Physics of the Earth and Planetary Interiors 149 (2005) 259–278 The combined influences of variable thermal conductivity, temperature- and pressure-dependent viscosity and core–mantle coupling on thermal evolution A.P. van den Berga,∗ , E.S.G. Raineyb,c , D.A. Yuenc a Department of Theoretical Geophysics, Institute of Earth Sciences, Utrecht University, 3508 TA Utrecht, Netherlands b Planetary Sciences Division, Code-150-21, Caltech, Pasadena, CA 91125, USA c Department of Geology and Geophysics and University of Minnesota Supercomputing Institute, University of Minnesota, Minneapolis, MN 55455-0219, USA Received 21 July 2003; received in revised form 27 August 2004; accepted 13 October 2004 Abstract Most convection studies of thermal history have not considered explicitly the thermal interaction between the mantle flow and the core. We have investigated the influences of variable thermal conductivity and variable viscosity (temperature- and pressuredependent) on the boundary layer and thermal characteristics of the D"" layer, and the evolution of the thermo-mechanical profiles of horizontally averaged viscosity and thermal conductivity. Viscosity contrast due to temperature dependence of up to 30,000 has been considered. Our results show clearly that variable thermal conductivity, though small in magnitude as compared to variations in the viscosity, does exert a significant delaying influence on mantle cooling, thereby keeping the Urey ratio low, reducing the growth of the bottom thermal boundary layer, and changing the viscosity profiles over time. A higher temperature at the core– mantle boundary increases the overall time-dependent behavior of the thermal boundary layers. Enhanced radiative conductivity results in faster cooling, opposite to the effect of the phonon conductivity component and a superadiabatic temperature gradient in the deep lower mantle. Finally, the initial value of the core–mantle boundary temperature can be inferred to wield a strong influence on the subsequent mantle thermal evolution in this model with both variable thermal conductivity and viscosity. We may conjecture that other rheological and conductivity complexities, such as grain-size dependence of mantle properties, would also have an impact on the current state of the mantle resulting from the primordial thermal condition. © 2004 Elsevier B.V. All rights reserved. 1. Introduction ∗ Corresponding author. E-mail addresses: [email protected] (A.P. van den Berg); [email protected] (E.S.G. Rainey); [email protected] (D.A. Yuen). 0031-9201/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.pepi.2004.10.008 For over three decades now, the thermal evolution of the mantle has been studied with the convection paradigm with the feedback mechanism of temperature-dependent viscosity being emphasized 260 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 (e.g. Tozer, 1972). The most extensive work using numerical convection models on thermal history has been conducted with depth-dependent viscosity (Butler and Peltier, 2002). While the effects of temperature- and pressuredependent viscosity are well known in the steady state (e.g. Christensen, 1985), not much work has been carried out in thermal evolution with variable viscosity, save for the work by DeLandro-Clarke and Jarvis (1997). Moreover, a model for mantle thermal conductivity based on phonon solid-state physics and infrared spectroscopy was developed by Hofmeister (1999). Contributions from both phonon and photon conductivity are included in this model. Recently on the basis of spectroscopic work, Badro et al. (2004) have argued for the importance of radiative thermal conductivity in the deep mantle. van den Berg et al. (2002) and van den Berg and Yuen (2002) have shown that variable thermal conductivity can delay the secular cooling of the mantle with a constant viscosity model. Since variations of viscosity in the course of thermal evolution are much greater than changes in the thermal conductivity, it is therefore important to evaluate the influence of variable viscosity on the effects of delayed secular cooling and also the stabilization of boundary layer activities at the core– mantle boundary (CMB) from increasing the radiative contribution to the thermal conductivity (Dubuffet et al., 2002). Another important aspect of thermal history is the influence from thermal coupling of the mantle to the core. This was first studied within the framework of one-dimensional parameterized convection models by Sharpe and Peltier (1978) and Schubert et al. (1979) and in fully two-dimensional (Steinbach et al., 1993; Honda and Yuen, 1994) and three-dimensional (Yuen et al., 1994) convection models. Buffett et al. (1992) and recently Buffett (2003) pointed out the importance of core–mantle interactions in thermal-chemical evolution. The heat flux at the core–mantle boundary (CMB) is of particular interest for planetary evolution. It controls the relative partitioning between bottom heating and internal heating in the lower mantle, and it also has important implications for the geodynamo and the chemical composition the core. The current CMB heat flux is relatively poorly constrained, but recent estimates indicate that the heat flux may be much higher than early estimates, perhaps as high as 12 TW (Buffett, 2003). An important minimum constraint on core heat flux is the heat necessary to drive the geodynamo and generate a magnetic field, which has probably existed for at least 3.5 Gyr. Best estimates indicate that prior to the solidification of the inner core, known sources of heat in the core are insufficient to drive a magnetic dynamo. 40 K, which is depleted in the mantle, was suggested as a possible source of radioactivity in the core that could provide heat necessary for the geodynamo (e.g., Hall and Murthy, 1971; Gessman and Wood, 2002). Although potassium was not thought be a siderophile, recent experimental evidence shows that 40 K can enter iron sulphide melts under core conditions (Murthy et al., 2003). The amount of 40 K in the core can be constrained by the core heat budget, which depends on how much heat is conducted from the core into the mantle. For obtaining better estimates of the heat flux at the CMB, it is necessary to use a model that includes a realistic mantle thermal conductivity, especially in the lower mantle, where radiative conductivity effects can be stronger than the phonon conductivity (Yuen et al., 2000) and can also be enhanced by iron concentration in the D"" layer (Manga and Jeanloz, 1996). In Section 2 we describe the models for 2D mantle convection and the thermal conductivity. In subsequent sections we focus respectively on the the effects of varying initial CMB temperatures, the enhancement of radiative thermal conductivity and the temporal development of the thermal structure of the mantle in the core-coupling model. In the final section we will state our conclusions and offer our perspectives for the role played by variable viscosity acting in concert with variable thermal conductivity in particular in view of the effects of enhanced radiative conductivity as indicated in recent mineral physics results (Badro et al., 2004), and core–mantle coupling in shaping the thermal history. 