regime, or has allowed a ridge to propagate more rapidly than would otherwise have been the case. The fact that rifting, when it does occur, is not randomly distributed but tends to follow relatively young orogenic belts has been noted by a number of authors [21,22], and this has been ascribed to local variations in the pre-rift strength of the continental lithosphere rather than to mantle convection [22,23]. Examples include the opening of the North Atlantic, which closely follows Palaeozoic mobile belts (see below), and the rifting of Antarctica from Africa, which was largely confined to Late Proterozoic mobile belts [15]. A final observation of importance is that the production of anomalously thick sequences of tholeiitic basalts such as those found within flood basalt provinces almost certainly requires their derivation from abnormally hot mantle [24,25]. 3. Uplift above a plume The thermal houyancy that causes plume material to rise results also in uplift of the superjacent surface, and the amount of uplift can be calculated provided the mean temperature difference between the plume and ambient mantle is known. For simplicity, this uplift can be considered in three separate parts, two of which relate to plume head provinces whereas the third is restricted to plume tail provinces. Within plume
R.I. HILL head provinces an initial uplift phase occurs above the centre of the plume while it is still rising [8,9], while a later phase of broader-scale uplift result as the plume head collapses and spreads laterally within the upper mantle [10]. Uplift above longlived plume tail conduits apparently results largely from injection of a wide ( - 1000 km) lens of hot plume material within the uppermost mantle [26,27]. I will restrict further discussion to plume head provinces. Elevation of the Earth's surface begins long before the rising plume head reaches the lithosphere, reaches a m a x i m u m and then declines slowly [8-10,12]. Significant uplift begins 10-20 Ma before the onset of basaltic volcanism (that is, when the top of the plume is still 500 km or more below the surface), and reaches a m a x i m u m when the plume is about 0.1-0.2 diameters from the surface, or when its leading edge is at a depth of about 150-200 km [9,12]. On the basis of a value for A T = +100°C, Campbell and Griffiths [12] estimate that the maximum uplift rate above the centre of the plume will be 20-40 m / M a , and that the m a x i m u m uplift of 500 1000 m should b e attained when the plume is still at a depth of 100 200 km, well below the depth at which significant melting begins ( - 70 km). An initial phase of uplift is thus predicted to have ceased by the time that voluminous basaltic eruption begins. Smaller values of AT ( + 50°C is considered more likely: see [12] and below) will give correspondingly smaller amounts of uplift. Campbell and Griffiths predict that uplift above the plume will be followed by subsidence as the plume head flattens within the uppermost mantle [12]. This is estimated to last for 20-60 Ma [12]. However, if the rising plume begins to replace lithospheric mantle the temperature variation between the plume and ambient lithosphere may increase considerably; the amount of uplift above a spreading plume head is thus sensitively related to the initial thermal structure of the overlying material. After spreading, the surface uplift resulting from emplacement of hot plume material into the uppermost mantle is given approximately by:
E=o~ A T L , where a is the coefficient of thermal expansion (taken as 3 × 1 0 5 oC 1), AT is the average ex-
S T A R T I N G PLUMES A ND C O N T I N E N T A L BREAK-UP
cess temperature anomaly (in °C), and L is the thickness of the anomalously hot layer. Griffiths and Campbell [10] estimate that the heads of starting plumes approximately double in diameter before the rate of lateral spreading decreases greatly. If the collapsed plume head has the shape of either a disc or a flattened ellipsoid, conservation of volume allows us to estimate that plume heads with initial diameters of 800, 1000 and 1200 km spread to form discs (values for flattened elipsoids are given in parentheses) of hot material 1600, 2000 and 2400 km across having maximum thicknesses of 130 (160), 170 (200) and 200 (240) km, respectively. These results come from experiments where spreading occurs in an environment of uniform viscosity [28]. The compositions of the most magnesian basalts derived from relatively recent plumes can be used to estimate that they come from source material up to 200°C hotter than ambient upper mantle [25]; however, most plume-derived basalts have much more ordinary compositions, implying lower source temperatures. This comes about because in order to rise or spread a plume must displace overlying or surrounding mantle, and this results in entrainment of colder material into the plume head [see 10]. The abundant tholeiites thus form by partial melting within the bulk of the relatively cool plume head, while the rarer picrites largely result from melting within the much hotter axial jet [13]. For purposes of illustration, I will use a value of 100°C for AT; although the actual value is likely to vary, for the modern Earth it apparently cannot exceed 200°C, the inferred temperature difference between the hottest axial conduit material and ambient asthenospheric mantle. The fluid dynamic experiments indicate that when it nears the Earth's surface a plume head originating at the c o r e - m a n t l e boundary will be comprised of 20-40% hot source and 80-60% cooler entrained material, implying AT = + 4 0 - 8 0 ° C relative to the asthenospheric mantle [12]. For the case where a plume head replaces cold lithosphere such as occurs beneath old cratons a value of AT = + 200°C might be more appropriate; more accurate estimates of AT depend on a better understanding of the mechanism by which the plume replaces lithospheric material than is currently available. For A T = + 100°C, the calculated maximum increase
401
in elevation is about 600 m for a large (2400 km diameter, 200 km thick) ellipsoidal plume head rising beneath continental crust, or about 900 m for the same plume rising beneath the oceans. Larger values of AT obviously increase these estimates; the observed average excess elevation at the centre of the East African Rift (which is here suggested to have formed above a starting plume) is about 800 m [29], consistent with the recent emplacement of a large ellipsoidal plume head 200 km thick and having a value for AT of - 130°C. Here the inferred topographic bulge is about 1400 ( E - W ) by 1700 ( N - S ) km, doming is estimated to have begun about 20 Ma ago [29], and it is possible that the plume head is still actively spreading (inferred from results in [10,12]). It is concluded that the maximum buoyancy-induced uplift that can be generated by unobstructed modern plumes is probably about 1000 and 1500 m for those rising beneath continents and oceans, respectively.
4. The relationship between uplift and extension In an important study, Houseman and England [16] used two-dimensional modelling to investigate the magnitude of horizontally-directed deviatoric stresses that could be induced by simple uplift of the Earth's surface. Although the details of their results are heavily dependent upon poorly understood parameters such as the rheology of Earth materials at elevated temperatures and pressures (see [16] for a discussion of the likely uncertainties), they calculated that for some combinations of input parameters uplift of order 1 - 2 km may result in horizontal deviatoric stresses of as much as - 1 0 0 and - 2 0 0 MPa within the crust and mantle, respectively. Stresses of this magnitude can result in the initiation of extension, and in fact these authors were most interested in delineating the conditions under which runaway extension did n o t occur. They inferred that there was a critical elevation (E~) below which extension was selflimiting and above which extension would lead to the formation of a new ocean basin. For the range of parameters used in their calculations E c varied from 800 m for a Moho temperature (TM) of 750°C to 2500 m for T M = 550°C (Fig. 2); likely uncertainties were estimated to be equivalent to a _+100°C uncertainty in the temperature. A smaller uncertainty ( + 50°C) has been shown in Fig. 2.
402
RI. HILL
come the strength of the crust and uppermost mantle except in areas where the crust and uppermost mantle is already anomalously hot ( T M > 700°C), or where extension driven by other processes is already taking place.
©
5. The response of the subcratonic lithosphere to the rise of a starting plume: evidence from southern Africa
O > O
O
v
500
600 700 Moho Temperature (TM), °C
800
Fig. 2. Graphical illustration of the relationship between elevation and Moho temperature predicted for the initiation of runaway extension (/3 = 1.47) by Houseman and England [16]. Solid circles represent the calculated conditions for their models; the width of the shaded band results from uncertainties in the temperature dependence of the rheological properties of rocks (see [16] for more detail).
