JOURNAL OF PETROLOGY VOLUME 40 NUMBER 2 PAGES 241–254 1999 A Two-stage Thermal Evolution Model of Magmas in Continental Crust TAKEHIRO KOYAGUCHI1∗ AND KATSUYA KANEKO2 1 EARTHQUAKE RESEARCH INSTITUTE, UNIVERSITY OF TOKYO, 113-0032 TOKYO, JAPAN 2 CENTRAL RESEARCH INSTITUTE OF ELECTRIC POWER INDUSTRY, 1646 ABIKO, ABIKO-SHI, 270-1194 CHIBA, JAPAN RECEIVED JULY 1, 1997; REVISED TYPESCRIPT ACCEPTED JUNE 26, 1998 When a basaltic magma is emplaced in a continental crust, a silicic magma is generated by melting of the crust. The light silicic magma forms a separate magma layer with little chemical interaction with the underlying dense basaltic magma layer. Extensive melting occurs at the boundary between the silicic magma and the crust while the basalt acts a heat source. The mass and heat transfer at the boundary between the silicic magma and the crust controls the thermal evolution of the silicic magma. The thermal evolution of the silicic magma after the basalt emplacement is divided into two stages. In the first stage, the temperature in the silicic magma rises above and then decays back to the melting temperature of the crust on a short timescale (102 years). The results of fluid dynamics experiments suggest that the silicic magma generally has a lower melting temperature than the crust because of fractional crystallization and mixing of partial melts during the first stage, and that it can be effectively liquid at the end of the first stage. In the second stage, the silicic magma cools slowly by heat conduction on a much longer timescale (105 years). Petrological features of the magma in the second stage are strongly constrained by petrological features of the surrounding crust as well as those of the supplied magma itself; its temperature remains at or just below the melting temperature of the crust for a long time because of the slow cooling rate; its phenocryst content reflects the difference in the melt fraction vs temperature relationships between the magma and the crust. Judging from the distinct cooling rate between the two stages, erupted magmas are statistically more likely to reflect the characteristics of magmas in the second stage. INTRODUCTION chamber; continental crust; crustal melting; magma residence times; fluid dynamics experiments When magma is emplaced into continental crust, it is believed to be stored in a ‘magma chamber’ for a certain period. Extensive studies have revealed how petrological and geochemical features of magmas are controlled by processes in magma chambers such as crystallization of magma, melting of crustal materials, and mixing of magmas (e.g. McBirney, 1980; Huppert & Turner, 1981; Sparks et al., 1984; Koyaguchi & Blake, 1991). This paper discusses how petrologic processes such as assimilation and fractional crystallization affect the thermal evolution and the residence times of magmas in magma chambers in continental crust. Some exploratory models for the residence time of a magma chamber in continental crust have been proposed on the basis of the physics of heat and mass transfer (Spera, 1979; Huppert & Sparks, 1988a; Marsh, 1989). To formulate the present problem it is helpful to introduce the model by Huppert & Sparks (1988a), in which the physics of crystallization and melting at the chamber roof as well as the effects of thermal convection in the magma are taken into account. According to their model, a basalt emplaced into continental crust results in extensive melting of the crustal materials at the chamber roof to form a silicic magma. The temperature in the silicic magma rises above and then decays back to the fusion temperature of the crust (in general ‘the effective fusion temperature’, the definition of which will be given in a later section) on a short timescale of 101–103 years after the basalt emplacement. Huppert & Sparks (1988a) also pointed out that the silicic magma generated by melting of the crust must be solidified when its temperature reaches the fusion temperature of the crust, because the silicic magma should have the same fusion temperature ∗Corresponding author. Telephone: 81-3-3812-2111, ext. 5755. Fax: 81-3-5802-3391. e-mail: [email protected] Oxford University Press 1999 KEY WORDS: magma JOURNAL OF PETROLOGY VOLUME 40 as the crust. This model, which implies that the residence times of silicic magmas in the continental crust will be short, is apparently in conflict with the model of long-lived magma chambers with cooling times of over 105 years (e.g. Spera, 1979), even though both models similarly took into account the effects of heat transfer because of magma convection and latent heat. These different predictions of the residence times have led to divergent views of the correct interpretation of geochemical and geological data from a silicic magma system (e.g. Halliday et al., 1989; Halliday, 1990; Mahood, 1990; Sparks et al., 1990). The purpose of this paper is to propose another view, which, we believe, reconciles the two hypotheses. We show that the melting and crystallization processes are rapid as long as the temperature of the silicic magma is higher than the fusion temperature of the crust, and that this is the key hypothesis of the Huppert & Sparks model. To examine the conditions under which long-lived magmas develop, we proceed by the following three steps. First, we review some theories of mass and heat transfer at crust–magma boundaries and show that the above paradox about the magma residence times results from differences in the boundary conditions in their numerical models; a magma can be long lived if it has a lower temperature than the fusion temperature of the surrounding crustal rocks. Second, we show the results of fluid dynamics experiments and numerical calculations to investigate possible mechanisms that might cause a magma to have a lower fusion temperature than the surrounding crustal rocks. These analyses imply that the thermal evolution of the crustal magma is generally divided into two distinct stages: a first stage of rapid cooling and a second stage of slow cooling. Third, petrological features of the long-lived magmas in the second stage will be discussed from the viewpoint of the phase equilibria of crustal materials. HEAT AND MASS TRANSFER BETWEEN MAGMA AND CRUST The factors that govern timescales of melting and solidification of magmas are considered here on the basis of simple theories of one-dimensional heat and mass transfer. Melting and solidification processes of multicomponent materials such