2. Description of the convection, conduction and viscosity models We use a 2D mantle convection model including thermal coupling to the core. Fig. 1 shows a diagram of the cartesian computational domain, illustrating the thermal coupling between mantle and core included in our model. An aspect-ratio of 2.5 for the compu- A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 261 Fig. 1. Domain diagram showing the earth’s thermally coupled mantle and core in a spherical configuration (top) and in a cartesian 2D box of aspect ratio 2.5, used in the numerical mantle convection model. The core is represented by an isothermal heat reservoir, thermally coupled to the convecting mantle. This core reservoir is cooled by the heat flux into the mantle driven by the temperature contrast δT across the bottom boundary layer of the mantle. In this model the temperature contrast across the convecting mantle TCMB (t) − Tsurface decreases with the cooling of the core. tational domain has been considered throughout. This same aspect-ratio was employed in our previous works (van den Berg and Yuen, 1998; van den Berg et al., 2002). The mantle convection model is based on the extended Boussinesq approximation for an infinite Prandtl number, incompressible fluid (Steinbach et al., 1989). In this model conservation of mass, momentum and energy and the constitutive rheological relation are expressed in the following non-dimensional equations ∂j u j = 0 (1) − ∂i #P + ∂j τij = αRaTδi3 (2) τij = η(T, P)(∂j ui + ∂i uj ) (3) DT = ∂j (κ(T, P)∂j T ) + αDiw(T + T0 ) Dt Di + ( + RH(t) Ra (4) Symbols used in (1)–(4) are defined in Table 1. In Eq. (4) D/Dt denotes the substantive derivative. H(t) is an exponential decaying function and R is a nondimensional measure of radiogenic strength. For the internal heating of the model we use a uniform distribution with exponential time dependence H(t) characterized by a half-life time of 2.5 Gyr, and an initial value of the internal heating number R equal to 20, corresponding to an internal heating, which is about a factor of two stronger than the present-day chondritic value. For the non-dimensionalization scheme we used the depth of the convecting layer h as the spatial scale and h2 /κ0 , a thermal diffusion time of the layer, as the time scale. The temperature scale #T corresponds to the initial temperature contrast across the layer, of an intermediate case of several models with different initial CMB temperature TCMB (0). We used temperature contrasts of 3000, 3500 and 4000 K, corresponding to initial core–mantle boundary temperatures TCMB (0) of 262 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Table 1 Physical parameters Symbol Definition Value Unit h z P #P T Tsurface #T ui eij = ∂j ui + ∂i uj "1/2 ! 1 e= 2eij eij w η(T, z) = η0 exp(cz − bT ) η0 τij = ηeij ( = ηe2 #α α(z) = [c(1 − z) + 1]3 #α = α(1) c = #α1/3 − 1 α0 ρ ρ0 cp k k0 a γ K0 K0" b0 b1 b2 b3 k κ= ρcp g # $ −t H(t) = H0 exp τ τ H0 H 0 h2 R= cp κ0 #T ρ0 α0 g#Th3 Ra = κ0 η0 α0 gh Di = cp qC (t) X Height of the mantle model Depth coordinate aligned with gravity Static pressure Dynamic pressure Temperature Surface temperature Temperature scale Velocity field component Strain rate tensor 3 × 106 – – – – 273 3500 – – m – – – – K K – – Second invariant of strain rate – – Vertical velocity aligned with gravity Temperature and pressure/depth dependent viscosity Viscosity scale value Viscous stress tensor Viscous dissipation function – – – – – – Pa s – – Depth dependent thermal expansivity – – – – K−1 – kg m−3 J K−1 kg−1 – W m−1 K−1 – – GPa – Thermal expansivity scale value Density Density scale value Specific heat Thermal conductivity Conductivity scale value Conductivity powe-law index Grueneisen parameter Bulk modulus Pressure derivative of bulk modulus Coefficient photon conductivity 2 × 10−5 – 4000 1250 – 4.7 0.3 1.2 261 5 1.7530 × 10−2 −1.0365 × 10−4 2.2451 × 10−7 −3.4071 × 10−11 Thermal diffusivity – – Gravitational acceleration 9.8 m s−2 Time-dependent internal heating – W kg−1 Dimensional decay time of radioactive heating Dimensional value of internal heating 3.6 – Gyr W kg−1 Non-dimensional internal heating number 20 – Rayleigh number – – Dissipation number 0.47 – Average heatflow density at the CMB Ratio of core to mantle heat capacity – 0.44 – – A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 3273, 3773 and 4273 K. The initial temperature field of the mantle in the cases with different TCMB (0) were computed by applying an appropriate uniform scaling factor to the initial mantle temperature of the intermediate case (TCMB (0) = 3773 K). The latter is obtained from a statistically steady-state equilibrium model run with enhanced (R = 40) and constant internal heating with a zero heat flux bottom boundary. For the thermal diffusivity κ = k(T, P)/ρcp we use the temperature- and pressure-dependent conductivity Hofmeister model (Hofmeister, 1999) # $ 298 a k(T, P) = k0 T ! # $ " 1 × exp − 4γ + α(P)(T − 298) 3 # $ 3 % K" P × 1+ 0 + fbi T i (5) K0 i=0 In (5) the first term gives the phonon contribution to the effective conductivity and the second term is the contribution from photon transport. The amplification factor f (e.g. van den Berg et al., 2002) of the photon term, valued f = 1 in the Hofmeister model (Hofmeister, 1999), is used here as a control parameter to vary the relative contribution of both mechanisms in the effective thermal conductivity. We consider in particular models with f values of 0, 1, 2 and 5 to investigate the impact of the radiative thermal conductivity on the model behavior. We have not considered the grain-size dependence of thermal conductivity, which can vary non-monotonically with depth (Hofmeister, 2004). The phonon term decreases with increasing temperature, ∂klat /∂T < 0 and increases with increasing pressure, ∂klat /∂P > 0. The photon term on the other hand increases with temperature ∂krad /∂T > 0 and is insensitive to pressure ∂krad /∂P = 0. To interpret the numerical modelling results we will also use the thermal resistivity, defined as the inverse of the thermal conductivity r = 1/k, in analogy with the theory of electricity. One-dimensional depth profiles of horizontally averaged resistivity can then be integrated from a boundary point zb to obtain a resistance profile R(z) through thermal boundary layers of the mantle, the lithosphere and the CMB region ( & z' 1 R(z) = (6) dz" ") k(z zb 263 This thermal resistance has been in use in geothermics as a means of obtaining reliable heatflow estimates from bore holes with strongly fluctuating conductivity profiles (Beardsmore and Cull, 2001; Bullard, 1939). For the rheological model we have chosen an exponential temperature and depth (pressure) dependent viscosity for Newtonian rheology η(T, z) = η0 exp(cz − bT ) (7) where c, b are defined in Table 1 in terms of the viscosity contrasts across the convecting layer due to depth (pressure) (#ηP ) and temperature (#ηT ), respectively. The value of #ηP is fixed at 100. For most cases, the value of #ηT is 3000, but for comparison we also show some contrasting cases. Eqs. (1), (2) and (4) are solved by using finite element methods for the spatial discretization, and applying a penalty function method for the continuity equation and Stokes momentum equations (1) and (2). The energy equation (4) which drives the time dependent system is integrated in time using a predictor corrector method (van den Berg et al., 1993). The finite element mesh consists of 150 × 140 nodal points in the horizontal and vertical direction, respectively. Mesh refinement was applied near the horizontal boundaries, where the vertical nodal point spacing was reduced to 6 km from a value of 30 km in the interior domain. Mesh refinements near the thermal boundary layers is essential in calculations using variable conductivity due to the occurrence of strong temperature gradients and similar sharp variations in the effective thermal conductivity in the boundary layer (see Yuen et al., 2000), which need to be resolved numerically. Especially the computation of the surface heat-flux requires a very high resolution of the finite element mesh (van den Berg et al., 2001). We use free-slip impermeable boundaries and a fixed top surface temperature of 273 K. On the vertical boundaries a zero heat flux symmetry condition was applied. The