Abstract: When basalt magmas are emplaced into continental crust, melting and generation of silicic magma can be expected. The fluid dynamical and heat transfer processes at the roof of a basaltic sill in which the wall rock melts are investigated theoretically and also experimentally using waxes and aqueous solutions. At the roof, the low density melt forms a stable melt layer with negligible mixing with the underlying hot liquid. A quantitative theory for the roof melting case has been developed. When applied to basalt sills in hot crust, the theory predicts that basalt sills of thicknesses from 10 to 1500 m require only 1 to 270 y to solidify and would form voluminous overlying layers of convecting silicic magma. For example, for a 500 m sill with a crustal melting temperature of 850 °C, the thickness of the silicic magma layer generated ranges from 300 to 1000 m for country rock temperatures from 500 to 850 °C. The temperatures of the crustal melt layers at the time that the basalt solidifies are high (900-950 °C) so that the process can produce magmas representing large degrees of partial fusion of the crust. Melting occurs in the solid roof and the adjacent thermal boundary layer, while at the same time there is crystallization in the convecting interior. Thus the magmas formed can be highly porphyritic. Our calculations also indicate that such magmas can contain significant proportions of restite crystals. Much of the refractory components of the crust are dissolved and then re-precipitated to form genuine igneous phenocrysts. Normally zoned plagioclase feldspar phenocrysts with discrete calcic cores are commonly observed in many granitoids and silicic volcanic rocks. Such patterns would be expected in crustal melting, where simultaneous crystallization is an inevitable consequence of the fluid dynamics. The time-scales for melting and crystallization in basalt-induced crustal melting (10—10 y) are very short compared to the lifetimes of large silicic magma systems (>10 y) or to the timescale for thermal relaxation of the continental crust (> 10 y). Several of the features of silicic igneous systems can be explained without requiring large, high-level, long-lived magma chambers. Cycles of mafic to increasingly large volumes of silicic magma with time are commonly observed in many systems. These can be interpreted as progressive heating of the crust until the source region is partially molten and basalt can no longer penetrate. Every input of basalt triggers rapid formation of silicic magma in the source region. This magma will freeze again in time-scales of order 10—10 y unless it ascends to higher levels. Crystallization can occur in the source region during melting, and eruption of porphyritic magmas does not require a shallow magma chamber, although such chambers may develop as magma is intruded into high levels in the crust. For typical compositions of upper crustal rocks, the model predicts that dacitic volcanic rocks and granodiorite/tonalite plutons would be the dominant rock types and that these would ascend-from the source region and form magmas ranging from those with high temperature and low crystal content to those with high crystal content and a significant proportion of restite. I N T R O D U C T I O N One of the central questions in igneous petrology concerns the generation of silicic magmas. There is now convincing evidence that most of the large plutonic complexes of granite in the continental crust are the result of crustal anatexis (Pitcher, 1987). There is also [Journal of Petrologf, Vol. 29, Ptn 3, pp 599-«24, 1988] © Oxford Umvcroty Prcu 19S8 600 HERBERT E. HUPPERT AND R. STEPHEN J. SPARKS widespread evidence that basaltic magma from the mantle is often intimately associated with the generation of silicic magmas (Hildreth, 1981). This association of mafic and silicic magmas can occur in orogenic belts above subduction zones, in continental hot-spots, and in regions of crustal extension. In plutonic complexes, mafic and intermediate igneous activity are recorded in contemporaneous dyke swarms, small satellite intrusions, and in mafic enclaves within the granites (Vernon, 1983; Pitcher, 1986, 1987). In silicic volcanic centres, evidence of basaltic magmatism is found in satellite lava fields and cinder cones, early lava shields and stratovolcano complexes prior to the main silicic volcanism (Lipman, 1984), and as mafic inclusions and bands within the silicic volcanic rocks (Smith, 1979; Bacon, 1986). Petrological and geochemical features of many silicic igneous rocks are also convincingly explained by admixture of a mantle-derived (mafic) component with a crustal melt. Regions of high temperature and low pressure metamorphism are commonly associated with granite plutonism and a plausible explanation of this association is that basalt is intruded into the crust, causing melting and high heat flow. Indeed basalt underplating of the crust is a currently popular idea to explain both large