as crustal rocks and magmas are complicated because they are composed of both solid and liquid phases between the solidus and liquidus temperatures. We start by describing some basic features of mass and heat transfer in simple systems where magma or crust has no melting interval but their melting temperatures are allowed to be different. We will discuss the effect of the melting interval later. The conditions at the boundary in the simple systems can be divided into the following four cases according NUMBER 2 FEBRUARY 1999 to whether there are phase changes at the solid–liquid boundary and the occurrence of convective heat flux from the liquid to the interface: Case 1: moving boundary because of melting or crystallization with convective heat flux from the liquid region to the interface; Case 2: moving boundary with conductive heat flux from the liquid region to the interface; Case 3: stationary boundary with convective heat flux from the liquid region to the interface; Case 4: stationary boundary with conductive heat flux from the liquid region to the interface. Case 4 is the simplest problem of heat conduction, and no further explanation is necessary here. The problems of heat conduction including boundary migration caused by phase changes (Case 2) are known as ‘Stefan Problems’, solutions of which are widely applied to geological situations (e.g. Turcotte & Schubert, 1982). In this case, the rate of boundary migration has to be calculated from an energy balance involving the heat flux on both sides of the boundary and the latent heat. The velocity of the boundary towards the solid side, ȧ, is given by ȧ= Fl−Fs qL (1) where Fl is the heat flux from the liquid to the interface, Fs is the heat flux from the interface to the solid, q is the density, and L is the latent heat. The rate of melting or solidification is limited by the difference in the conductive heat fluxes between the liquid and the solid media [Fl – Fs in equation (1)] in this case. In Case 1, the rate of melting or solidification at the boundary is also determined by equation (1); however, because the rate of heat release from the liquid to the interface, Fl, is determined by the convective heat flux in the liquid, ȧ can be much greater than in the case of pure conduction. This situation represents, for example, mass and heat transfer at the roof of a magma chamber where the magma temperature is higher than the melting temperature of the wall rock (Huppert & Sparks, 1988a). When the liquid undergoes high Rayleigh number thermal convection, the heat flux is expressed (Turner, 1973) as Fl=kkl ag jlm 1/3 (Tl−Ti)4/3 (2) where Tl and Ti are the temperatures of the liquid and interface, respectively, a, kl, jl and m are the coefficients of thermal expansion, heat conduction, thermal diffusivity and the kinematic viscosity of the liquid, respectively, g 242 KOYAGUCHI AND KANEKO THERMAL EVOLUTION OF MAGMAS is the gravitational acceleration, and k is a constant. In a quasi-steady state, the rate of melting can be solved analytically (Huppert & Sparks, 1988b) to yield ȧ= Fl q[c(Ti−Tx)+L] (3) where c is the specific heat and Tx is the temperature of the solid far from the interface. Because the temperature at the interface is fixed at the melting temperature of the wall rocks in Case 1, rapid melting occurs at the boundary at a rate limited by the convective heat flux from the liquid region given by equation (2), until the liquid temperature reaches the melting temperature of the wall rocks [i.e. Tl – Ti ~ 0 in equation (2)]. On the other hand, after the liquid temperature reaches the melting temperature, the system will evolve in the manner of Case 2 or Case 3, which will be introduced next. In Case 3 no melting is allowed at the boundary, and the rate of heat release from the liquid layer to the solid is determined by the convective heat flux [Fl in equation (2)] as in Case 1. This situation represents, for example, the case where the melting temperature of the solid is higher than the temperature of the liquid. The most important difference from Case 1 is that the temperature at the interface, Ti, is not fixed to the melting temperature in Case 3. If the convective heat flux is initially greater than the conductive heat flux in the solid, the temperature at the interface will increase until the convective heat flux in the liquid becomes equal to the conductive heat flux in the solid [see the Appendix of Huppert & Sparks (1988b) and Marsh (1989)]. Therefore, the heat flux from the liquid to the solid is eventually limited by the conductive heat flux in the solid. Figure 1 shows the results of calculations of the cooling time of a 1 km thick liquid layer for the above three cases on the basis of the previously published formulae [for detailed derivations, see Huppert & Sparks (1988a, 1988b) for Case 1, Turcotte & Schubert (1982) for Case 2, and the Appendix of Huppert & Sparks (1988b) for Case 3]. Two distinct modes of cooling are recognized; the timescale is as short as 102 years in Case 1, whereas it is 104–105 years in the other cases. It is particularly important that, even though convection occurs, the timescale of cooling in Case 3 is much closer to the purely conductive cases (e.g. Case 2) and much longer than in the case where convection and melting simultaneously occur (Case 1). The two different hypotheses on the residence times of crustal magma chambers (Spera, 1979; Huppert & Sparks, 1988a) are therefore interpreted to have resulted from different views on the appropriate boundary conditions (i.e. Case 3 and Case 1, respectively). The difference in timescales between Case 1 and the other cases is explained by the difference in limiting 243 Fig. 1. Variation of liquid temperature with time after a contact of a hot-liquid–cold-solid pair under various boundary conditions (Cases 1, 2, and 3; see text for their definitions) on the basis of one-dimensional mass and heat transfer models. The average temperature in the liquid is given for Case 2. The initial thickness and temperature of the liquid layer are 1000 m and 1200°C, respectively. The circles, squares and diamonds represent the results for initial solid temperatures of 150, 500, and 675°C, respectively. The melting temperature is assumed to be 850°C for the calculations for Cases 1 and 2. Values of other variables used are listed in Table 1. processes. The rate of decrease in liquid temperature is limited by heat loss because of wall rock conduction in Cases 2, 3 and 4, whereas