model runs were started from a statistically steady state obtained for a zero heat flux bottom boundary and constant internal heating. Thermal coupling between mantle and core is represented by an isothermal heat reservoir of the core, shown in Fig. 1, where the temperature TC is controlled by the average heat-flow from the core–mantle bound- 264 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 ary qC described by the ordinary differential equation (Steinbach et al., 1993) dTC A qC (t) =− dt ρC cPC VC (8) where A is the area of the core–mantle boundary surface and CC = ρC cPC VC is the total heat capacity of the core. The core heat capacity is expressed as a fraction X of the mantle heat capacity CM , resulting in an O.D.E. for TC (t): dTC 1 =− qC (t) dt Xρcp h (9) where ρ and cp are the mantle values of density and heat capacity. Parameter values are given in Table 1. At each time step TC is updated by integrating (9), using the average heatflow value qC , computed from the finite element solution for the mantle temperature field. This is done by forward extrapolation in time of the heat flux in an Euler type scheme. The updated uniform core temperature is then taken as a time dependent boundary condition for the finite element computation of the mantle temperature in the next time step, so TCMB (t) = TC (t) in (9). We note that analytical asymptotic methods (Solomatov and Zharkov, 1990) for treating thermal history, though capable of handling variable viscosity, may be hard-pressed to apply in the case of variable thermal conductivity. 3. Results A comparison of a series of temperature snapshots for typical secular cooling runs for two different values of the TCMB (0) = 3273 and 4273 K is shown in Fig. 2. The snapshots represent a time span of about the age of the Earth, illustrating the effects of different initial TCMB (0), showing faster cooling for the hotter initial temperature case. In previous work we have investigated the impact of variable conductivity on the secular cooling of the convecting mantle, restricted to isoviscous models (van den Berg and Yuen, 2002) and to models with temperature dependence of the viscosity limited to around 1000 (van den Berg et al., 2004). Here we consider models including pressure and temperature dependent viscosity given by (7) with a higher value of the temperature dependence. In contrast to previous work we also focus on the opposite role of the different heat transport mechanisms (phonons versus photons) in the composite conductivity model. Finally we introduce here thermal coupling between the convecting mantle and a thermal reservoir representing the core. Photon conductivity is likely to be dominant in the lower mantle and hence important for thermal core/mantle coupling. Phonon conductivity is suppressed by the 1/T temperature dependence. The effect of varying #ηT on the overall cooling history of the coupled mantle and core system is presented in Figs. 3 and 4. The cooling curves in Fig. 3, show the volume averaged temperature of the mantle, comparing also the variable conductivity cases with the corresponding constant conductivity cases. The corresponding constant conductivity cases have the same surface conductivity value as the variable conductivity cases. For the conductivity model with f = 1 this also corresponds to an approximately similar value of the volume average conductivity. This was also shown in previous work, Fig. 7 of van den Berg and Yuen (2002). The main feature of the results shown in Figs. 3 and 4 is a similar cooling delay of the variable conductivity models with respect to the corresponding constant conductivity models. This amounts to an accumulated delay time of almost two billion years at a model time of 4.5 Gyr. These results indicate that the delay we found earlier in secular cooling in isoviscous models with variable conductivity is a robust phenomenon, also in models including variable viscosity. The results also show that the trend in the cooling rate, for increasing #ηT , is non-monotonic. It appears that cooling is slightly slower for #ηT = 3 × 103 (Fig. 3b) than for #ηT = 3 × 102 (Fig. 3a). Increasing the temperature to #ηT = 3 × 104 the cooling rate increases also (Fig. 3c). This non-monotonic trend is the same for constant conductivity models represented by the dashed curves. The small difference in the overall cooling behavior between the three contrasting viscosity cases is surprising inview of the significant difference of the interior viscosity shown for these cases shown in Fig. 4. It appears from these results that one cannot simply apply the temperature dependence of the viscosity to predict an increased cooling rate for increasing #ηT . The pressure dependence of the viscosity complicates the details of the resulting temperature distribution which feeds back into the viscosity, resulting in this non-monotonic behavior. A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 265 Fig. 2. Snapshots of the temperature field for two different initial CMB temperatures TCMB = 3273 and 4273 K. The variable conductivity model used is the same in both cases with f = 1. Different temperature scales have been used between the initially hotter and cooler model cases. The difference in the thermal evolution of the convecting mantle is illustrated by a series of snapshots spanning the age of the earth. The hot model in the righthand column shows a significantly faster cooling than the initially cooler model. The hot model also shows smaller scale convective features then the cooler model. Depth profiles of temperature and viscosity are shown in Fig. 4 for three cases with contrasting temperature dependence of the viscosity #ηT = 300, 3000 and 30,000. The conductivity model used is the same in all three cases, corresponding to the Hofmeister (1999) model with f = 1. The initial value TCMB (0) is 3773 K in all cases and the snapshot corresponds to an integration time of 4.428 Gyr. Internal temperatures for the three cases shown are roughly similar, with a larger temperature difference of several hundred degrees in a layer of 500 km above the CMB. The variation in the corresponding horizontally averaged viscosity is between one and two orders of magnitude between the different model cases inline with the differences in #ηT . In order to investigate the mechanism behind the cooling delay of the variable conductivity models we have applied a 1D depth dependent conductivity model. The corresponding conductivity, ka (z), is computed from the horizontally averaged conductivity, taken from a variable (Hofmeister, 1999) conductivity model with f = 1 substituted in (5). The 1D profile is defined as the time-averaged value, for an averaging time window of 5 Gyr, of horizontally averaged conductivity snapshots. The result of this space and time averaging of the conductivity is shown in Fig. 5a. The small variation of the effective conductivity profiles over time, due to the secular cooling, is illustrated by the width of the bundle of black curves. The time averaged profile ka (z) is represented by the red curve. We compared the thermal history of the variable conductivity model with the model based on the 1D profile ka (z). The viscosity model is kept the same in this com- 266 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Fig. 3. Thermal evolution curves showing volume averaged temperature of the convecting mantle against integration time %T (t)&, for three models characterized by different values of the temperature dependence of the viscosity, #ηT = 300, 3000 and 30,000. The variable conductivity model used is the same in all cases with f = 1. Results for variable k, f = 1 are compared with corresponding constant conductivity models. A strong delaying effect on secular cooling of variable