scale crustal melting and the strongly layered character of the lower crust. While there may be some silicic magmas that are generated by processes without the aid of basaltic input, such as tectonic thickening of radioactive crust (England & Thompson, 1984; Pitcher, 1987), this paper takes the position that in many cases the additional thermal energy of basalt is essential. The continental crust is strongly layered in terms of its composition, density, and mechanical behaviour. The upper crust is cold and brittle whereas the lower crust is hotter, has a higher density, deforms in a ductile manner, and is commonly characterized by prominent horizontal layering. Basalt magma can be emplaced into the continental crust as dykes and sills and, in some cases where the rate of magma input is high, these intrusions can coalesce to form larger magma chambers. Dyke emplacement does not seem an efficient way of generating large volumes of silicic magma, because dykes are usually small in width and much of the potential heat for melting will not be utilized if the mafic magma erupts. Sills provide a more promising situation in which extensive crustal melting can occur. Horizontal intrusions concentrate their heat at a particular level in the crust and do not dissipate their heat over a large depth range. Sills are intrinsically more efficient than dykes in this respect. Dykes may play an important role in heating up the crust to initiate melting. However, once a region of the crust has become hot, ductile, and partially molten, conditions for dyke propagation become less favourable. A layer or region of partially molten crust provides an effective density barrier and we suggest that basalt magma reaching such a level will spread out as horizontal intrusions. An additional factor which promotes sill formation in the lower parts of the crust is its strongly layered character providing a structural environment in which horizontal intrusions are favoured. For these reasons this paper is concerned principally with the heat transfer and fluid dynamics of sills intruded into hot continental crust. We consider the cooling and crystallization of basaltic sills emplaced into the continental crust. In particular, we emphasize the situation where the roof of the sill is composed of rock which has a fusion temperature that is lower than the magma temperature and the roof rock consequently melts. This is likely to be the normal situation where basalt intrudes into the typical rock types of mature and ancient middle and upper crust which are already at high temperature. However, the concepts developed in this paper are also likely to be applicable to conditions in immature continental crust such as in island arcs, to more refractory lower crust and to lower crust formed by slightly older or even contemporaneous episodes of basalt underplating. In each of these latter cases, lithologies which have relatively low fusion temperatures can form by differentiation processes and can be remelted by further intrusion THE GENERATION OF GRANITIC MAGMAS 601 of basalt. Thus the model is not confined to the origin of granites, but should be relevant to the origins of intermediate rocks such as tonalites and evolved alkaline rocks such as syenite. We present experimental studies on the melting of the roof of a sill. We develop a quantitative model of the melting process at the roof, which describes the rates at which a new layer of roof melt forms and the rates at which the underlying liquid layer solidifies. We discuss possible mechanisms by which the melts can be mixed together and also their implications for magma genesis within the continental crust. A companion paper (Huppert & Sparks, 1988a) describes the melting of the roof of a chamber from a detailed fluid mechanical point of view. Throughout this paper the magma will be considered to be Newtonian. Although magma in reality can be non-Newtonian, especially when it is rich in crystals (McBirney & Murase, 1984) its nonlinear Theological properties and the consequences of its non-Newtonian rheology are poorly understood. Two effects may be evident: there may exist a yield strength, so that for a sufficiently low applied stress the magma will not move; and the nonlinear viscosity may alter the heat flux transferred by a convecting magma. Because of the relatively large values of the Rayleigh number that result in most of our calculations, we anticipate that the yield strength will be exceeded by quite a margin. The alterations in the heat flux are at the moment difficult to anticipate and we suggest that the reader views our quantitative results as an indication of the calculated quantity rather than as a precise value. It may be possible to examine non-Newtonian effects with greater insight in the future, but a Newtonian description illuminates many of the fundamental effects and is a necessary first step in order to form the basis for any comparison. EXPERIMENTAL STUDIES The geological problem i