it is determined by the increasing proportion of initially cold melts generated by melting which convectively mix with the hot liquid region in Case 1. It is reasonable, therefore, to define two modes of mass and heat transfer in terms of the limiting processes; hereafter referred to as ‘convection and melting mode (C&M mode)’ and ‘conductive cooling mode (CC mode)’. Whether a system evolves in C&M mode depends more on whether the magma temperature is higher than the melting temperatures of the wall rocks than on other detailed physics. To apply the above argument to the natural multicomponent system, the effects of melting interval should Table 1: Values used for calculation in Fig. 1 Quantity Value used Unit q 2·6 × 103 (s), 2·4 × 103 (l) kg/m3 k 2·4 (s and l) W/m per K c 1·3 × 103 (s and l) J/kg per K m 1·0 × 102 m2/s L 2·9 × 105 J/kg a 5·0 × 10–5 per K g 9·8 m/s2 k 0·1 The letters ‘s’ and ‘l’ represent solid and liquid, respectively. JOURNAL OF PETROLOGY VOLUME 40 be taken into account. In the context of mass and heat transfer, we must specify the temperature at which a material behaves as a liquid in a fluid mechanical sense (in this paper we use the term ‘liquid’ in the fluid mechanical sense and the term ‘melt’ as a phase equilibria term). The concept of ‘effective fusion’ (e.g. Huppert & Sparks, 1988a) is useful in describing the physical properties of such solid–liquid mixtures. When the temperature of a crustal material is only slightly higher than its solidus temperature and the melt fraction is so low that the crystalline phases form an interconnected framework, the crust can be regarded as partially molten solid. At temperatures much higher than the solidus temperature, the melt fraction is greater, and beyond a critical narrow range of melt fractions the connectedness of the crystalline framework is destroyed and the rock behaves as liquid suspending crystals (i.e. magma). Experimental and theoretical studies indicate that a drastic transition in the mechanical properties of a solid–liquid mixture occurs at a narrow range of melt fraction between 30 and 50% (e.g. Marsh, 1981; Rutter & Neumann, 1995). We tentatively take a critical value of 50% melt fraction in this study (Fig. 2). Because the melt fraction of a crustal material with a given chemical composition at a fixed pressure is a function of temperature alone, we can define ‘effective fusion temperature (EFT)’ as a function of chemical composition. Mode of heat and mass transfer greatly changes at the EFT. For example, in the case of melting and solidification at the roof of a magma chamber, the heat and mass can be efficiently conveyed by vigorous convection above the EFT, whereas they migrate only by conduction and diffusion below it (Huppert & Sparks, 1988a). The relationships between melt fraction and temperature for the crustal materials must be specified to determine their EFTs (Fig. 2). Huppert & Sparks (1988a) assumed a single monotonic relationship between melt fraction and temperature for each melting and crystallization event in most of their calculations (Fig. 2a). In fact, this was one of the most critical assumptions which led to the hypothesis of short residence times of silicic magmas. A silicic magma generated by melting of a crust crystallizes very quickly in C&M mode until its temperature returns to the EFT of the crust. Because of the above assumption, melt fraction and temperature of the silicic magma change along a single melt fraction vs temperature relationship, and so when the temperature returns to the EFT the silicic magma is inevitably effectively solidified (Fig. 2a). However, the evolution of the melt fraction vs temperature relationship during a melting and crystallization event in the natural continental crust is generally more complex. For example, if a crustal material has a eutectic composition, then its melt fraction vs temperature relationship is discontinuous and a magma of up to 100% liquid can exist at its ‘EFT’ (Fig. 2b). NUMBER 2 FEBRUARY 1999 Furthermore, the EFT of magma may become lower than that of its surrounding crust because of shifts of melt fraction vs temperature relationships as a result of compositional changes during melting and crystallization processes in C&M mode (Fig. 2c). In such cases magmas can remain as liquids at or below the EFT of the surrounding crust. The analyses in this section indicate that, if a magma can remain as a liquid at the EFT of the surrounding crust, subsequent cooling is a very slow process (e.g. Case 2 or Case 3); in other words, the magma is long lived. Consequently, the evolution of the magma and the continental crust after each input of a high-temperature magma is conceptually divided into two stages. In the first stage, a silicic magma is rapidly generated by melting and as long as its temperature is above the EFT of the crust it crystallizes in C&M mode. In the second stage, the silicic magma is cooled slowly in CC mode. Whether or not CC stage actually exists depends on whether the magma has a melt fraction vs temperature relationship such that the magma has a lower EFT than the surrounding crust (Fig. 2c). We outline below some examples of petrological processes that result in the shifts of melt fraction vs temperature relationships as in Fig. 2c. We will also point out in a later section that the situation that is qualitatively equivalent to the discontinuous melt fraction vs temperature relationships as in Fig. 2b can occur in a wide compositional range of hydrous crusts. POSSIBLE MECHANISMS TO GENERATE LONG-LIVED MAGMAS After basalt emplacement, the light silicic magma generated by crustal melting forms a magma layer overlying the dense basaltic magma layer. Extensive melting occurs at the boundary between the silicic magma and the crust while the basalt acts a heat source (Campbell & Turner, 1987; Huppert & Sparks, 1988a; see Fig. 3a). We focus on the process at the boundary between the silicic magma and the crust. Selective assimilation (fluid dynamics experiments) Huppert & Sparks (1988a) assumed that the silicic magma and the crust have the same chemical compositions, and hence the same melt fraction vs temperature relationship, throughout the melting process on the basis of the experimental results of roof melting. When melting occurs at the roof of a magma chamber, the silicic magma generated forms a separate layer with little chemical interaction with the underlying basaltic magma because of their large density