conductivity is clearly shown by these results. Furthermore this delay is robust for increasing values of the temperature dependence of the viscosity. parison, #ηT = 3 × 103 , #ηP = 102 . Time series of global quantities for both cases are shown in Fig. 5b and c. Fig. 5b shows the evolution of the volume averaged mantle temperature and the temperature of the core heat reservoir, indicating almost identical thermal evolution. Fig. 5c shows a corresponding times series of the CMB and surface heat flux. The equivalence of the heat flux level for both models shown in this frame is in agreement with the coincidence of the thermal history curves in Fig. 5b. Note that the heat flux level has the right order of magnitude for Earth. Experiments with purely pressure dependent conductivity, using the same surface value, have shown a negligible cooling delay compared to the full pressure and temperature dependent conductivity models (van den Berg et al., 2004). These results show that the delay in secular cooling of the variable conductivity models can be ascribed to the average 1D structure of the conductivity profile and in particular to the low conductivity zone (LCZ) at shallow depth. The effect of this LCZ is to increase the thermal resistance of the lithosphere, which suppresses conductive heat transport through the lithosphere and results in delayed secular cooling. 3.1. Effect of initial CMB temperature TCMB (0) on mantle evolution In Fig. 6 we show the evolution over time of the volume averaged temperature %T (t)&, for three differ- Fig. 4. 1D depth profiles of horizontally averaged temperature (left) and viscosity (right), for the same models as in Fig. 3. The results shown correspond to an integration time of 4.428 Gyr and an initial CMB temperature TCMB (0) = 3773 K for all models. The main differences in the temperature are in the bottom boundary layer, which is also reflected in the shape of the viscosity hill in the deep lower mantle. The difference in the viscosity profiles are up to about two orders of magnitude, in line with the small temperature differences and the viscosity parameter values. A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 267 Fig. 5. Lefthand frame: (a) (black) Snapshots, evenly spaced in time between 0 and 4.5 Gyr, of horizontally averaged conductivity profiles, (red) time averaged conductivity profile computed from the black profiles. Righthand frames: Time series comparing the thermal evolution of a variable conductivity model (f = 1 denoted by black curves) and a time averaged 1D depth dependent conductivity model ka (z), shown in frame (a), including a shallow low conductivity zone, denoted by red curves. (b) Core temperature (top curves) and average mantle temperature (bottom curve). (c) Heat flux through CMB (lower curves) and Earth’s surface (top curves). Fig. 6. Thermal evolution curves showing volume averaged temperature of the convecting mantle against integration time %T (t)&. The three curve groups are labeled with the corresponding values of the initial CMB temperature, TCMB (0) = 3273 K, 3773 K and 4273 K. The different curves in each group represent models with different conductivity models. The effect of increasing TCMB (0) is an increased slope representing a higher cooling rate. Furthermore the constant conductivity model cases (dashed lines) show the fastest secular cooling. The largest cooling delay is obtained for the models with f = 0, corresponding to absence of radiative conductivity. The effect of increasing the radiative component (increasing f ) is to speed up the cooling rate. ent values of the initial CMB temperature TCMB (0). For each value of TCMB (0), represented by the three curve groups in Fig. 6, we also show a comparison of results for different conductivity models, including a constant conductivity case represented by the dashed curves and four variable conductivity cases for different values of the multiplication factor f for the radiative component of thermal conductivity. In general, the constant conductivity models show faster mantle cooling than most of the variable conductivity cases, only surpassed by the variable k models with strongly enhanced conductivity (f = 10). Among the variable conductivity models the model with f = 10 shows the fastest cooling which is related to the different structure of the low conductivity zone (shown below) for shallow depth. The models with f = 0 and 1 show a similar cooling history with an overall cooling delay of about 2 Gyr with respect to the corresponding constant conductivity case. The effect of the initial CMB temperature, TCMB (0), results in an increase of the overall cooling rate as is apparent from the increasing slope of the %T (t)& curves. At the same time the cooling delay between the variable k and constant k remains fairly constant. These results with variable viscosity corrobarate our earlier results for simpler models with constant viscosity and without thermal coupling between mantle and core (van den Berg et al., 2002; van den Berg and Yuen, 2002). Cooling rates of the convecting mantle are shown in Fig. 7 for the same three values of the initial CMB temperature TCMB (0). The cooling rates, d%T &/dt, were computed by a central difference approximation of the time series of the volume averaged temperature %T (t)&. For each value of TCMB (0) a variable conductivity model with f = 1 is compared with a corresponding constant conductivity case. The trend between the different panels is an increase of the cooling rate in line with a similar trend in the slopes of %T (t)& shown in Fig. 6. Increasing values of the initial CMB temperature are also reflected in the degree of time dependence. The higher TCMB (0) cases are characterized by rapid fluctuations of the cooling rate, in contrast to the smooth curves of the volume averaged temperature. This is re- 268 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Fig. 7. Time evolution of the rate of secular cooling, d%T &/dt, for different initial CMB temperatures, TCMB (0) = 3273 K (a), 3773 K (b) and 4273 K (c). A comparison is made between constant conductivity (black curves) and variable conductivity (f = 1). The lower absolute values of the variable conductivity models is consistent with the delay in secular cooling of the corresponding curves in Fig. 6. Increasing the initial CMB temperature results in a higher cooling rate in line with Fig. 6 and a much stronger time dependence of the the cooling rate. lated to the increasing fluctuations in the surface heat flow, which contribute to the secular cooling, with increasing Rayleigh number (van den Berg et al., 1993; van den Berg and Yuen, 2002). The delay in secular cooling of the variable conductivity cases, apparent in Fig. 6, is reflected in Fig. 7 by the consistently lower cooling rates for the variable conductivity curves labeled f = 1. In Fig. 8 we show time series of the temperature of the core heat reservoir for the same models as used for Fig. 6, i.e. three different initial core temperatures and five different conductivity models. These core cooling results show a similar trend as in Fig. 6 for the mantle. However, there is a difference in the sensitivity to the contribution of the radiative conductivity expressed in the amplification factor f. The core temperature seems more sensitive for increased f values. This is due to the fact that models with f = 5 (red curves) show already faster core cooling than the constant conductivity runs, whereas Fig. 6 shows that the switch to faster mantle cooling for increased f occurs later, between f = 5 and 10. An explanation of this different behavior is that the accelerating effect of f on mantle cooling, resulting in an increasing effect on the temperature contrast between mantle and core, is compounded