These models have obvious application to the problem of the link between plumes and continental break-up. The results indicate quite explicitly that the tensional effects imposed by a rising plume may lead to crustal extension, and that progression through to continental break-up is dependent upon a complex array of factors including the amount of excess elevation induced by the plume as well as the thermal structure of the lithosphere prior to the arrival of the plume. In particular, hot lithosphere (and crust) will extend much more readily than will cold lithosphere (and crust). However, for most stable continental regions inferred Moho temperatures of < 500°C are well below those needed for runaway extension given the magnitude of the uplift that can be generated above axisymmetric starting plumes (maximum of - 1000 m). Note also that the maxim u m elevation is likely to be attained where the Moho is coolest, further limiting the ability of plume heads to initiate runaway extension. Plume initiation alone is thus unlikely to lead to continental break-up because the gravitationally induced horizontal stresses are insufficient to over-
The voluminous basalts of the Karoo Province of southern Africa are inferred to have resulted from melting within the head of a starting plume [2,11,12]. However, the patchy distribution of outcrops makes estimation of the original volume, areal extent, and location of the feeders of Karoo magmatism a difficult proposition. Nevertheless, if the basalt source regions extended beneath the old cratons, then the generation of what must have been a very considerable volume of basaltic magma 193 Ma ago would be expected to have been accompanied by a pronounced disturbance of the thermal regime within the uppermost mantle beneath the southern African cratons. The evidence from xenoliths brought to the surface in post-Karoo kimberlites indicates that this is not the case, and that relatively cold (and undisturbed) mantle lithosphere extends to depths of 200 250 km beneath much of the old shields [30,311. This is illustrated in Fig. 3, a diagram showing inferred pressure-temperature ( P - T ) conditions for xenoliths from Kaapvaal kimberlites and a " L e s o t h o " geotherm calculated using crustal model LCA of Nicolaysen et al. [32] and a chemically homogeneous mantle. Also shown are adiabats for mantle with potential temperatures of 1280 and 1480°C, the approximate m a x i m u m range inferred from the compositions of modern basalts [25], and the upper limit of the diamond stability field. For the simple steady-state model assumed in the calculation, the intersection of the geotherm with the 1280°C adiabat at - 2 4 0 km depth gives the approximate thickness of the Kaapvaal lithosphere beneath the generally diamond-bearing kimberlite pipes, and the excellent concordance between the model geotherm calculated from the heat flow and the xenolith P - T data indicates that there has been little change in the thermal regime of the sub-Kaapvaal litho-
STARTING
PLUMES AND CONTINENTAL
403
BREAK-UP
1800
1800 Zoneof Basalt Productionin ModemMantle
1600
Tp= 1480
&
-
1600
-
1400
OO 1400
I200
-- 1200 solidus
1000
--
~'~ 800
1000
800
©
~600
600
O OnCraton Frank Smith 116Ma Kimberly 84-88Ma NorthernLesotho 87-95Ma Premier 1200Ma
Lesotho 400
Graphite
•
200
Off Crat0n
Hebron E. Griqualand 0
q
I
0
50
I
I 100
I
I
150
200
0 250
Depth, kin Fig. 3. Temperature-depth diagram showing the thermal regime of the lithosphere beneath the Kaapvaal craton of southern Africa. Lesotho geotherm is calculated using the observed surface heat flow and crustal model LCA [32] and with the assumption of a chemically homogeneous mantle. Xenolith P - T estimates and position of graphite-diamond phase boundary are from [33]; dry solidus is from [25]; adiabats for potential temperatures (Tp; see [25] for definition) of 1280 and 1480°C are calculated using an adiabatic gradient of 0.6°C km 1 [25].
sphere since Cretaceous time. D a t a from xenoliths from the 1200 Ma Premier pipe show that there has, in fact, been little change in the thermal structure of the K a a p v a l lithosphere since Late Proterozoic time. The controversy a b o u t the reality or otherwise of the " k i n k " in the geotherm [e.g., 33,34] does n o t affect the c o n c l u s i o n that the lithospheric m a n t l e b e n e a t h the central part of the K a a p v a a l craton c a n n o t have u n d e r g o n e more than m i n o r thermal d i s t u r b a n c e d u r i n g the K a r o o event. T h e data from pipes m a r g i n a l to the craton ( H e b r o n , East G r i q u a l a n d ) do indicate the pres-
ence of c o n s i d e r a b l y hotter (and thus t h i n n e r ) lithosphere away from the craton, a n d these are the areas overlain by the vast bulk of the K a r o o basalts, a n d through which passed most of the reported feeders to the K a r o o m a g m a t i s m [35]. In summary, the bulk of the K a r o o basalts apparently were generated away from the c r a t o n or b e n e a t h the craton margins, a n d were e r u p t e d onto the y o u n g e r mobile belts ( i n c l u d i n g the Late A r c h a e a n L i m p o p o mobile belt) that s u r r o u n d the cratons. The m a i n K a r o o dyke swarms are also f o u n d along the craton margins, in the L e b o m b o
404 area and in the Limpopo mobile belt between the Kaapvaal and Zimbabwe cratons [35]. The P , T observations available from the xenolith data carry three strong implications. (1) The lithosphere beneath the craton is much thicker (perhaps as much as 100 km) and colder than that beneath the younger mobile belts marginal to the craton. The sub-cratonic lithosphere was little disturbed by the Karoo plume, whereas the lithosphere beneath the mobile belts marginal to the cratons may have undergone extensive modification. (2) Steady even spreading of a starting plume beneath the continent, as envisaged by [14], cannot have occurred beneath the Karoo since the xenolith data provide strong evidence that the Kaapvaal lithosphere is not the remnant of a proposed Karoo starting plume (the temperatures are simply not high enough). (3) It is highly unlikely that the Karoo basalts were derived from the Kaapvaal lithosphere because samples of that lithosphere show no evidence that the high temperatures required for basalt production ( > 1300°C) were reached at the appropriate time. This does not rule out the possibility that small volumes of H 2 0 - or CO2-rich melts formed at much lower temperatures than those of " n o r m a l " basalts were derived from the lithosphere immediately above the rising and spreading plume head. The xenolith data thus provide strong evidence that the rising Karoo plume was displaced laterally by the old, cold, and stiff Kaapvaal lithosphere, and instead preferentially displaced lithosphere from beneath the relatively young mobile belts surrounding the old cratons. The cold, stiff and possibly low density (because it is Si- and Mg-rich [36]) lithosphere beneath the old craton has thus served to protect it from both thermal and structural reworking associated with the rise of a major starting plume, and this may be an important factor in the preservation of Archaean shield areas. Some of the implications that arise from this conclusion will be discussed further below. 6. The sequence of events above a starting plume
The fact that, given a suitable environment, initiation of extension leading to rifting is depen-
R.1. HILL dent upon the temperature profile through the upper mantle and crust as well as on the amount of elevation leads to significant differences in the sequences of events above plumes rising beneath "cold" versus " h o t " lithosphere. Table 1 lists the expected sequences of events associated with the rise of starting plumes beneath stable continental crust having a "cool"thick ( - 200 km) lithosphere, and beneath crust that is already " h o t " and has a thinner ( - 120 km) lithosphere. Of particular importance for testing the model is the prediction that uplift will begin some time (10-20 Ma) prior to the initiation of voluminous basaltic magmatism [12], and this holds for both cases examined. By way of contrast, the relationship between the initiation of voluminous basaltic magmatism and the beginning of active rifting is sensitively dependent on the temperature of the uppermost mantle and overlying crust. If the lower crust and upper mantle are already hot, active extension may be apparent before eruption of basalts begins. However, where the plume rises beneath crust having a thick mantle lithosphere, it may take some time for sufficient heat to be transported by conduction from the top of the plume into the overlying mantle and crust in order to weaken them sufficiently for rapid extension to begin. For such a case, extension may postdate the beginning of basalt eruption by 10-20 Ma (or more for plumes that stabilize at greater depths). 7. The opening of the Atlantic