contrast. Furthermore, although 244 KOYAGUCHI AND KANEKO THERMAL EVOLUTION OF MAGMAS Fig. 2. Schematic diagram showing the definition of the effective fusion temperature (a) and possible melt fraction vs temperature relationships in continental crust (a, b and c). The relationship may be a single monotonic function (a), a discontinuous one for eutectic compositions (b), or it may shift because of a certain fractionation process during the C&M stage (c). the crust is partially melted, evolved light interstitial melt cannot segregate from the residual crystals at the roof into the main silicic magma. The crustal material will mix with the main silicic magma as a whole when it becomes effectively liquid. In this case, the assumption of the same composition between the crust and the silicic magma may be justified unless the effect of crystal settling in the magma is significant. On the other hand, when melting occurs at the floor or the side-walls of a magma chamber, evolved melts formed by partial melting may segregate from the boundary and selectively mix with the main silicic magma body (e.g. Woods, 1991; Kerr, 1994). The compositional fractionation because of the floor or side-wall melting modifies the melt fraction vs temperature relationship of the magma and affects subsequent mass and heat transfer between the silicic magma and the surrounding crust. We present some results of fluid dynamics experiments to show how a compositionally uniform solid–liquid pair fractionates by floor and/or side-wall melting. To simulate the melting behaviour at the floor and side-wall of a magma chamber, experiments were conducted using aqueous solutions and solids in the NH4Cl– H2O binary system (Fig. 3). The NH4Cl–H2O binary system is a simple eutectic system and the eutectic temperature and composition are –15·4°C and 19·7 wt % NH4Cl, respectively (Fig. 3c). A solid–liquid pair of a uniform chemical composition of 28 wt % NH4Cl is used as starting material. Under this condition of 28 wt % NH4Cl, the density of the released liquid decreases along the liquidus surface towards the eutectic composition and the density of the liquid generated at the solidus temperature is the minimum in the liquid layer (Fig. 3c). This density and compositional relationship is qualitatively identical to the typical crust and silicic magma pair with a uniform initial composition. The apparatus used consists of a Perspex rectangular tank (10 cm × 10 cm × 30 cm for the floor melting experiments and 10 cm × 20 cm × 15 cm for the sidewall experiments) and thick foam plastic plates covering 245 JOURNAL OF PETROLOGY VOLUME 40 NUMBER 2 FEBRUARY 1999 Fig. 3. A schematic sketch of floor melting and side-wall melting of a magma chamber in a continental crust (a), and the design of the analogue experiments in the NH4Cl–H2O binary system (b, c). The contours in the phase diagram (c) show density of the liquid (g/cm3). the tank for thermal insulation (Fig. 3b). The configurations of the experiments are designed to extract the essence of the physical processes at the floor and sidewall and do not necessarily represent the actual configuration of a magma chamber itself. The experiments were started with a solid layer (typically 18 cm deep for the floor melting experiments and 10 cm thick for the side-wall experiments) at temperatures slightly below the solidus temperature (about –16°C). A dyed hot aqueous solution of identical bulk composition at liquidus temperature (24°C) was rapidly poured and the temperatures inside the tank were monitored at 20–26 fixed positions and the compositions of liquid were measured at five heights during the experiments. Other experiments with various solid compositions (19·7 wt % and 73 wt % NH4Cl) and temperatures (down to –45°C) were also carried out for comparison and systematic quantitative analyses, some results of which were reported elsewhere (Koyaguchi & Kaneko, 1995; Kaneko, 1996; Kaneko & Koyaguchi, 1996). Figure 4 shows the results of the floor melting experiments. Immediately after the start of each run, a mushy layer developed between the liquid and solid layers by crystallization of NH4Cl, and the boundary between the solid and mushy layers moved downwards by partial melting (Fig. 4a and b). At the boundary 246 KOYAGUCHI AND KANEKO THERMAL EVOLUTION OF MAGMAS Fig. 4. (a). between the solid and mushy layers, the solid layer partially melted to form eutectic liquid leaving 10% residual NH4Cl solid at the eutectic temperature. Many plumes rose from the mushy layer and resulted in vigorous convective motions in the liquid region. Temperature and composition of the liquid region were almost uniform because of the vigorous convection in the earliest stage and subsequently weak thermal and compositional gradients developed; the temperature rose and the NH4Cl concentration decreased upwards (Fig. 4c). The average NH4Cl content in the liquid region decreased with time because of crystallization of NH4Cl crystals in the mushy region and simultaneous mixing with partial melts generated at the solid–mush interface. On the other hand, the bulk NH4Cl content of the partially molten zone estimated from the mass balance between the liquid region and the mushy layer increased with time (Fig. 4c). The dyed liquid penetrated at a much higher rate than is expected by diffusion alone to the level below the original solid–liquid boundary, suggesting that there was convective exchange between interstitial liquid of the partially molten solid layer and the overlying liquid. Figure 5 shows the results of the side-wall melting experiments with a liquid–solid pair of uniform composition (28 wt % NH4Cl). A vertical mushy layer developed between the liquid and solid layers by melting and crystallization. It collapsed and slumped down from time to time and a pile of NH4Cl crystals formed at the bottom of the liquid region (Fig. 5). The interstitial melt in the mushy layer mixed with the main liquid body while the ‘mush avalanche’ was falling down. Many plumes rose from the crystal pile at the bottom, which resulted in vigorous convective motions in the liquid region. Double diffusive convective layers (e.g. Turner, 1973) with stepwise thermal and compositional gradients developed in the liquid region (Fig. 5). The NH4Cl concentration in the uppermost convective layer decreased with time and it became as low as 21 wt % NH4Cl. The initially dyed liquid layer penetrated into the mushy layer also in the side-wall experiments, suggesting the presence of