with the enhanced cooling by radiative heat transport across the CMB for increased values of f. This sensitivity is also increasing for higher initial core temperatures, in agreement with the temperature dependence of the radiative conductivity. Time evolution of the CMB heat flux is shown in Fig. 9, for several model cases with the same initial core temperatures as in Figs. 6 and 8. The results show that the variable conductivity with f = 0 produces the lowest core heat flux. For the constant conductivity case, represented by the dashed line, the core heat flux is relatively high and for the variable conductivity cases with enhanced conductivity, for 10 the highest core heat flux is obtained. The trend in these core heat flux results is consistent with the corresponding core temperature curves shown in Fig. 8, in agreement with the fact that core temperature is obtained in our model by integrating the cmb heat flux according to (9). Evolution of the heat flux from the core has been studied mainly in parameterized models characterized by smooth time variations (Buffett, 2003). Our model results show a remarkably high fluctuation level of the core heat flux and one could speculate that such fluctu- A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Fig. 8. Thermal histories of the core for the same models as shown in Fig. 5. Three curve groups are labeled with the initial core temperature. Each group includes results for different conductivity models. The trends in these curves illustrate the key role of radiative conductivity in controlling core cooling. The slowest core cooling is obtained for the f = 0 model without any radiative conductivity, krad = 0, and the cooling rate increases with increasing relative contribution of krad controlled by the amplification factor f. Furthermore this effect is stronger in a hotter earth in line with the temperature dependence ∂krad /∂T > 0. ations could impact the geodynamo process and leave their marks in the paleomagnetic intensity record. 3.2. Influence from enhanced thermal radiative conductivity As discussed above, the parameter f represents a measure of the radiative contribution to the thermal conductivity, with f = 1 having the same value as the model presented by Hofmeister (1999). Fig. 10 compares the two-dimensional temperature fields for values of constant thermal conductivity, and variable conductivity with f = 1 and 0 (purely lattice conductivity) and f = 5 (enhanced radiative conductivity). It is clear that with variable k, the entire convective region is hotter, in comparison to constant k (Dubuffet et al., 1999, 2002). As was observed in the previous work using constant viscosity (van den Berg et al., 2002), we find that the Hofmeister variable conductivity results in a reduced overall convective vigor. The constant k case has stronger downwellings and earlier, betterdeveloped plumes. For different values of f, there is 269 Fig. 9. Heat flow through the core–mantle boundary against time, for the same model cases as in Fig. 7. The core heatflow increases with the relative contribution of the radiative conductivity krad in agreement with the core cooling histories shown in Fig. 7. Core heat flux is highly time dependent in these models with peak to peak values of about 100%. This is a result of the strong time dependence of cold downwellings cooling the hot core in these models which are largely cooled from above. still present a noticeable difference in the temperature fields between the constant k and variable k models, although less so in the enhanced krad case f = 5, which has the fastest cooling rate of the variable k models considered here (Fig. 6). The viscosity profiles for the entire mantle are shown in Fig. 11 for constant conductivity, f = 1 and 5 and three different values of initial CMB temperature TCMB (0). It is interesting to note that a sharper low viscosity valley is produced by a lower temperature at the CMB. For constant viscosity Dubuffet et al. (2002) have noted that there is a bifurcation in the behavior of the convective solution, as f is increased beyond a certain value, which depends on the TCMB and on the amount of internal heating. In Fig. 12 we show the one-dimensional profiles of the horizontally averaged temperature %T &, the viscosity %η& and the thermal conductivity %k& for f = 0, 1 and 5. The temperature profiles show that the effect of introducing variable conductivity is to make the thermal 270 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Fig. 10. Temperature (left) and streamfunction (right) snapshots for time t = 4.428 Gyr for models with initial CMB temperature TCMB (0) = 3773 K, and different thermal conductivity. Fig. 11. Global 1D depth profiles of horizontally averaged viscosity, for integration time t = 5.161 Gyr, for different conductivity models, constant k (a), variable conductivity with f = 1 (b) and f = 5 (c). In each frame different curves are shown for models with different initial CMB temperature TCMB (0) = 3273, 3773 and 4273 K. A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 271 Fig. 12. Global 1D depth profiles of horizontally averaged temperature (left), conductivity (middle) and viscosity (right). Results are for the same initial CMB temperature of TCMB (0) = 3773 K and for an integration time t = 2.952 Gyr. boundary layer thinner at the top and thicker at the bottom, compared to the constant conductivity case. This can be interpreted in terms of the different systematics of the thermal resistance profiles defined in (6), through the top and bottom boundary layers, as discussed in more detail below. Increased thermal resistance of the top boundary layer has in increasing effect on the temperature contrast across the lithosphere. Similarly the decreased resistance near CMB results in a decreased temperature contrast across CMB. We see that the viscosity profiles are all very similar but that the conductivity and temperature profiles reveal sharp changes with the amount of enhanced radiative conductivity from subadiabatic to superadiabatic gradient. Thus one cannot casually employ a constant value thermal gradient in the lower mantle for determining the viscosity profile (Yamazaki and Karato, 2001). There is a dramatic variation in the shape of %k& for values of f exceeding 3. This ‘transition’ is also reflected in the character of the temperature gradient in the bottom part of the mantle. The temperature gradient for f = 5 shows a superadiabatic character, in contrast to the models with lower f values, which show a subadiabatic geotherm in the bottom parts of the mantle. We obtained similar results in models with a zero heat flux bottom boundary condition (van den Berg et al., 2002). The super adiabatic geotherm is consistent with the mineral physics result of da Silva et al. (2000) on the basis of the bulk modulus variation with depth. This superadiabatic character of the geotherm may indicate an enhanced radiative heat transfer in the deep mantle. This high temperature gradient in the lower mantle, due to enhanced radiative heat transfer, is also reminiscent of temperature distributions resulting from an abyssal source of radiogenic heating invoked in the deep mantle model by Kellogg et al. (1999). The evolution of the thermal-mechanical structure near the CMB is shown in Fig. 13, where we plot the %T &, %η& and %k& profiles for constant conductivity, f = 0, 1 and 5. The effect of variable thermal conductivity is to retard the growth of the thermal boundary layer. With larger values of f the growth rate of the boundary layer approaches that associated with a constant thermal conductivity. We see that the more efficient heat transfer in the case of f = 5 gives rise to a cooler lower mantle temperature and hence a shallower trough in the viscosity