The opening of the Atlantic Ocean is of particular interest in attempting to assess the importance of plumes in continental break-up not only because it has been relatively well studied, but also because initiation of rifting has been episodic. Also of interest is that periods of rapid development of new spreading ridge occur shortly after periods of extensive basaltic magmatism. The initiation of spreading within the Central Atlantic 175 Ma ago [17] followed a period of considerable basaltic magmatism within eastern North America and possibly also within western Africa 200 Ma ago [37-40]; the rapid opening of much of the South Atlantic beginning 132 Ma ago may have been more or less coincident with eruption of the flood basalt sequences of the Paraha (South America) and Etendeka (southern Africa) [2,11,
STARTING PLUMES AND CONTINENTAL
BREAK-UP
19]; and the opening of the North Atlantic about 58 Ma ago [18] followed closely upon the eruption of basalts signalling the initiation of the Iceland hotspot 62 Ma ago [11,12,19]. Prior to the inception of seafloor spreading within the Atlantic Ocean, the area of what is now eastern North America and western Africa was within the interior of the supercontinent of Pangaea. Rifting was preceded by, and generally occurred within the limits of, a series of mostly non-marine sedimentary basins indicative of at least gentle subsidence and possibly slow extension prior to ocean formation. The best documented of these precursory basins are those containing the sedimentary rocks of the Newark Supergroup of eastern North America [41]; however, most summaries of the kinematics of the opening of the Atlantic have brief descriptions of their presence [17-19]. This widespread evidence for slow continental extension prior to rifting is here considered important because it provides evidence that the plate-scale motions were suitably arranged for break-up to occur before the arrival of any of the inferred plumes. What about the broader-scale tectonic framework that existed over the period during which the Atlantic Ocean formed? Arc-type volcanism has been inferred to have begun along the western coast of North America 190-200 Ma ago [42], while to the east of what is now the North Atlantic region lay the Tethys Ocean and an extensive landmass across much of Europe and northern Asia [43]. If plate tectonics are driven largely by cooling and subduction of oceanic crust [44], then it is unclear whether rifting within Pangaea both immediately prior to and following the opening of the Atlantic was driven directly by subduction of oceanic parts of the Pangeaean plate, or whether it was driven indirectly, by secondary motions associated with return flows driven by the subduction of other plates [45]. Proper assessment of these two possibilities requires a thorough evaluation of any evidence for subduction around the margins of Pangaea over the appropriate time interval, and this is beyond the scope of the present paper. 7.1. The North A tlantic
Events in the opening of the North Atlantic, and the essential elements in the rifting of Europe
405
from Greenland have been the subject of recent reviews [18,19]. Spreading in the North Atlantic (arbitrarily defined as north of the AzoresGibraltar Fracture Zone [18]) was in progress south of the G r a n d Banks by 126 Ma ago (the time scale of Kent and Gradstein [46] is used throughout), had propagated northwards to begin opening the Labrador Sea by 92 Ma ago (with northwards p r o p a g a t i o n involving rifting of the Early Archaean West G r e e n l a n d - L a b r a d o r craton), moved eastwards to begin a new rift between Greenland and Europe 58 Ma ago, and by 56 Ma had progressed northwards so that over 1000 km of the Lomonosov Ridge was active as well [18]. The total length of new ridge formed within the first few million years of spreading following transfer of the ridge across Greenland was in excess of 5000 km. Spreading in the Labrador Sea between Greenland and Canada was preceded by a long period of slow extension prior to the beginning of rifting, underwent a major change in direction 56-59 Ma ago, slowed down considerably about 50 Ma ago, and had stopped completely by 36 Ma ago [18,47]. Finally, when the episode of rifting that formed the present-day North Atlantic began about 58 Ma ago, break-up was almost completely confined to the Palaeozoic mobile zones between the old cratonic blocks (Fig. 4). On-land basaltic volcanism began in Britain and West Greenland 62 Ma ago and extended to the East Greenland region about 4 Ma later, about 58 Ma ago. Seafloor spreading also began to the east of Greenland at this time, concluded by [19] to be the peak of volcanic activity. The onshore basalts were extruded over shallow marine and freshwater sediments, and the basaltic sequences remained above sea level during most of their formation. Thicknesses reach 6 km or more on both the east and west Greenland coasts. At least some of the West Greenland basalts were extruded from an offshore source. Both the Greenland and European margins of the North Atlantic are characterised by thick sequences (to 8 km) containing seaward-dipping reflectors interpreted as thick sequences of basaltic extrusives [19]. These basaltic rocks overlie both thinned continental and oceanic crust, and have geochemical characteristics consistent with derivation from abnormally hot mantle. The average excess temperature required for the development
406
R.1. HILL
of the excess v o l c a n i s m has been e s t i m a t e d to be - 5 0 ° C [19]; the presence of relatively h i g h - M g b a s a l t s (including picritic varieties) within the b a s a l t i c sequences i m p l i e s local t e m p e r a t u r e a n o m a l i e s c o n s i d e r a b l y in excess of this value.
7.2. A model of dynamic lithosphere extension above a starting plume I have i n c o r p o r a t e d these o b s e r v a t i o n s into the m o d e l for the response of the lithosphere a n d crust to the rise of a new starting p l u m e o u t l i n e d in T a b l e 1. The m o d e l is most satisfactorily ill u s t r a t e d using an actual example, a n d I have chosen the rise of the I c e l a n d p l u m e b e n e a t h G r e e n l a n d a b o u t 63 M a ago with the s u b s e q u e n t o p e n i n g of the N o r t h A t l a n t i c (Figs. 4 a n d 5). Initial uplift p r i o r to 63 M a ago o c c u r r e d in r e s p o n s e to the rising h e a d of the I c e l a n d plume, a n d was d i s t r i b u t e d m o r e or less equally regardless of the tectonic age of the overlying l i t h o s p h e r e a n d crust (Fig. 5a). This initial p h a s e of uplift m a y have been u n d e r w a y b y 70 M a ago. However, once the p l u m e b e g a n to interact with the lithosphere b e n e a t h the cratons, p l u m e m a t e r i a l was deflected laterally to p r e f e r e n t i a l l y d i s p l a c e lithosphere from b e n e a t h the Palaeozoic m o b i l e belts (Fig. 5b). This resulted in the d e v e l o p m e n t of relative elevation differences within the u p l i f t e d areas, with the m o b i l e belts staying high while the o l d shield areas sank. By 63 M a ago the u p p e r m o s t p o r t i o n s of the now irregular top of the p l u m e had risen sufficiently high (to within - 70 k m of the surface) for p r o d u c t i o n of basaltic melts to begin. A s the p l u m e h e a d s p r e a d laterally it intersected the existing ocean ridge system to the west of G r e e n l a n d (Fig. 5b). I n c o r p o r a t i o n of this hot m a t e r i a l into the plate-scale flow resulted in the a n o m a l o u s l y thick sequences of basalts o b s e r v e d in W e s t e r n G r e e n l a n d . A l t h o u g h c o n c o m i t a n t uplift a p p a r e n t l y resulted ( p r o d u c i n g the s u b a e r i a l b a s a l t s d e r i v e d from an offshore source n o t e d above), it is inferred that this was less than the a m o u n t of uplift p r o d u c e d a l o n g the eastern coast of G r e e n l a n d which was closer to the centre of the p l u m e , a n d where the p l u m e was replacing relatively cold lithosphere. T h e result was that the g r a v i t a t i o n a l p o t e n t i a l of the elevated crust of the m o b i l e belts s e p a r a t i n g the G r e e n l a n d a n d
Fig. 4. Sketch geological map of the North Atlantic Province at anomaly 25 time (59 Ma ago), immediately prior to the opening of the North Atlantic. Palaeozoic mobile belts shown by diagonal ruling, areas of extensive 58-62 Ma volcanism by stipple, position of active ridge at 54 Ma by heavy solid line, and present-day 2000 m isobath by dashed line. The position of fragments of the Archean craton on either side of the Labrador Sea gives an indication of the extent of rifting prior to 59 Ma. Solid circle marks inferred position of hotspot at 62 Ma ago [59]; note that [19] places the hotspot position (at 54 Ma) about 200 km further to the south. Large circle indicates the approximate size of the plume head, and has a diameter of 2500 km [12]. Modified from [18,19,47].