liquid exchange between the partially molten solid layer and the liquid region. The solution in the liquid region was dyed from the top to the bottom of the tank throughout the experiment, suggesting that partial melts mixed with the dyed liquid while they were incorporated into the liquid region. The experimental conditions and the natural system are different in some respects, such as dimensions, viscosities of liquid, kinetics of crystallization and phase equilibria of constituent materials. For example, the efficient fractional crystallization and the melt segregation from residual crystals seen in the experiments may be largely due to the low viscosity of the solution. In spite of these differences, there are at least two important implications of the above experimental results. First, there was considerable mixing of partial melts from the floor and side-wall boundaries with the main liquid region. Second, regardless of details of the processes, both mixing of the partial melts at the floor and/or side-walls and fractional crystallization increase the low melting point component in the liquid region. This means that the melt fraction vs temperature relationship of the magma shifts in the direction which decreases its EFT compared with the original solid wall rock (see Fig. 2c). Fractional crystallization (numerical analyses) The effect of crystal settling plays a role regardless of the presence or absence of the floor and/or side-wall melting, because crystallization occurs inside the magma chamber, even if melting occurs at the roof of the chamber alone (see Huppert & Sparks, 1988a). As crystals settle, they form a cumulate pile at the bottom of the chamber, leaving a magma of higher melt fraction. This effect is evaluated by a simple model in a binary system (Appendix). To separate the effect of fractional crystallization from other complications such as mixing with partial melts and eutectic melting, crystallization of a given mass 247 JOURNAL OF PETROLOGY VOLUME 40 NUMBER 2 FEBRUARY 1999 Fig. 4. A photograph (a), a schematic diagram (b), and the evolution of gradient in NH4Cl content (c) of the floor melting experiments. The NH4Cl contents in the mushy and partially molten layers are based on the mass balance assuming a linear concentration gradient in these layers. of an initially completely liquid magma along a linear liquidus surface at a constant cooling rate is considered here. Crystals are assumed to be homogeneously distributed by vigorous convection in the magma layer because of high Rayleigh number, but to be separated from the magma layer to the cumulate pile at the bottom at the rate of its settling velocity (see Martin & Nokes, 1988). It is also assumed that the cumulate pile has the assumed critical melt fraction bounding effective liquid and solid (i.e. 50 vol. %) and that there is no chemical interaction between the interstitial melt and the main magma body after the melt is trapped in the cumulate pile. Bulk composition, melt fraction, and depth of the magma layer evolve with time, as the crystals and the less evolved interstitial melts are incorporated into the cumulate pile. In the present context, the melt fraction and the depth of the magma layer at the moment when the temperature reaches the EFT of the crust are of primary interest. These quantities depend on the ratio of the settling timescale to that for the temperature to reach the EFT of the crust (Fig. 6; see Appendix for derivations). If the timescale of settling is much the smaller of the two (i.e. the case of efficient crystal settling), a crystal-poor magma of approximately 1/4 of the initial depth would remain at the EFT of the crust. On the other hand, if the temperature reaches the EFT of the crust before efficient settling, a large amount of highly porphyritic magma results. Crystals that are suspended at this moment settle further while the temperature is kept at or just below the EFT of the crust on a long timescale. The depth of final, crystal-free magma layer 248 KOYAGUCHI AND KANEKO THERMAL EVOLUTION OF MAGMAS Fig. 7. Timescale of C&M stage [i.e. the timescale for (T – Te)/(T0 – Te) to reach 0·05] and timescale of crystal settling (H/vs in the Appendix) for magma body with 1 km thickness as a function of viscosity of magma. Spherical crystals of 10–2 m in diameter are assumed in the calculations of settling velocities. The other values used in the calculations are the same as in Fig. 1 (see Table 1). Fig. 5. A photograph of the side-wall melting experiments. inversely proportional to the rate of incorporation of the cold surrounding crust, which is proportional to the boundary velocity expressed by equation (3). The timescale of settling, on the other hand, is proportional to viscosity, because of the form of Stokes’ law. The ratio of the two timescales depends on the dimension of the magma body only very weakly and hence we express it in the form r>Am2/3 Fig. 6. Melt fraction and depth of magma layer at the moment when the temperature reaches the EFT of the crust, and depth of final, crystal-free magma layer after complete crystal settling as a function of ratios of the timescale of crystal settling to that for temperature to reach the EFT of the crust [i.e. H/(vssF) in the Appendix]. The depth of the magma layer is normalized such that initial thickness is unity. after complete crystal settling decreases with the increasing ratios of timescale of settling to that for temperature to reach the EFT of the crust (Fig. 6). Although the simple situation considered in the Appendix is not identical to the situation where melting occurs at the same time, it would be reasonable to consider that the above timescale for the temperature to reach the EFT of the crust roughly represents the timescale of the C&M stage. The results of numerical calculations indicate that both the timescale of the C&M stage and that of crystal settling are approximately proportional to initial thickness of the magma layer, and that they increase as magma viscosity increases (Fig. 7). The timescale of the C&M stage is approximately proportional to the cube root of viscosity, because it is (4) where A is a constant. The ratio, r, is unity when the viscosity is of order 104–105 Pa s (Fig. 7) for crystals of 10–2 m diameter under the conditions given in Table 1. These results suggest that low-viscosity magma forms a high-melt fraction magma, and high-viscosity magma tends to form a highly porphyritic magma at the end of the C&M stage. The