at the CMB. The time evolution of the temperature contrast δT across the thermal boundary layer at the CMB is shown in Fig. 14. These temperature contrasts where calculated as the difference, δT (t) = TCMB (t) − TA (zCMB , t), between the actual CMB temperature and the extrapolated CMB temperature TA (zCMB , t) of a mantle adiabat obtained by a least squares estimate, for the depth range between 1000 and 1800 km depth, 272 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Fig. 13. Profiles of horizontally averaged temperature (left column), conductivity (middle column) and viscosity (righthand column) for the bottom 300 km of the mantle. Models shown are for an initial CMB temperature TCMB (0) = 3773 K. Different curves in each panel correspond to different integration times, t = 0, 2.36 and 5.02 Gyr, and the direction of increasing time is indicated by the arrows. The columns shown correspond to different conductivity models, constant k (top), f = 1, 0 and 5. A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Fig. 14. Evolution in time of the temperature contrast across the thermal boundary layer at the CMB. Different initial CMB temperature are shown TCMB (0) = 3273, 3773, 4273. Each panel shows results for different conductivity models, constant k, f = 0, 1 and 5. 273 based on the horizontally averaged 1D temperature profile. The constant k models show the highest temperature contrast, as are also illustrated in the temperature profiles of Fig. 12. The temperature contrast for the variable k cases increases with f between f = 0 and 5. A more rapid development of a higher temperature contrast δT may explain a larger tendency towards early plume formation from the CMB in the constant conductivity cases. Considering the effects of the structure of the bottom thermal boundary layer on the cooling of the core, we see that the high δT value for the constant k case is apparently compensated by a lower conductivity value, as shown in Fig. 12(middle), resulting in an intermediate core heat flux, which can be observed clearly in Fig. 9. More insight can be obtained in the trends in the model results for enhanced radiative conductivity by comparing profiles of the 1D thermal resistance defined Fig. 15. Vertical profiles of horizontally averaged conductivity %k& (left), resistivity %1/k& (middle) and corresponding thermal resistance R(z) (right). Blue and green curves correspond to snapshots, evenly spaced in time from 0 to 4.55 Gyr, for two models with contrasting contribution of the radiative conductivity, krad = 0, f = 0 (green), amplified krad , f = 10 (blue). The top row of frames shows a zoom in on the top 500 km of the model. The bottom row shows the bottom boundary layer. 274 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 in (6). This quantity represents an integrated measure of the variation with depth of the parameters controlling conductive heat transport, conductivity and resistivity. Since the depth variation of the resistance is monotonic, its dependence on the parameterization of the conductivity model, like the f parameter value, is more straight forward than the conductivity or resistivity profiles. Like the electric resistance of a layer corresponds to the electric voltage required to drive a unit current through the layer, the thermal resistance corresponds to the necessary temperature contrast to drive a unit heat flow through the layer. The simple behavior of the thermal resistance in the thermal boundary layers may provide a basis for the development of parameterized convection models including effects of variable conductivity on thermal history. Fig. 15 shows 1D depth profiles of conductivity, resistivity and corresponding thermal resistance through the top and bottom thermal boundary layers for different mantle convection models. Two cases with contrasting contribution of the radiative conductivity are shown. The green curves correspond to a purely phonon conduction case (f = 0) and the blue curves are for strongly enhanced radiative conductivity (f = 10). The black curves indicate the constant conductivity reference case. The trend in the models for enhanced radiative conductivity is clearly reflected in the monotonic resistance profiles. In the top thermal boundary layer, top frames, the contribution from the radiative conductivity controls the resistance profile (c) where the constant conductivity case is intermediate between the contrasting variable conductivity models. This reflects the trend in the cooling curves for the mantle shown in Fig. 5, which explains the strong impact of krad in speeding up secular cooling, through its influence on the low conductivity zone and the resulting thermal resistance of the lithosphere. In the bottom boundary layer, bottom frames, a similar relation exists except that the constant conductivity case has the highest resistance values, corresponding to the minimum value of the conductivity shown in frame (d). 4. Discussion and conclusions We have developed a core–mantle coupling convection model within the framework of a cartesian 2D geometry. This model has many realistic transport proper- ties built in, such as variable thermal conductivity and variable viscosity. It does not have surface plates, phase transitions and chemical heterogeneities. But, nonetheless, this study will shed some light on the nature of the thermo-mechanical structure in the deep lower mantle. 4.1. Summary of important findings Variable thermal conductivity affects both conductive and convective cooling mechanisms in the mantle. Introducing pressure- and temperature-dependent thermal conductivity along with temperature- and pressuredependent viscosity into the mantle convection model results in several important changes in the cooling behavior and mantle flow patterns: 1. The secular cooling rate of the mantle is lower, using the Hofmeister conductivity model (Hofmeister, 1999) (f = 1) than with constant thermal conductivity. Heat flux at the surface is reduced. Increased f (greater radiative contribution to thermal conductivity) tends to increase the cooling rate relative to the Hofmeister model. This thermal conductivity mechanism, acting in concert with partial melting (Korenaga, 2003) can help to retain a lot of the primordial heat of the Earth. Therefore, variable thermal conductivity can keep the Urey ratio low, which is consistent with highly depleted heat-producing elements in the mantle (Jochum et al., 1983), favored by geochemists. 2. We have shown that the cooling delay of the variable conductivity models is closely linked to the formation of a low conductivity zone (LCZ) at shallow depth. This LCZ results from the negative temperature derivative of the dominant lattice conductivity. These results imply that purely pressuredependent conductivity models, characterized by a monotonic increase of the conductivity with depth are not suitable for long-term thermal history calculation (Anderson, 1987; Steinbach, 1991; Solheim and Peltier, 1994; Tackley, 1996; van Keken, 2001; Butler and Peltier, 2002). 3. Our model results show a high fluctuation level of the heat flow from the core into the mantle, with higher fluctuations for models with enhanced radiative conductivity. We speculate that such fluctuations could leave an imprint in the paleomagnetic A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 4. 5. 6. 7. 