E u r o p e a n c r a t o n s was greater t h a n that of the uplifted L a b r a d o r Sea, a n d e v e n t u a l l y this forced a shift in the p o s i t i o n of the s p r e a d i n g centre. D e c o m p r e s s i o n m e l t i n g within the p l u m e h e a d p r o d u c e d large v o l u m e s of tholeiitic b a s a l t as the u p p e r p a r t of the p l u m e h e a d p a s s e d t h r o u g h the " d r y " b a s a l t solidus at a d e p t h of a b o u t 70 k m ( c o r r e s p o n d i n g to Tp = 1330°C). Basalt p r o d u c tion was restricted to areas w h e r e the p l u m e m a t e r i a l c o u l d rise high e n o u g h for m e l t i n g to begin, a n d was thus localised b e n e a t h the m o b i l e belts or b e n e a t h the existing s p r e a d i n g ridge.
STARTING PLUMES AND CONTINENTAL BREAK-UP
407
TABLE 1 Summary showing how the sequences of events occuring during the rise of plumes differs according to the initial state of the lithosphere (from [9,10,12, this paper]) "Cool" lithosphere ( T M = 500 ° C)
" H o t " lithosphere (T M = 700 o C)
1. Beginning of surface uplift when the top of the plume is still > 500 km from the Earth's surface. For A T = 5 0 ° C (average within the plume head), m a x i m u m uplift will be of order 500 m, and uplift may continue for up to 20 Ma before any other surface expression of plume ascent is visible. Little or no extension at this stage. May be minor amounts of volatile-rich magmatism.
Beginning of surface uplift when the top of the plume is still > 1000 km from the Earth's surface. For A T = 5 0 ° C (average temperature within the plume head), the m a x i m u m uplift of 500 m, attained when the top of the plume is still at a depth of 100-200 km, may be sufficient to initiate considerable extension. For the hot case, initiation of extension will thus pre-date initiation of voluminous basaltic magmatism, although there may be minor a m o u n t s of early volatile-rich magmatism.
2. Beginning of spreading of plume head, and beginning of displacement of cold lithosphere. Uplift resulting from lithosphere displacement more than balances depression resulting from collapse of the plume head, so that the surface above the centre of the plume continues to rise. Overall area of uplift expands, to give ultimately an uplifted area up to 2500 km across 20-60 Ma after the initiation of spreading of the plume head.
Spreading of plume head may be associated with some initial decrease in surface elevation. However, as the plume begins to displace the lithosphere, uplift begins again. Elevation decrease may be accompanied by a short-term waning of the rate of extension.
3. Beginning of volcanism in a restricted area above the hot central conduit when the top of the plume reaches a depth of - 120 kin. Initial products are predicted to have picritic composition. Still no extension, although uplift continues.
Beginning of picritic volcanism as with the "cold" case. However, volcanism may occur during active tectonic extension and continued broad-scale uplift.
4. Beginning of areally extensive basaltic volcanism when the top of the plume is - 7 0 km from the surface. Uplift continues as the plume is still displacing cold lithosphere.
As for the "cold" case, except that rapid extension m a y be occurring.
5. Plume ascent volcanism.
ceases. Waning of
Plume ascent finishes and uplift ceases. Continued extension, possibly accompanied by minor basaltic volcanism.
6. Heat conduction into the crust from the top of the plume can weaken the uppermost mantle and lowermost crust sufficiently that after a delay of 2-10 Ma (depending on the distance between the top of the plume and the Moho) rapid extension may begin. If the plate-scale arrangement is appropriate, extension may continue through to sea-floor spreading.
Conduction of heat into the upper mantle and lower crust can lead to almost immediate runaway extension (given an appropriate plate-scale stress environment), and, after a delay of 1-5 Ma, to voluminous crustally-derived anatectic m a g m a t i s m if appropriate sources are available.
7. If sea-floor spreading begins quickly, early basaltic magmatism will come from a source with AT = + 50 ° C relative to normal ridges. The first basalts produced at the new ridge system will post-date the beginning of melting within the plume head by 2-10 Ma.
If extension leads to sea-floor spreading, early basalts will come from hot ( A T = + 50 o C) mantle.
8. Continued conduction of heat into the crust can lead to the production of crustally-derived anatectic magmas. This can occur regardless of whether extension leads to rifting.
An extensive period of bimodal volcanism can continue, regardless of whether rifting occurs.
9. Slow cooling of the plume head (on a time scale of > 100 Ma) leads to a long-lived period of thermal subsidence.
Slow cooling of plume head leads to a long-lived period of thermal subsidence.
Where basaltic
relative liquids
areas downslope
finishes, and
elevation flowed onto
uplift
differences from
developed,
high-standing
the old cratons.
vent
Conduction the plume
of sufficient
heat
to raise the Moho
ciently for runaway
extension
from
the top
temperature
of
suffi-
to begin took a few
408
R.1. HILL up]ift
Labrador Sea Canada
~
Paieozo]c mobi]e be]ts
Greenand
melting
II,l
Europe
I ~ M m e l t i n g
-
:iiii
...... I i
transfer of plate-scale motions Melting of plurne material within new ridqe system
Atlantic Ocean
iiiii!iiiiiiiiiiii;iiii;ii......... i ii (d)
s!ow formation of new 'lthosphere
Incorporated into lithosphere
Fig. 5. Diagrammatic cross section showing the inferred stages in flood basalt production and continental rifting in the North Atlantic following arrival of the Iceland plume. (a) Long wavelength doming as top of plume begins to flatten upon reaching the lithosphere beneath the Greenland shield regions. (b) Sideways displacement of hot plume material away from cold, thick shield lithosphere into the areas beneath the Palaeozoic mobile belts, and into the existing ridge system in the Labrador Sea. This results in differential uplift of mobile belts, and excess volcanism to the west of Greenland. (c) Movement of plate-scale convective system across Greenland, with initial volcanism in the new spreading centre resulting from movement of hot plume material upwards through the melt zone beneath the new ridge. (d) Slow subsidence of the ridge as the hot plume material cools, ultimately yielding the North Atlantic Ridge as it is seen today.