above experimental results and the numerical analyses suggest that a silicic magma generated by melting of crust inevitably has a lower EFT than the surrounding crust at the end of the C&M stage as a result of selective assimilation and/or fractional crystallization, and that the two-stage thermal evolution is a natural consequence after each input of a high-temperature magma in continental crust. PETROLOGICAL IMPLICATIONS There are obviously many factors that govern the evolution of magmas in continental crust, and no single model can explain all the petrological features of observed magmas. Nevertheless, the fact that the cooling rate of 249 JOURNAL OF PETROLOGY VOLUME 40 magma is two or three orders of magnitude greater in the C&M stage than in the CC stage may statistically reflect petrological features of erupted magmas. Marsh (1981) has pointed out that the probability of a magma being erupted can be closely related to the cooling rate of the magma; for example, assuming that eruptions sample magmas randomly in time, the chance of observing a magma within a temperature interval of rapid cooling is smaller than that within a temperature interval of slow cooling. This means that the chance of observing magmas in the CC stage (we call them ‘long-lived magmas’ hereafter) may be greater than that of observing magmas in the C&M stage. To test the present hypothesis of the two-stage thermal evolution model from the viewpoint of petrology, petrological features which characterize long-lived magmas must be specified. The characteristics of long-lived magmas are determined by two factors: (1) temperature is fixed at or just below the EFT of the surrounding crust for a long time because of the slow cooling rate during the CC stage; (2) phenocryst content depends on the degree of the shift of the melt fraction vs temperature relationships during the C&M stage (see Fig. 2c). The shift of melt fraction vs temperature relationships must be quantified from the viewpoint of the phase equilibria of crustal materials, although the fact that the magma has a lower EFT than that of the surrounding crust is robust regardless of the detail of the phase equilibria. According to recent experimental studies (e.g. Wolf & Wyllie, 1995), low-viscosity hydrous silicic melts generated by dehydration melting of hydrous minerals can be efficiently segregated from the host, and these would be the most probable candidate of the selective assimilant. We have analysed the variation of the melt fraction vs temperature relationships caused by mixing of hydrous silicic melt with natural crustal materials using the thermodynamic model ‘MELTS’ of Ghiorso & Sack (1995). Melt fraction increases markedly around 700°C with increasing mixing ratio of the hydrous silicic melt (Fig. 8). The shifts of the melt fraction vs temperature relationships result in decrease in phenocryst content in magmas at the EFT of the crust (Fig. 9). The rate of decrease in phenocryst content at the same mixing ratio as the hydrous silicic melt depends on the bulk composition of the crust involved; it is greater for gabbroic compositions compared with tonalitic or granodioritic compositions (Fig. 9). These features of the shifts of the melt fraction vs temperature relationships and their dependence on the bulk compositions are not artefacts of a particular thermodynamic model, but can be understood as common features of multicomponent eutectic systems. The amounts of melt around 700°C in Fig. 8 are basically determined by the eutectic melting of quartz, orthoclase and plagioclase for silicic compositions (e.g. Piwinskii & 250 NUMBER 2 FEBRUARY 1999 Fig. 8. Variation of melt fraction vs temperature relationships for a granodioritic crust caused by mixing of a hydrous granitic melt (6 wt % water) for mixing ratios from 10 to 50% on the basis of thermodynamics model ‘MELTS’ of Ghiorso & Sack (1995). The pressure is assumed to be 200 MPa. The chemical compositions of the experimental starting materials (Sample 766 with SiO2 70%, and Sample 705 with SiO2 77%) of Piwinskii & Wyllie (1968, 1970) are used for the calculations, to cross-check the thermodynamics model and the experimental results. Although there are slight discrepancies between the two approaches, partly because of the lack of a thermodynamic model of hornblende and its dehydration melting in MELTS code (Ghiorso & Sack, 1995) and partly because of large errors in modal analyses of the experimental products, the results agree fairly well especially in the range of melt fraction >50%. Initial water content of the crustal materials is assumed to be 0·5 wt %. Fig. 9. Variation in phenocryst content of long-lived magmas at the EFT of the crusts as a function of mass fraction of added hydrous silicic melts on the basis of the thermodynamics model. The chemical compositions of the experimental starting materials are used; these are gabbro (SiO2 51%, High-Al basalt 82-66) of Sisson & Grove (1993), tonalite (SiO2 62%, Sample 1213) of Piwinskii & Wyllie (1968), granodiorite (SiO2 70%, Sample 766) of Piwinskii & Wyllie (1968), and granite (SiO2 77%, Sample 705) of Piwinskii & Wyllie (1970). Initial water content of the crustal materials is assumed to be 0·5 wt %. The depth of final, crystal-free magma layer normalized by the initial depth is also shown on the vertical axis at the right-hand side. Wyllie, 1968). Because the hydrous silicic melt has a composition near the ternary minimum of the natural KOYAGUCHI AND KANEKO THERMAL EVOLUTION OF MAGMAS mode at the eutectic temperature. During the CC stage phenocrysts settle leaving a crystal-poor granitic magma with the eutectic composition in the upper part of the magma chamber. Judging from the effects of the shift and the form of melt fraction vs temperature relationships as well as effect of crystallization during the C&M stage, possible petrological features of long-lived magmas can be summarized as follows: (1) Magmas with viscosities of less than 104–105 Pa s (e.g. mafic magmas in a gabbroic crust or hydrous magmas) can be crystal-poor long-lived magmas because of crystal settling during the C&M stage. (2) More than 10% of hydrous silicic melts or 1% water is required as a selective assimilant to form a crystal-poor long-lived magma with <25% crystals. The rate of decrease in crystal content is less remarkable in a crust of intermediate composition than in a gabbroic crust. (3) A crystal-poor long-lived magma at the eutectic temperature forms by eutectic