8. intensity record, through a possible influence on the geodynamo. A higher value of the initial CMB temperature, TCMB (0), leads to a more stable boundary layer due to the increase in radiative conductivity with temperature. It is important to note that core–mantle boundary temperature is also important for boundary layer stability (Sevre et al., 2002). A higher initial core–mantle boundary temperature leads to faster secular cooling of the mantle and faster core cooling. The local maximum of viscosity (viscosity hill) in the lower mantle and the region of low viscosity below the hill also change with variable k. The viscosity hill is smaller for variable k compared with constant k. Increased TCMB (0) (or decreased viscosity) also leads to a decrease in the size of the viscosity hill in the lower mantle and at the same time maintains a shallower viscosity gradient in the lower mantle. There is a greater predominance of smaller convective upwellings for variable k, especially in cases with a high initial TCMB (0). In this model, large plumes are less prevalent, and the lower mantle may have less capacity to create superplumes. The geotherm in the deep mantle becomes superadiabatic (da Silva et al., 2000) with enhanced values of radiative thermal conductivity, which controls the magnitude of the thermal gradient in the lower mantle (Yamazaki and Karato, 2001). Recent suggestion by Badro et al. (2004) has added support to the importance of radiative heat transfer in the deep mantle. These results clearly demonstrate that temperatureand pressure-dependent thermal conductivity, though small in magnitude, can not be neglected in mantle dynamics and planetary thermal evolution, through the combined nonlinear feedback interactions among thermal conductivity, temperature and viscosity. 4.2. Applications to the core We have found that with core coupling the heat flux at the CMB depends on both conductivity in the lower boundary layer and the evolving temperature difference across the lower boundary layer, both of which are significantly impacted by variable thermal conduc- 275 tivity. The heat flux at the core–mantle boundary shows large fluctuations and is reduced for variable thermal conductivity with a normal amount of radiative conductivity (f = 1). These results would suggest that the heat flux out of the core may actually be less than that predicted by models with constant conductivity. If less heat is conducted from the core into the mantle, this would delay draining the core’s internal energy supply. This would decrease the minimum necessary concentration of radioactive elements such as 40 K in the core. However, the CMB heat flux with variable conductivity is strongly dependent on initial CMB temperature and on the relative strength of radiative conductivity in the lower mantle, which is represented in this model by the parameter f (e.g. Dubuffet et al., 2002). These parameters are not known with enough certainty to provide tight enough constraints on the CMB heat flux. 4.3. The role in 1D parameterized convection calculations Because of the nonlinearities in mantle rheology, mantle thermal conductivity and mantle convection, thermal evolution with convection is definitely influenced by its initial temperature condition for a period usually called the thermal adjustment time (Solomatov, 2001). After the thermal adjustment time, thermal histories calculated from different initial conditions should converge. Early thermal evolution models based on parameterized convection, found a thermal adjustment time of less than 1 Gyr (e.g. Schubert et al., 1979); however, when more complex rheology is taken into account, the thermal adjustment time for parameterized models becomes much longer (see discussion by Solomatov, 2001). Our 2D convection results for different values of initial CMB temperature indicate that the initial thermal state affects mantle thermal evolution for a period of time greater than the age of the Earth. After around 5 Gyr, there is still a significant difference in interior temperature, core temperature, and thermal profiles for cases with different initial core CMB temperature. This implies that for a fully convective system with strongly temperature- and pressure-dependent viscosity and variable thermal conductivity, the initial thermal state is an important parameter to be considered, in modeling thermal evolution and this cannot be neglected, as in early hypotheses of fast thermal adjust- 276 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 ment with only temperature-dependent viscosity (e.g., Tozer, 1972). 4.4. Limitations of the model and perspectives Since small-scale upwellings appear to be important convective cooling mechanisms for the model with variable conductivity, especially in cases with a high initial TCMB (0), it is important to use a grid with high enough resolution to accurately represent small-scale features. A major limitation in the applicability of the results discussed to the real Earth is a result of the two-dimensional rectangular geometry of the model domain. For a spherical domain, the ratio of interior volume to outer surface area is larger than in the rectangular case. This means that in a spherical domain there will be a reduction in interior temperature (Zhang and Yuen, 1996). Core–mantle dynamics are also influenced by the particular geometry. In a rectangular domain, the size of the core- mantle boundary interface is equal to the size of the mantle-surface interface. These are not equal in a domain with curved geometry: for cylindrical or spherical geometry, the size ratio of the core–mantle boundary to the mantle-surface boundary is less than one. This will impact the coupled cooling of the mantle and core system (van Keken, 2001). Jarvis (1993) investigated the effects of using a curved domain in mantle models and found that the relative thickness of upper and lower boundary layers and the temperature drop across the boundary layers depended on the degree of curvature. For the curvature of the Earth’s mantle boundaries, the interior temperature is also lower by several hundred degrees (Jarvis, 1993). Thus, to apply the results of numerical convection models to the Earth with a reasonable degree of accuracy, it is preferable to use a 3D spherical convection model (e.g. Tackley et al., 1993; Monnereau and Yuen, 2002) but with variable properties built-in. But this will be a computational challenge because of the high spatial resolution involved, with at least 108 grid points. In future work the role of the new post-perovskite phase change near the bottom of the mantle (Murakami et al., 2004) needs to be explored since large changes in the physical properties, including thermal conductivity (Badro et al., 2004), may be expected (Tsuchiya et al., 2004). Finally, our results point to the potential role played by complex transport properties, such as variable ther- mal conductivity and grain-size dependence of material properties (Solomatov, 1996, 2001; Hofmeister, 2004), on making the thermal evolution of the mantle more dependent on the initial condition, because of the slower secular cooling, which induce other modes of planetary evolution (Sleep, 2000). Acknowledgments We acknowledge thorough and constructive reviews by Paul Tackley and Thorsten Becker which greatly helped to improve the manuscript. We thank discussions with Tomo K.B. Yanagawa, Anne M. Hofmeister, Fabien W. Dubuffet, Marc Monnereau, Renata M. Wentzcovitch, Slava Solomatov and Erik O.D. Sevre. This work has been supported by the geophysics program of the National Science Foundation and the Dutch NWO. References Anderson, D.L., 1987. A seismic equation of state II, shear properties and thermodynamics of the lower mantle. Phys. Earth Planet. Inter. 45, 307–323. Badro, J., Rueff, J.