million years. Estimates as to how long this would take vary from 2 - 1 0 to Ma, depending on the assumptions made about initial plume and Moho temperatures, and the depth to the top of the
plume head. Break-up was restricted almost completely to the young mobile belts for two reasons. First, the intrusion of hot plume material beneath the mobile belts (but not beneath the cratons)
409
S T A R T I N G PLUMES A N D C O N T I N E N T A L BREAK-UP
resulted in substantial and sustained uplift, possibly in excess of 1000 m. Second, heat was transferred by conduction from this hot material into the uppermost mantle and crust, thereby changing the rheology and lowering the critical elevation at which runaway extension could begin (Fig. 2). The preferential replacement of the lithosphere beneath the mobile belts with hot plume material thus resulted in both the elevation needed to initiate extension, and the changes to mantle and crustal rheologies that allowed extension to continue. Once reorganisation of the plate-scale convective system had been completed, with the ridge moving to the east of Greenland, a second mechanism for moving hot plume material upwards into the zone of melting became available. I interpret the early volcanism (58-62 Ma) as resulting from melting within the rising plume head, while the slightly younger ( - 5 8 Ma) voluminous flood basalt volcanism that produced the unusually thick basaltic sequences of the thinned continental margins and early formed oceanic crust resulted from decompression melting of plume material as it was moved upward by the newly established plate-scale flow regime (Fig. 5c). Finally, as the remaining, unmelted plume material cooled and extension continued, the area of the new ocean basin began to subside, leading ultimately to the North Atlantic margins seen today (Fig. 5d; see [48,49] for an analysis of the physics of this problem). Thus, when viewed in the regional context, arrival of the new plume head resulted only in a minor rearrangement of an existing plate-scale convective system that was systematically opening a new ocean basin from the south. The fact that the Lomonosov Ridge, north of Greenland and well outside the region affected by the plume head (Fig. 4), became active simultaneously with the transfer of the plate-scale motions across Greenland indicates that the plate-scale convective system was organised appropriately for continental break-up to occur. It is important to consider the rifting of Europe from Greenland in context, that is, that it resulted from ridge jumping near the active end of an existing propagating rift system. The opening of the North Atlantic has thus resulted from the coincidence of five factors: an existing plate-scale convective system in which
rifting was already underway; the possibility for rapid rearrangement of this system as a result of the development of an area with greater gravitational potential above the rising starting plume; the presence of appropriately aligned young mobile belts with lithosphere susceptible to thermal erosion by hot plume material; the production of horizontal deviatoric stresses by uplift of the surface above the introduced plume material that led to extension within the young mobile belts; and the thermal weakening of the upper mantle and crust of the mobile belts that resulted from emplacement of hot plume material. 7.3. Central Atlantic region
In detail, the model developed above applies only to the situation where a rising and spreading plume head intersects an existing spreading ridge. The probability of this happening is small; in general, rising starting plumes are unlikely to rise under an existing spreading ridge. Such has been suggested to be the case for the opening of the central section of the Atlantic Ocean, for which it has been concluded that the new spreading ridge may have intersected the extreme edge of a contemporary mantle plume, and that the observable effects from this intersection are minimal [19]. It is thus worth examining the opening of the central Atlantic as a possible example of the development of a spreading ridge unaffected by a plume. 7.3.1. Early extension and basaltic magmatism within eastern North America The stratigraphy of the Newark Supergroup has recently been revised [50]. Sedimentation began in the more southerly basins in North Carolina and Virginia in Late Ladinian time (about 230 Ma ago) and progressed northwards, so that deposition in the Hartford and Deerfield Northfield basins of Connecticut and Massachusetts did not begin until Late Carnian time ( - 225 Ma ago). The youngest sediments (of Pliensbachian and Toarcian age: - 1 9 3 Ma) are found also in the more northerly basins. With the exception of the Culpepper Basin of Virginia and Maryland, sedimentation in the more southerly basins had ceased by Middle Norian time ( - 210-215 Ma). The oldest igneous rocks (basalts) appear in early Hettangian time. Although two groups of
410
dates (at - 1 9 5 and - 1 7 8 Ma) have been reported for the basaltic rocks of the Newark Supergroup [39,51], Sutter [40] has recently concluded that the younger group results from hydrothermal resetting of the K-Ar systems in these generally extensively altered diabases, and the best age of basaltic magmatism is consequently that given by U-Pb and 4°Ar/39Ar dates of 200-202 Ma [40,52]. A short time interval for magmatism, perhaps only 0.6 Ma [53], is consistent with restriction of basaltic volcanism to Hettangian time, the age limits of which need to be adjusted slightly in light of these new dates [40]. The combined stratigraphic and geochronological evidence indicates-that basaltic magmatism was restricted to a short time interval 201 +_ 2 Ma ago. The North American basaltic rocks of earliest Jurassic age have been suggested to be a fragment of a once more extensive basaltic province that has been disrupted by the formation of the Atlantic Ocean [37]. Two features of these probable Early Jurassic basalt provinces (Fig. 6) are that they extend considerable distances inland, and, as noted by May [37], the associated dikes form a radial pattern centred on an area to the east of Florida.
Fig. 6. Pre-drift distribution of Late Triassic and Early Jurassic diabase dikes around the Central Atlantic. Circle with a diameter of 2500 km is partly illustrative, as the exact position of the postulated hotspot is not know. Modified from [37]; inferred distribution of subsurface basalts (diagonal shading) in the southeastern United States from [60]. Stippled area shows post Early Jurassic sedimentary rocks.
R.I. HILL
Both geochronological and palaeomagnetic data from basaltic rocks in Liberia and Morocco indicate that they are probably of similar age to the American occurrences [38], although the possibility t h a t b o t h the g e o c h r o n o l o g y and palaeomagnetic systems of the African rocks have undergone similar post-crystallisation modification to that recorded by their American counterparts [40] means that much of the data should be used with caution.
7.3.2. Seafloor spreading within the Central Atlantic The geology pertinent to the opening of the Central Atlantic region has been reviewed and summarised by Klitgord and Schouten [17], who conclude that seafloor spreading was initiated about 175 Ma ago, more or less simultaneously from a position south of Florida northwards to the Gibraltar Fracture Zone, a distance of about 2700 km. An unknown length of ridge south and west of this (in the Caribbean region) is also inferred to have been initiated at this time, in order to link Atlantic spreading through to a postulated transform fault (the M o j a v e - S o n o r a megashear) through Central America. Once begun, Central Atlantic spreading progressed steadily until about 132 Ma ago, when the ridge system began to propagate northwards towards the Rockall Trough southeast of Greenland. A little later, about 118 Ma ago, the initiation of rifting between western Africa and South America resulted in the establishment of a single throughgoing ridge system that extended the length of the Atlantic. Thus, with minor rearrangements, spreading within the Central Atlantic which began 175 Ma ago has continued up until the present time, and an initial spreading system has propagated both northwards and southwards to evolve into the current spreading system about 57 Ma ago. 7. 3. 3. Evidence for a possible precursor plume The possibility that the opening of the Central Atlantic may have been at least influenced by hot material emanating from a mantle plume was discussed by White and McKenzie [19], who concluded that although there was some evidence for plume involvement, it was small. They concluded that " t h e central North Atlantic margins are an
STARTIN(~ PLIJMES AND CONTINENTAL
411
BREAK-UP
example of a continental split above asthenosphere that was just a little warmer than normal", and suggested that the plume involved may have been the Cape Verde hotspot, inferred by extrapolation from [54] to have lain about 1000 km northwest of the United States east coast at this time. It is worth remembering that these authors [19] are probably here thinking in terms of the long lens of material emplaced into the upper mantle by flow of material up a narrow conduit (a plume tail or hotspot track) rather than the head of a starting plume. The main reason for the conclusion that any plume effects were minor comes largely from the lack of evidence for thick sequences of early basalts within the early rift sequences of the central Atlantic. The identification of features related to the earliest stages of rifting is difficult beneath the thick sediment sequences that characterise the North American margin, and even more difficult beneath the extensive salt deposits of the African margin. However, even after allowing for this, there appear to be far less basalts within the early parts of the central Atlantic than there are within the northern Atlantic. Three possibilities must be considered. The first, that the basalts are there but are unrecognised is considered unlikely. Resolution of the s e c o n d - - t h a t the basaltic liquids were trapped beneath the sediments and salt of the pre-rift b a s i n s - - i s difficult to evaluate. The third and favoured possibility is that the basalts are not there [19]. This last conclusion is in apparent conflict with the evidence reviewed above for a brief but extensive episode of basaltic volcanism 201 + 2 Ma ago within what became the margins to the central Atlantic only 25 Ma later. The areal extent, short duration, and geochemistry and isotope geochemistry of this basaltic event [55,56] are in combination characteristic of plume-related basaltic volcanism [12,15]; further, the presence of an inferred hotspot track (Fernando de Noronha) in approximately the right place at the right time [2], adds support to the starting plume interpretation. Finally, the time gap between the cessation of sedimentation in the southern Newark basins 210-215 Ma ago (interpreted as signalling the beginning of uplift), and subsequent basaltic magmatism at 201 Ma is within the range predicted by the plume model.
Two important questions need to be answered before the exact nature of the link between the " N e w a r k " plume and opening of the Central Atlantic can be properly determined. Why is there a time gap of 25 Ma between plume-related volcanism and initiation of oceanic rifting? Why are there no thick sequences of early-rift basalts resulting from introduction of hot plume material into the ridge melting system? The second question has a fairly simple answer, whereas the first is more difficult to resolve at this time. For the north Atlantic case discussed above the rising plume "captured" an existing spreading ridge, and hot plume material was being processed within the newly reorganised ridge system within a few million years. However, within the central Atlantic the 25 Ma of slow extension that ensued between arrival of the plume and initiation of seafloor spreading could have resulted in considerable attenuation of the lens of hot plume material, so that only a small volume of plume material was passed through the ridge melting system before it was replaced with underlying asthenospheric mantle. In general, where a lengthy period of slow extension occurs between plume arrival and seafloor spreading, thick early-rift sequences should not be expected. What caused the delay? Although a number of possible factors can be suggested (cold lithosphere, need to rift across cratons in northern South America western Africa, presence of a relatively small or cool plume, magnitude of stresses resulting from the plate-scale motions, presence of an east-dipping subduction zone on the western side of Pangaea), the relative importance of these or other possibilities cannot be assessed at this stage.