melting in hydrous granodioritic to granitic crusts with water content more than a few weight per cent. These results suggest a general tendency of the petrological features of long-lived magmas: they can be crystal poor when they are mafic or granitic compositions, but they tend to be porphyritic when they have intermediate compositions. The observation that andesites tend to be more porphyritic than basalts or rhyolites (Ewart, 1982) may support the two-stage thermal evolution model and the ubiquity of long-lived magmas as erupted magmas. Fig. 10. Variation of melt fraction vs temperature relationships for a granodioritic crust because of addition of water on the basis of the thermodynamics model. The compositions used in the calculations are same as in Fig. 8. Fig. 11. Variation in phenocryst content of long-lived magmas at the EFT of the crust as a function of added water content. Initial water content of the crustal materials is assumed to be 0·5 wt %. The compositions used in the calculations are same as in Fig. 9. The depth of final, crystal-free magma layer normalized by the initial depth is also shown on the vertical axis at the right-hand side. system, the amount of melt formed by melting at the eutectic temperature increases with the increasing mixing ratio of the hydrous silicic melt. Melt fraction at a given temperature for crustal materials also increases greatly with increasing water content (Fig. 10). A long-lived magma with <20 vol. % phenocrysts can form by mixing of at most 1 wt % water in granitic or gabbroic crusts (Fig. 11). Figures 8 and 10 indicate that the magma is still effectively liquid at the eutectic temperature in a hydrous granodioritic to granitic crust with water content more than a few weight per cent. The thermal evolution of the magma in such crusts is similar to that of the crust with the eutectic composition (see Fig. 2b) in the sense that the magma slowly solidifies by losing heat in CC Future problem The hypothesis of the two-stage thermal evolution model can be tested in the future by several different kinds of observations and theoretical modelling. The shift of melt fraction vs temperature relationship is controlled by melt segregation of partial melts and vapour circulation around a magma chamber. Observations and experimental works on melt segregation in the crust [for a summary of the recent progress, see Brown et al. (1995)], and studies on the physics of melt segregation in partially molten materials or the mushy layer at the floor and wall (e.g. Tait & Jaupart, 1989, 1992; Woods, 1991; Worster, 1991; Kerr, 1994) will give constraints on the segregation process. Isotope geochemistry of magma such as d18O and other geochemical studies on volcanic gases and fluids in geothermal areas will also be useful to test the contribution of hydrothermal water as selective assimilant. The probability of observing long-lived magmas as erupted magmas is dependent on timings of basaltic inputs and eruptions. Chronological approaches such as those of radioactive disequilibria of the U–Th series would give 251 JOURNAL OF PETROLOGY VOLUME 40 information on the timings of basalt inputs and eruptions. In the present study, simplified models for mass and heat transfer are applied to extract the fundamental physics; however, more realistic models which take into account the effects of fracturing of chamber wall and kinetics of crystallization and melting (e.g. Bergantz, 1995; Hort, 1997) may be necessary to quantitatively compare the model with these natural observations. ACKNOWLEDGEMENTS We thank Steve Sparks for his constructive comments on an earlier version of the manuscript. Comments by Steve Tait and two anonymous referees are also greatly appreciated. K.K. was supported by the Japanese Society for the Promotion of Science. REFERENCES Bergantz, G. W. (1995). Changing techniques and paradigms for the evaluation of magmatic processes. Journal of Geophysical Research 100, 17603–17613. Brown, M., Rushmer, T. & Sawyer, E. W. (1995). Introduction to special section: mechanisms and consequences of melt segregation from crustal protoliths. Journal of Geophysical Research 100, 15551– 15563. Campbell, I. & Turner, J. S. (1987). A laboratory investigation of assimilation at the top of a basaltic magma chamber. Journal of Geology 95, 155–172. Ewart, A. (1982). The mineralogy and petrology of Tertiary–Recent orogenic volcanic rocks: with special reference to the andesitic– basaltic compositional range. In: Thorpe, R. S. (ed.) Andesites: Orogenic Andesites and Related Rocks. New York: John Wiley, pp. 25–95. Ghiorso, M. S. & Sack, R. O. (1995). Chemical mass transfer in magmatic processes IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119, 197–212. Halliday, A. N. (1990). Reply to comment of R. S. J. Sparks, H. E. Huppert & C. J. N. Wilson on ‘Evidence for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in precaldera lavas of Glass Mountain’. Earth and Planetary Science Letters 99, 390–394. Halliday, A. N., Mahood, G. A., Holden, P., Metz, J. M., Dempster, T. J. & Davidson, J. P. (1989). Evidence for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in precaldera lavas of Glass Mountain. Earth and Planetary Science Letters 94, 274–290. Hort, M. (1997). Cooling and crystallization in sheet-like magma bodies revisited. Journal of Volcanology and Geothermal Research 76, 297–317. Huppert, H. E. & Sparks, R. S. J. (1988a). The generation of granitic magmas by intrusion of basalt into continental crust. Journal of Petrology 29, 599–624. Huppert, H. E. & Sparks, R. S. J. (1988b). Melting the roof of a chamber containing a hot, turbulently convecting fluid. Journal of Fluid Mechanics 188, 107–131. Huppert, H. E. & Turner, J. S. (1981). A laboratory model of a replenished magma chamber. Earth and Planetary Science Letters 54, 144–152. NUMBER 2 FEBRUARY 1999 Kaneko, K. (1996). The processes of the thermal and material evolution of a magma system in a crust. Ph.D. Thesis, University of Tokyo. Kaneko, K. & Koyaguchi, T. (1996). Evolution of magma system in the crust—investigations based on thought experiments and analogue experiments (in Japanese with English abstract). Memoirs of the Geological Society of Japan 46, 29–41. Kerr, R. C. (1994). Melting driven by vigorous compositional convection. Journal of Fluid Mechanics 280, 255–285. Koyaguchi, T. & Blake, S. (1991). Origin of mafic enclave: constraints on the magma mixing model from fluid dynamics experiments. In: Didier, J. & Barbarin, B. (eds) Enclaves and Granitic Petrology. Amsterdam: Elsevier, pp. 415–429. Koyaguchi, T. & Kaneko, K. (1995). Thermal and petrological evolution of magma system. Abstracts of Todai Symposium ‘The Role of Magmas in the Evolution of the Earth’. Tokyo: University of Tokyo, pp. 36–37. Mahood, G. A. (1990). Second reply to comment of R. S. J. Sparks, H. E. Huppert & C. J. N. Wilson on ‘Evidence for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in precaldera lavas of Glass Mountain’. Earth and Planetary Science Letters 99, 395–399. Marsh, B. D. (1981). On the crystallinity, probability of occurrences and rheology of lava and magma. Contributions to Mineralogy and Petrology 78, 85–98. Marsh, B. D. (1989). On convection style and vigor in sheet-like magma chambers. Journal of Petrology 30, 479–530. Martin, D. & Nokes, R. (1988). Crystal settling in a vigorously convecting magma chamber. Nature 332, 534–536. McBirney, A. R. (1980). Mixing and unmixing of magmas. Journal of Volcanology and Geothermal Research 7, 357–371. Piwinskii, A. J. & Wyllie, P. J. (1968). Experimental studies of igneous rock series: a zoned pluton in the Wallowa batholith, Oregon. Journal of Geology 76, 205–234. Piwinskii, A. J. & Wyllie, P. J. (1970). Experimental studies of igneous rock series: felsic body suites from the Needle Point Pluton, Wallowa batholith, Oregon. Journal of Geology 78, 52–76. Rutter, E. H. & Neumann, D. H. K. (1995). Experimental deformation of partially molten Westerly granite under fluid absent conditions, with implications for the extraction of granitic magmas. Journal of Geophysical Research 100, 15697–15751. Sisson, T. W. & Grove, T. L. (1993). Experimental investigations of the role of H2O in calc-alkaline differentiation and subduction zone magmatism. Contributions to Mineralogy and Petrology 113, 143–166. Sparks, R. S. J., Huppert, H. E. & Turner, J. S. (1984). The fluid dynamics of evolving magma chambers. Philosophical Transactions of the Royal Society of London, Series A 310, 511–534. Sparks, R. S. J., Huppert, H. E. & Wilson, C. J. N. (1990). Comment on ‘Evidence for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in precaldera lavas of Glass Mountain’ by A. N. Halliday, G. A. Mahood, P. Holden, J. M. Metz, T. J. Dempster, & J. P. Davidson. Earth and Planetary Science Letters 99, 387–389. Spera, F. J. (1979). Thermal evolution of plutons: a parameterized approach. Science 207, 299–301. Tait, S. & Jaupart, C. (1989). Compositional convection in viscous melts. Nature 338, 571–574. Tait, S. & Jaupart, C. (1992). Compositional convection in a reactive crystalline mush and melt differentiation. Journal of Geophysical Research 97, 6735–6756. Turcotte, D. L. & Schubert, G. (1982). Geodynamics: Applications of Continuum Physics to Geological Problems. New York: John Wiley. Turner, J. S. (1973). Buoyancy Effects in Fluids. Cambridge: Cambridge University Press. 252 KOYAGUCHI AND KANEKO THERMAL EVOLUTION OF MAGMAS Wolf, M. B. & Wyllie, P. J. (1995). Liquid segregation parameters from amphibolite dehydration melting experiments. Journal of Geophysical Research 100, 15611–15621. Woods, A. W. (1991). Fluid mixing during melting. Physics of Fluids A3, 1393–1404. Worster, M. G. (1991). Natural convection in a mushy layer. Journal of Fluid Mechanics 224, 335–359. Fig. A1. Fractional crystallization model in a binary system with a linear liquidus surface. (See text for explanation.) APPENDIX: FRACTIONAL CRYSTALLIZATION MODEL IN A BINARY SYSTEM Let us consider a material in a binary system with a linear liquidus surface (Fig. A1), where compositional scale is normalized, such that the solid composition is C = 1 and the liquid composition at the EFT is C = 0. For simplicity it is assumed that the density difference between the solid and liquid is small enough for values of volume fraction to approximately represent those of mass fraction. The initial bulk composition is, therefore, C = φ, where φ is the critical melt fraction between effective solid and liquid. We take φ = 0·5 in the following calculations. If the material of C = φ has initially a higher melt fraction (say, u0) and the crystallization proceeds as temperature decreases at a given cooling rate, then the crystals settle to form a cumulate pile at the bottom and the depth of ‘magma’ (i.e. the region in an effectively 253 JOURNAL OF PETROLOGY VOLUME 40 NUMBER 2 FEBRUARY 1999 liquid state) decreases. Three more assumptions are made: (1) the melt fraction of the cumulate pile has also the critical value, φ; (2) the main magma body is homogenized; (3) no chemical interaction between the interstitial melt and the main magma body is allowed. Under these assumptions, the rate of change in melt composition is approximated as u −φ dC =− 0 dt u0sF (A1) where sF is the timescale to reach the EFT. The growth rate of the cumulate pile is given by 1 1−u dz =− vs dt H u−φ (A2) Fig. A2. Composition of magma that is incorporated into the cumulate pile against the height from the bottom for various ratios of timescale of settling to that for temperature to reach the EFT of the crust [H/(vssF)]. where H is the total height, vs is the settling velocity of crystals, z is the normalized depth of the magma (see Fig. A1), and u is the melt fraction of the magma. The mass balance at the front of the growing cumulate pile can be expressed as u dC d(1−u) (1−u+uC)−(1−φ+φC) dz = − dt z(1−C) dt 1−C dt (A3) where C is the melt composition. From equations (A1)– (A3), we can numerically obtain the evolutions of composition of magma that is incorporated into the cumulate pile (Fig. A2), phenocryst content, and the depth of magma layer (Fig. A3). For a given initial condition of u0, the behaviours of these evolutions depend on the ratios of the timescale of settling to that for temperature to reach the EFT [i.e. H/(vssF)] alone. To evaluate the effects of fractional crystallization during the C&M stage, the values at the moment when the temperature reaches the EFT of the surrounding crust against varying ratios of the two timescales are the most meaningful, and these are shown in Fig. 6. 254 Fig. A3. The relationships between depth and melt fraction of magma layer for various ratios of timescale of settling to that for temperature to reach the EFT of the crust [H/(vssF)]. The depth of magma layer starts at z = 1, decreases with time as crystallization proceeds, and terminates at the position of circles when temperature reaches the EFT of the crust.
© Copyright 2026 Paperzz