-P., Vanko, G., Monaco, G., Fiquet, G., Guyot, F., 2004. Electronic transitions in perovskite: possible nonconvecting layers in the lower mantle. Science 305, 383– 386. Beardsmore, G.R., Cull, J.P., 2001. Crustal Heat Flow: A Guide to Measurement and Modeling. Cambridge University Press. Buffett, B.A., Huppert, H.E., Lister, J.R., Woods, A.W., 1992. Analytical model for solidification of the Earth’s core. Nature 356, 329–331. Buffett, B.A., 2003. The thermal state of the Earth’s core. Science 299, 789–791. Bullard, E.C., 1939. Heat flow in South Africa. Proc. Roy. Soc. Lond. A 173, 474–502. Butler, S.L., Peltier, W.R., 2002. The thermal evolution of the earth: models with time dependent layering of mantle convection which satisfy the Urey ratio constraint. J. Geophys. Res. 107, B6, doi:10.1029/2000JB000018. Christensen, U.R., 1985. Thermal evolution models for the Earth. J. Geophys. Res. 90, 2995–3008. da Silva, C.R.S., Wentzcovitch, R.M., Patel, A., Price, G.D., Karato, S.I., 2000. The composition and geotherm of the lower mantle: constraints from the elasticity of silicate perovskite. Phys. Earth Planet. Inter. 118, 103–109. DeLandro-Clarke, W., Jarvis, G.T., 1997. Numerical models of mantle convection with secular cooling. Geophys. J. Int. 129, 183– 193. A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 Dubuffet, F., Yuen, D.A., Rabinowicz, M., 1999. Effects of a realistic mantle thermal conductivity on the patterns of 3D convection. Earth Planet. Sci. Lett. 171 (3), 401–409. Dubuffet, F., Yuen, D.A., Rainey, E.S.G., 2002. Controlling thermal chaos in the mantle by positive feedback from radiative thermal conductivity. Nonlinear Process. Geophys. 9 (3–4), 311–323. Gessman, C.K., Wood, B.J., 2002. Potassium in the Earth’s core? Earth Planet. Sci. Lett. 200, 63–78. Hall, H.T., Murthy, V.R., 1971. The early chemical history of the Earth: some critical elemental fractionations. Earth Planet. Sci. Lett. 11, 239–244. Hofmeister, A.M., 1999. Mantle values of thermal conductivity and the geotherm from phonon lifetimes. Science 283, 1699– 1706. Hofmeister, A.M., 2004. The dependence of diffusive transfer on grain-size, temperature, and Fe-content: implications for mantle processes. J. Geodynamics, submitted for publication. Honda, S., Yuen, D.A., 1994. Cooling model of mantle convection with phase changes: effects of aspect ratio and initial conditions. J. Phys. Earth 42, 165–186. Jarvis, G.T., 1993. Effects of curvature on two-dimensional models of mantle convection: cylindrical polar coordinates. J. Geophys. Res. 98, 4477–4485. Jochum, K.P., Hofmann, A.W., Ito, E., Seufert, M., White, W.M., 1983. K, U and Th in mid-ocean ridge basalt glasses and heat production K/U and K/Rb in the mantle. Nature 306, 431– 436. Kellogg, L.H., Hager, B.H., van der Hilst, R.D., 1999. Compositional stratification in the deep mantle: towards a hybrid convection model. Science 283, 1881–1884. Korenaga, J., 2003. Energetics of mantle convection and the fate of fossil heat. Geophys. Res. Lett. 30 (8), doi:10.129/ 2003GL016982. Manga, M., Jeanloz, R., 1996. Implications of a metal-bearing chemical boundary layer in D"" for mantle dynamics. Geophys. Res. Lett. 23 (22), 3091–3094. Monnereau, M., Yuen, D.A., 2002. How flat is the lower-mantle temperature gradient? Earth Planet. Sci. Lett. 207, 171–183. Murakami, M., Hirose, K., Kawamura, K., Sata, N., Ohishi, Y., 2004. Post-perovskite phase transition in MgSiO3 . Science 304, 855– 858. Murthy, V.R., van Westrenen, W., Fei, Y., 2003. Experimental evidence that potassium is a substantial radioactive heat source in planetary cores. Nature 423, 163–165. Schubert, G., Cassen, P., Young, R.E., 1979. Subsolidus convective cooling histories of terrestrial planets. Icarus 38, 192– 211. Sevre, E.O.D., Dubuffet, F., Yuen, D.A., Rainey, E.S.G., 2002. A stabilizing dynamical influence in the deep mantle due to the radiative thermal conductivity and a high temperature at the core– mantle boundary. Electron. Geosci. 7. http://link.springer.de/ link/service/journals/10069/free/discussion/sevre/index.html. Sharpe, H.N., Peltier, W.R., 1978. Parameterized mantle convection and the Earth’s thermal history. Geophys. Res. Lett. 5, 737– 740. Sleep, N.H., 2000. Evolution of the mode of convection within terrestrial planets. J. Geophys. Res. 105, 17,563–17,578. 277 Solheim, L., Peltier, W.R., 1994. Avalanche effects in phase transition modulated thermal convection: a model of the Earth’s mantle. J. Geophys. Res. 99, 6997–7018. Solomatov, V.S., Zharkov, V.N., 1990. The thermal regime of Venus. Icarus 84, 280–295. Solomatov, V.S., 1996. Can hotter mantle have a larger viscosity? Geophys. Res. Lett. 23, 937–940. Solomatov, V.S., 2001. Grain size-dependent viscosity convection and the thermal evolution of the Earth. Earth Planet. Sci. Lett. 191, 203–212. Steinbach, V., Hansen, U., Ebel, A., 1989. Compressible convection in the earth’s mantle: a comparison of different approaches. Geophys. Res. Lett. 16, 633–635. Steinbach, V., 1991. Numerische Experimente zu Konvektion in kompressiblen Medien. Ph.D. Thesis. Universität zu Köln. Steinbach, V., Yuen, D.A., Zhao, W., 1993. Instabilities from phase transitions and the timescales of mantle evolution. Geophys. Res. Lett. 20, 1119–1122. Tackley, P.J., Stevenson, D.J., Glatzmaier, G.A., Schubert, G., 1993. Effects of an endothermic phase transition at 670 km depth on spherical mantle convection. Nature 361, 699–704. Tackley, P.J., 1996. Effects of strongly variable viscosity on threedimensional compressible convection in planetary mantles. J. Geophys. Res. 101, 3311–3332. Tozer, D.C., 1972. The present thermal state of the terrestrial planets. Phys. Earth Planet. Inter. 6, 182–197. Tsuchiya, T., Tsuchiya, J., Umemoto, K., Wentzcovitch, R.M., 2004. Phase transition in MgSiO3 perovskite in the earths lower mantle. Earth Planet. Sci. Lett. 224, 241–248. van den Berg, A.P., van Keken, P.E., Yuen, D.A., 1993. The effects of a composite non-Newtonian and Newtonian rheology in mantle convection. Geophys. J. Int. 115, 62–78. van den Berg, A.P., Yuen, D.A., 1998. Modelling planetary dynamics by using the temperature at the core–mantle boundary as a control variable: effects of rheological layering on mantle heat transport. Phys. Earth Planet. Inter. 108, 219–234. van den Berg, A.P., Yuen, D.A., Steinbach, V., 2001. The effects of variable thermal conductivity on mantle heat-transfer. Geophys. Res. Lett. 28 (5), 875–878. van den Berg, A.P., Yuen, D.A., Allwardt, J.R., 2002. Nonlinear effects from variable thermal conductivity and mantle internal heating: implications for massive melting and secular cooling of the mantle. Phys. Earth Planet. Inter. 129, 359–375. van den Berg, A.P., Yuen, D.A., 2002. Delayed cooling of the Earth’s mantle due to variable thermal conductivity and the formation of a low conductivity zone. Earth Planet. Sci. Lett. 199, 403– 413. van den Berg, A.P., Yuen, D.A., Rainey, E.S.G., 2004. The influence of variable viscosity on delayed cooling due to variable thermal conductivity. Phys. Earth Planet. Inter. 142, 283–295. van Keken, P.E., 2001. Cylindrical scaling for dynamical cooling models of the Earth. Phys. Earth Planet. Inter. 124, 119–130. Yamazaki, D., Karato, S.-I., 2001. Some mineral physics constraint on the rheology and geothermal structure of Earth’s lower mantle. Am. Mineralogist 86, 385–391. Yuen, D.A., Reuteler, D.M., Balachandar, S., Steinbach, V., Malevsky, A.V., Smedsmo, J.L., 1994. Various influences on 278 A.P. van den Berg et al. / Physics of the Earth and Planetary Interiors 149 (2005) 259–278 three-dimensional mantle convection with phase transitions. Phys. Earth Planet Inter. 86, 185–203. Yuen, D.A., Vincent, A.P., Bergeron, S.Y., Dubuffet, F., Ten, A.A., Steinbach, V.C., Starin, L., 2000. Crossing of scales and nonlinearities in geophysical processes. In: Boschi, E., Ekstrom, G., Morelli, A. (Eds.), Problems for the New Millennium. Editrice Compositori, Bologna, Italy, pp. 403–463. Zhang, S., Yuen, D.A., 1996. Various influences on plumes and dynamics in time-dependent, compressible, mantle convection in 3D spherical shell. Phys. Earth Planet. Inter. 94, 241–267.
© Copyright 2026 Paperzz