7.4. The South Atlantic
Spreading in the south Atlantic began 132 Ma ago, with southern Africa beginning to separate from South America over a 5000 km or greater distance from the vicinity of the Falklands Plateau to as far north as the Benu6 Trough [17]. The final rifting of West Africa from the northern part of South America is inferred to have begun in anomaly M0 time (about 118 Ma ago) with the rapid establishment of a ridge system southwards
412
from the Caribbean to the Benu6 Trough, a distance of over 2000 km [17]. A number of workers have noted that the episode of rifting that led to the opening of the south Atlantic as far north as the Benu6 Trough was immediately preceded by eruption of the ParahaEtendeka basalts [2,11,19]. Parallels have been drawn with the arrival of the Iceland plume and the opening of the North Atlantic [11,19], and it is suggested that the model developed above and illustrated using the North Atlantic example is applicable to the South Atlantic, with the modification that the tensional area above the plume provided an environment into which the existing ridge system could propagate with relatively little impediment. Once the central and southern Atlantic parts of the rift were established, the geometry of the system apparently required them to join, even although this required the rifting of an old craton. This was accomplished - 118 Ma ago. 8. Discussion
The model presented above provides a simple physical explanation for the observed link between plume initiation, flood basalt volcanism, and continental extension. It also provides an explanation for the observation that rifting associated with plumes (and thus, potentially, flood basalt provinces) tends to be localised within relatively young mobile belts rather than split ancient cratons. This leads to the conclusion that rifting that does split old shield areas either results from another process, or is required by the geometry of extension initiated within neighbouring but unconnected mobile belts. Two separate classes of models now exist for the origin of flood basalt provinces. In the starting plume model, basalts are formed by decompression melting within the actively rising head of an anomalously hot mantle starting plume [11,12]. In the competing passive rifting model of White and McKenzie [19] the basalts are apparently formed (again by decompression melting) when hot plume material previously injected into the uppermost mantle through a hotspot conduit (or plume "tail") passes beneath either an active spreading centre or an area already undergoing extension but which has not yet proceeded through to continental
R.I. HILL
break-up, or when a new plume rises into one of these two environments. Flow of plume material into the melt zone is accomplished passively, within the plate-scale flow. Although White and McKenzie note that uplift of - 1 0 0 0 m is commonly observed above hotspots, and that the gravitational potential thus developed is significant and will greatly assist rifting [19], these authors do not develop this further. Rather, they develop a model of melt generation during externally imposed rifting, and thus do not consider the origin of the driving forces that result in basalt formation, continental rifting and oceanic spreading. The model developed in this paper is to some extent a hybrid one, in that it allows for basalt production from hot plume material by the imposed rifting mechanism after uplift-driven extension has forced local rearrangement of plate boundaries, or the acceleration of plate-scale motions. As most plumes do not rise adjacent to oceanic ridges (but may still yield flood basalt provinces, albeit smaller ones), the production of the huge volumes of basalt that result when this coincidence occurs is truly remarkable, and leads to the apparent relationship between flood basalt volcanism and continental break-up remarked upon by previous authors [2,19]. The fundamental problem of whether the extension that is sometimes observed to coincide with the development of flood basalt provinces is internally or externally generated has important implications in the application of the two classes of models to continental break-up. If extension is externally imposed, as is the case for the White and McKenzie model [19], then there is no obvious reason why there are flood basalt provinces that are not related to the formation of ocean basins. The externally imposed rifting model, for instance, provides no obvious explanation for the production of the basalts of the Siberian Traps, where hot plume material apparently underwent considerable partial melting beneath a stable continent. Other relatively recent situations where flood basalt eruption has not been followed more or less immediately by continental break-up include the Karoo of southern Africa, the FerrarTasmanian province of Antarctica and Australia, and the Columbia River Plateau of western North America.
STARTING PLUMES AND CONTINENTAL
BREAK-UP
A further problem with the externally imposed rifting model is that in certain situations it is incapable of producing, within a reasonable period of time, the huge volume of hot source material required by the more extensive volcanic provinces. For example, the volume of source material required to yield the - 1 0 7 km 3 of volcanic products within the early stages of the opening of the North Atlantic [19] must be 5 x 1 0 7 km 3 or more. For the situation where rifting intersects a pre-existing plume swell, there are two ways of estimating the time required to establish the necessary source volume. The first uses estimates of the flow-rate of material up modern hotspot conduits [12], and yields a time of 8 Ma (for the current Iceland flux of 200 m 3 s 1 [27]) provided that all of the hot material produced within this time actually passes through the region of melting beneath the new ridge system. This is considered unlikely. The second approach is to use the development of Iceland over the past 15 Ma to estimate directly the amount of basalt produced by injection of hot plume material into an ocean ridge system. At the estimated production rate of 0.02 km 3 yr-~ (0.6 m 3 s 1) [11], it would take 500 Ma to generate the source volume for the North Atlantic Province basalts. After appraisal of the uncertainties and assumptions contained within these two estimates, it is suggested that a realistic time scale for the formation of a plausible source volume is probably at least a few tens of million of years. The alternative is that the basaits form by melting of material carried into the lithosphere within the head of a starting plume [11,12]. White and McKenzie do mention, in passing [19, p. 7689] that the start of a new plume is associated with the arrival of a large blob containing about 5 X 1 0 6 km (diameter: d = 230 km) of material 100 200°C hotter than "standard" mantle. Such a blob could conceivably yield - 1 0 6 km 3 of basalt, an order of magnitude less than they infer to have erupted within the North Atlantic province during the initial stages of rifting. If the source of the hot material that resulted in enhanced volcanism within the North Atlantic was the new Iceland plume, as envisaged by them, then it is simple to calculate that the plume head must be an order of magnitude larger than their estimate (that is, at least 5 × 107 km3: d = 500 km). Even this estimate assumes that all of the plume material is advected
413
through the melt zone beneath the ridge, which, given the geometry of the system (see Fig. 5b and c), is highly unlikely. If enhanced volcanism within the North Atlantic is related to initiation of the Iceland hotspot, as is strongly suggested by the geological evidence, then the only viable model for the production of the hot source layer that satisfies the volume constraints is that of Campbell and Griffiths [12], who propose that volcanism results from melting within a rising plume head with a diameter (before spreading) of - 1 0 0 0 km (volume: 5 × 108 km3). As outlined above, the arrival of a plume head of these dimensions is likely to have other consequences, not least being the development of sufficient gravitational potential within the overlying crust to result in important local (1000 km scale) reorganisation of plate scale motions. Further, the evidence from the Atlantic suggests that this extra gravitational potential may be sufficient to actually move an area from one of slow plate scale driven extension to one of active rifting, presumably with the gravitational potential added to the forces that were already resulting in weak pre-rift extension. It is concluded that the plume model outlined in this paper and elsewhere [10,12] is an improvement over the externally imposed rifting model [19], not only because the former is based on quantitative dynamical models, but also because the latter is in conflict with a number of important geological observations.
8.1. The relative importance of plumes and platescale motions It has been already observed that the arrival of a plume does not by itself a new ocean make. What mechanism then, is responsible for "true" continental break-up? For example, what caused the initiation of rifting within the central Atlantic, the break-up of Antarctica and Australia, or the drift of Madagascar from southern Africa and then both away from Antarctica? The situation is perhaps most clear for the central Atlantic, where slow extension had been in progress for some time (possibly 40 Ma or more) before the initiation of break-up about 175 Ma ago [17]. As with the North Atlantic 115 Ma later ( - 60 Ma ago), final rifting through to the initiation of steady spread-
414
ing was preceded by the eruption or emplacement of basaltic magma over an approximately circular area about 2500 km across [37], which I interpret as a possible starting plume province associated with initiation of the Fernando mantle plume and hotspot track (see [2]). The opening of the Central Atlantic thus bears many similarities to the opening of the North Atlantic--weak and perhaps not well organised plate scale motions that were already resulting in slow extension were transformed into full oceanic spreading ridges by the arrival of a plume, although after a much longer time interval (25 Ma rather than 5 Ma). These observations focus attention on the plate-scale convective system as being the dominant force that ultimately drives plate motions. This system is responsible for most of the heat loss from the Earth [44] and recent dynamical models of mantle convection demonstrate that it is driven largely by cooling of the upper boundary layer, that is, through the cooling and sinking of the oceanic plates [26,44]. Continental extension apparently can be initiated in the middle of plates which contain no sinking portion to drive directly plate motions, and the geological evidence is that the formation of new hotspots alone do not initiate continental extension, although they play an important role in determining the exact position of break-up when it does occur. The observation that continents without attached subducting oceanic crust (such as Africa today, or possibly North America-Europe in Jurassic time) may still rift implies an important role for secondary motions in the overall dynamics of plate movement, that is, that all plate-scale motions are induced by flow resulting from the sinking of a relatively few and often distant cold slabs (see [45]). If gravitational potential derived from buoyancy-induced uplift of the Earth's surface is important in driving tectonics at the scale of - 1000 km or so, as is suggested here, then there may be situations apart from those related to the development of spreading ridges where evidence may be found to further support this model. Certainly, the gravitational potential of overthickened continental crust has been suggested to be an important factor in the development of the tectonics of continental margins [55,56]; conceivably, the addition of both volatiles and melts, and the upward movement of heat contained within magmas rising
RA. HILL
above subducting slabs could provide sufficient gravitational potential to result in the splitting of the overlying arc, with the consequent formation of back-arc basins. 9. Conclusions Although there is a common link between the initiation of a new mantle plume and the presence of flood basalt provinces, in detail the further link with continental break-up suggested by White and McKenzie [19] is at best weak. Although the rise of a plume head beneath a region already undergoing extension driven by plate-scale motions commonly does result in either rapid acceleration of the tensional regime through to the opening of a new ocean basin (e.g., Central Atlantic, South Atlantic), or the start of a new spreading ridge following ridge-jumping (North Atlantic, Carlsberg Ridge), there are sufficient examples of flood basalt provinces not associated with ocean basins (Siberian Traps, Columbia River, Karoo) to be able to argue that plumes alone do not initiate continental break-up [see also 19]. This is understandable in terms of the combined results of the models of H o u s e m a n and England [16] and the starting plume models of Griffiths and Campbell [10], which show that the magnitude of the uplift above a plume head rising unobstructed beneath a continent (maximum of - 1 0 0 0 m) may produce sufficient gravitational potential to force rearrangement of local tectonics but is insufficient to lead to runaway extension. These models also allow a simple explanation of why ridges form along young mobile belts rather than split old cratons; as well as producing uplift that results in extension, emplacement of hot plume material beneath the mobile belts leads to thermal weakening of the overlying lithosphere and crust, increasing the amount of rifting that will result from a given amount of uplift. The recognition of the important role played by old subcontinental lithosphere in channelling a spreading plume head provides a ready explanation as to why continental break-up may occur some distance from the centre of the initiating plume. The apparently near-complete deflection of the rising Karoo plume away from the old cratons to beneath the younger surrounding mobile belts is particularly instructive, as later con-
STARTING PLUMES AND CONTINENTAL
415
BREAK-UP
t i n e n t a l b r e a k - u p o c c u r r e d in t h e p l u m e - a f f e c t e d mobile belts rather than by splitting of the old c r a t o n s . T h e r e s i s t a n c e o f old, t h i c k c o n t i n e n t a l l i t h o s p h e r e to p e n e t r a t i o n by rising p l u m e material m a y have t e n d e d to p r o t e c t A r c h e a n c r a t o n s f r o m Phanerozoic
reworking.
Deflection
of
plume
material away from sub-cratonic lithosphere may act
to increase b o t h
maximum
the
areal
extent
and
the
t h i c k n e s s o f h o t p l u m e m a t e r i a l as it
s p r e a d s b e n e a t h c o m p l e x c o n t i n e n t a l crust.
Be-
cause the length scale of m o s t old c r a t o n i c blocks ( 1 0 0 - 1 0 0 0 k m ) is r e l a t i v e l y s m a l l w h e n c o m p a r e d with the s p r e a d w i d t h of starting p l u m e s (up to 2500 k m ) , c r a t o n s m a y b e p r e s e r v e d as " r a f t s " within
areas
of
thermally
and
t e c t o n i c a l l y re-
worked younger crust.
Acknowledgements The ideas presented here have been developed d u r i n g d i s c u s s i o n s o n t h e r o l e o f m a n t l e p l u m e s in tectonic processes with my colleagues Ian Campbell, G e o f f D a v i e s a n d R o s s G r i f f i t h s .
References 1 W.J. Morgan, Convection plumes in the lower mantle, Nature 230, 42-43, 1971. 2 W.J. Morgan, Hotspot tracks and the opening of the Atlantic and Indian oceans, in: The Sea, Vol. 7, C. Emiliani, ed., pp. 443-487, Wiley, New York, N.Y., 1981. 3 R.D. Hyndman, Evolution of the Labrador Sea, Can. J. Earth Sci. 12, 1041-1045, 1973. 4 J.A. Whitehead and D.S. Luther, Dynamics of laboratory diapir and plume models, J. Geophys. Res. 80, 705-717, 1975. 5 J.N. Skilbeck and J.A. Whitehead, Formation of discrete islands in linear island chains, Nature 272, 499-500, 1978. 6 R.W. Griffiths, Particle motions induced by spherical convective elements in Stokes flow, J. Fluid Mech. 166, 139159, 1986. 7 R.W. Griffiths, Dynamics of mantle thermals with constant buoyancy or anomalous internal heating, Earth Planet. Sci. Lett. 78, 435-446, 1986. 8 P. Olson and I.S. Nam, Formation of seafloor swells by mantle plumes, J. Geophys. Res. 91, 7181-7191, 1986. 9 R.W. Griffiths, M. Gurnis and G. Eitelberg, Holographic measurements of surface topography in laboratory models of mantle hotspots, Geophys. J. 96, 477-495, 1989. 10 R.W. Griffiths and I.H. Campbell, Stirring and structure in mantle plumes, Earth Planet. Sci. Lett. 99, 66-78. 11 M.A. Richards, R.A. Duncan and V.E. Courtillot, Flood basalts and hot-spot tracks: plume heads and tails, Science 246, 103-107, 1989.
12 I.H. Campbell and R.W. Griffiths, Implications of mantle plume structure for the origin of flood basalts, Earth Planet. Sci. Lett. 99, 79-93, 1990. 13 I.H. Campbell, R.W. Griffiths and R.I. Hill, Melting in an Archaean mantle plume: heads it's basalts, tails it's komatiires, Nature 339, 697-699, 1989. 14 I.H. Campbell and R.I. Hill, A two-stage model for the formation of the granite-greenstone terrains of the Kalgoorlie-Norseman area, Western Australia, Earth Planet. Sci. Lett. 90, 11-25, 1988 15 R.I. Hill, I.H. Campbell and R.W. Griffiths, Plume tectonics and the development of stable continental crust, Explor. Geophys. 22, 185-188, 1991. 16 G. Houseman and P. England, A dynamical model of lithosphere extension and sedimentary basin formation, J. Geophys. Res. 91, 719-729, 1986. 17 K.D. Klitgord and H. Schouten, Plate kinematics of the central Atlantic, in: The Geology of North America, Vol. M, The Western North Atlantic Region, P.R. Vogt and B.E. Tucholke, eds., Geol. Soc. Am.,pp. 351-378, 1986. 18 S.P. Srivastava and C.R. Tapscott, Plate kinematics of the North Atlantic, in: The Geology of North America, Vol. M, The Western North Atlantic Region, P.R. Vogt and B.E. Tucholke, eds., Geol. Soc. Am,pp. 379-404, 1986. 19 R. White and D. McKenzie, Magmatism at rift zones: The generation of volcanic continental margins and flood basalts, J. Geophys. Res. 94, 7685-7729, 1989. 20 A.K. Martin and C.J.H. Hartnady, Plate tectonic development of the south-west Indian Ocean: A revised reconstruction of East Antarctica and Africa, J. Geophys. Res. 91, 4767-4786, 1986. 21 J.T. Wilson, Evidence from oceanic islands suggesting movement in the Earth, Phil. Trans. R. Soc. London Ser. A 25, 145 165, 1965. 22 L.R. Sykes, Intraplate seismicity, reactivation of preexisting zones of weakness, alkaline magmatism, and other tectonism postdating continental fragmentation, Rev. Geophys. Space Phys. 16, 621-688, 1978. 23 J.A. Dunbar and D.S. Sawyer, Continental rifting at pre-existing lithospheric weaknesses, Nature 333, 450-452, 1988. 24 J.-G. Schilling, Iceland mantle plume: geochemical study of Reykjanes Ridge, Nature 242, 565-571, 1973. 25 D. McKenzie and M.J. Bickle, The volume and composition of melt generated by extension of the lithosphere, J. Petrol. 29, 625-679, 1988. 26 G.F. Davies, Ocean bathymetry and mantle convection, 1. Large-scale flow and hot-spots, J. Geophys. Res. 93, 10467-10480, 1988. 27 N.H. Sleep, Hotspots and mantle plumes: Some phenomenology, J. Geophys. Res. 95, 6715-6736. 28 R.W. Griffiths, Written communication, 1990. 29 C.J. Ebinger, Tectonic development of the western branch of the East African rift system, Geol. Soc. Am. Bull. 101, 885 903, 1989. 30 F.R. Boyd, A pyroxene geotherm, Geochim. Cosmochim. Acta 37, 2533-2546, 1973. 31 F.R. Boyd, J.J. Gurney and S.H. Richardson, Evidence for a 150-200 km thick Archaean lithosphere from diamond inclusion thermobarometry, Nature 315, 387-389, 1985.
416
32 L.O. Nicolaysen, R.J. Hart and N.H. Gale, The Vredefort radioelement profile extended to supracrustal strata at Carletonville, with implications for continental heat flow, J. Geophys. Res. 86, 10653-10661, 1981. 33 A.A. Finnerty and F.R. Boyd, Thermobarometry for garnet peridotites: basis for the determination of thermal and compositional strcuture of the upper mantle, in: Mantle Xenoliths, P.H. Nixon, ed., pp. 3 8 1 - 4 0 , Wiley, New York, N.Y., 1987. 34 D.A. Carswell and F.G.F. Gibb, Garnet lherzolite xenoliths in the kirnberlites of northern Lesotho: revised P T equilibration conditions and upper mantle palaeogeotherm, Contrib. Mineral. Petrol. 97, 473-487. 35 J. Erlank, ed., Petrogenesis of the volcanic rocks of the Karoo Province, Spec. Publ. Geol. Soc. S. Afr. 13, 1984. 36 W.F. McDonough, Constraints on the composition of the continental lithospheric mantle, Earth Planet. Sci. Left. 101, 1 18, 1990. 37 P.R. May, Pattern of Triassic-Jurassic diabase dykes around the North Atlantic in the context of predrift position of the continents, Geol. Soc. Am. Bull. 82, 1285-1292, 1971. 38 T.E. Smith and H.C. Noltimier, Paleomagnetism of the Newark trend igneous rocks of the North Central Appalachians and the opening of the Central Atlantic Ocean, Am. J. Sci. 279, 778-807, 1979. 39 J.F. Sutter and T.E. Smith, 4°Ar/39Ar ages of diabase intrusions from Newark trend basins in Connecticut and Maryland: initiation of Central Atlantic rifting, Am. J. Sci. 279, 808-831, 1979. 40 J.F. Sutter, Innovative approaches to the dating of igneous events in the early Mesozoic basins of the Eastern United States, U.S. Geol. Surv. Bull. 1776, 194 199, 1988. 41 A.J. Froelich and G.R. Robinson, Jr., eds., Studies of the Early Mesozoic Basins of the Eastern United States, U.S. Geol. Surv. Bull. 1776, 423 pp., 1988. 42 B.C. Burchfiel and G.A. Davis, Triassic and Jurassic tectonic evolution of the Klamath Mountains-Sierra Nevada geologic terrane, in: The Geotectonic Development of California, W.G. Ernst, ed., pp. 50-70, Prentice-Hall, Englewood Cliffs, N.J., 1981. 43 A.G. Smith, A.M. Hurley and J.C. Briden, Phanerozoic Paleocontinental World Maps, 102 pp., Cambridge University Press, Cambridge, 1981. 44 G.F. Davies, Role of the lithosphere in mantle convection, J. Geophys. Res. 93, 10451 10466, 1988. 45 Y. Ricard and C. Vigny, Mantle dynamics with induced plate tectonics, J. Geophys. Res. 94, 17543-17559, 1989.
R,I. H I L L
46 D.V. Kent and F.M. Gradstein, A Jurassic to recent chronology, in: The Geology of North America, Vol. M, The Western North Atlantic Region, P.R. Vogt and B.E. Tucholke, eds., Geol. Soc. Am., 45-50, 1986. 47 W.R. Roest and S.P. Srivastava, Sea-floor spreading in the Labrador Sea: A new reconstruction, Geology 17, 10001003, 1989. 48 N.H. Sleep, Thermal effects of the formation of Atlantic continental margins by continental break-up, Geophys. J. R. Astron. Soc. 24, 325-350, 1971. 49 N.H. Sleep and N.S. Snell, Thermal contraction and flexure of Mid-Continent and Atlantic marginal basins, Geophys. J. R. Astron. Soc 45, 125 154, 1976. 50 J.P. Smoot, A.J. Froelich and G.W. Luttrell, Uniform symbols for the Newark Supergroup, U.S. Geol. Surv. Bull. 1776, 1-5, 1988. 51 R.D. Dallmeyer, The Palisades sill: a Jurassic intrusion? Evidence from 4°Ar/39Ar incremental release ages, Geology 3, 243-245, 1975. 52 G.R. D u n n i n g and J.P Hodych, U / P b zircon and baddeleyite ages for the Palisades and Gettysburg sills of the northeastern United States: Implications for the age of the Triassic/Jurassic boundary, Geology 18, 795 798. 1990. 53 P.E. Olsen, N.H. Shubin and M.H. Anders, New early Jurassic tetrapod assemblages constrain Triassic-Jurassic extinction event, Science 237, 1025-1029, 1987. 54 J.W. Morgan, Hot spot tracks and the early rifting of the Atlantic, Tectonophysics 94, 123-139, 1983. 55 A.J. Froelich and D. Gottfried, An overview of Early Mesozoic intrusive rocks in the Culpepper Basin, Virginia and Maryland, U.S. Geol. Surv. Bull. 1776, 151 165, 1988. 56 W.J. Pegram, Development of continental lithospheric mantle as reflected in the chemistry of the Mesozoic Appalachian Tholeiites, U.S.A., Earth Planet. Sci. Lett. 97, 316331. 57 P. England and G. Houseman, Extension during continental convergence, with application to the Tibetan Plateau, J. Geophys. Res. 94, 17561 17579, 1989. 58 J.P. Platt and R . L M . Vissers, Extensional collapse of thickened continental lithosphere: A working hypothesis for the Alboran Sea and Gibraltar arc, Geology 17, 540-543, 1989. 59 G.E. Vink, 1984, A hotspot model for Iceland and the Voting Plateau, J. Geophys. Res. 89, 9949-9959, 1984. 60 J.H. McBride, K.D. Nelson and L.D. Brown, Evidence and implications of an extensive early Mesozoic rift basin and b a s a h / d i a b a s e sequence beneath the southeast Coastal Plain, Geol. Soc. Am. Bull. 101,512 520, 1989.