Hostname: page-component-7b7567c8c-z4zkx Total loading time: 0 Render date: 2026-11-16T10:10:52.204Z Has data issue: false hasContentIssue false

The effects of a Robin boundary condition on thermal convection in a rotating spherical shell

Published online by Cambridge University Press:  17 May 2021

Thibaut T. Clarté*
Affiliation:
Univ. Grenoble Alpes, Univ. Savoie Mont Blanc, CNRS, IRD, IFSTTAR, ISTerre, 38000 Grenoble, FranceUniversité de Lyon, ENSL, UCBL, Laboratoire LGLTPE, 15 parvis René Descartes, BP7000, 69342 Lyon, CEDEX 07, France
Nathanaël Schaeffer
Affiliation:
Univ. Grenoble Alpes, Univ. Savoie Mont Blanc, CNRS, IRD, IFSTTAR, ISTerre, 38000 Grenoble, France
Stéphane Labrosse
Affiliation:
Université de Lyon, ENSL, UCBL, Laboratoire LGLTPE, 15 parvis René Descartes, BP7000, 69342 Lyon, CEDEX 07, France
Jérémie Vidal
Affiliation:
Univ. Grenoble Alpes, Univ. Savoie Mont Blanc, CNRS, IRD, IFSTTAR, ISTerre, 38000 Grenoble, France
*
†Email address for correspondence: thibaut.clarte@ens-lyon.fr

Abstract

Convection in a spherical shell is widely used to model fluid layers of planets and stars. The choice of thermal boundary conditions in such models is not always straightforward. To understand the implications of this choice, we report on the effects of the thermal boundary condition on thermal convection, in terms of instability onset, fully developed transport properties and flow structure. We impose a Robin boundary condition, enforcing linear coupling between the temperature anomaly and its radial derivative, with the Biot number ${{\textit {Bi}}}$ as a proportionality factor in non-dimensional form. Varying ${{\textit {Bi}}}$ allows us to transition from fixed temperature for ${{\textit {Bi}}}=+\infty$, to imposed heat flux for ${{\textit {Bi}}}=0$. We find that the onset of convection is only affected by ${{\textit {Bi}}}$ in the non-rotating case. Far from onset, considering an effective Rayleigh number and a generalized Nusselt number, we show that the Nusselt and Péclet numbers follow standard universal scaling laws, independent of ${{\textit {Bi}}}$ in all cases considered. However, for the non-rotating limit, the large-scale flow structure keeps the signature of the boundary condition with more vigorous large scales for smaller ${{\textit {Bi}}}$, even though the global heat transfer and kinetic energy are the same. For all practical purposes, the Robin condition can be safely replaced by a fixed flux when ${{\textit {Bi}}} \lesssim 0.03$ and by a fixed temperature for ${{\textit {Bi}}}\gtrsim 30$. For turbulent rapidly rotating convection, the thermal boundary condition does not seem to have any impact, once the effective numbers are considered and a reference temperature profile has been chosen.

Information

Type
JFM Papers
Copyright
© The Author(s), 2021. Published by Cambridge University Press
Figure 0

Table 1.Dimensionless numbers characterising a terrestrial global magma ocean, maximal and minimal values are given, computed from Solomatov (2007), Maas & Hansen (2015) and Snyder, Gier & Carmichael (1994) (for typical values of the thermal conductivity). The extremal values for each parameter cover the full range between a fully liquid to a quasi-solid (because of a high proportion of crystals in suspension) magma ocean. Finally, the range of parameters used in this study is precised.

Figure 1

Figure 1.Three-dimensional visualization of radial velocity (a) and temperature anomaly (b) of a nonlinear computation. Simulation run for ${{\textit {Ra}}}=9\times 10^{6}$, ${{\textit {Pr}}}=3$, ${{\textit {Ek}}}=10^{-4}$ and ${{\textit {Bi}}}=0$. Internal and external surfaces have been defined slightly outside boundary layers.

Figure 2

Figure 2.Critical Rayleigh number ${{\textit {Ra}}}_C$ of FF and FT cases as a function of Ekman number ${{\textit {Ek}}}$ for Prandtl number values ${{\textit {Pr}}}=0.3$, $3$ and $30$. For evanescent ${{\textit {Ek}}}$, ${{\textit {Ra}}}_C$ scales as ${{\textit {Ek}}}^{-4/3}$ (represented by the dotted line). The inset is a compensated plot of the same data points.

Figure 3

Figure 3.Values of the critical azimuthal wavenumber $m$ as a function of ${{\textit {Ek}}}$ for both FF (${{\textit {Bi}}}=0$) and FT (${{\textit {Bi}}}=\infty$) cases. For small enough values of ${{\textit {Ek}}}$, $m$ scales as ${{\textit {Ek}}}^{-1/3}$ (represented by the dashed line). The inset is a compensated plot of the same data points.

Figure 4

Figure 4.Plot of $\varXi$ (i.e. difference of ${{\textit {Ra}}}_C$ between FF and FT cases for a given ${{\textit {Ek}}}$ normalised by the non-rotating difference (cf. (3.1))) as a function of ${{\textit {Ek}}}$. Here $\varXi$ decreases with the same general behaviour independently from ${{\textit {Pr}}}$.

Figure 5

Figure 5.Plot of $\widetilde {{{\textit {Ra}}}_C}$ (cf. (3.2)) as a function of ${{\textit {Bi}}}$ for ${{\textit {Pr}}}=3$. The amplitude of the continuous transition from FF to FT is decreased when ${{\textit {Ek}}}$ is reduced. By definition, $\widetilde {{{\textit {Ra}}}_C} \rightarrow 1$ when ${{\textit {Bi}}} \rightarrow \infty$ (i.e. when the boundary condition goes to FT). The axis of abscissas uses a hybrid linear-logarithmic scale.

Figure 6

Figure 6.Plot of $\varGamma = ({{{\textit {Ra}}}_C({{\textit {Bi}}})-{{\textit {Ra}}}_C({{\textit {Bi}}}=0)})/({{{\textit {Ra}}}_C({{\textit {Bi}}}=\infty )-{{\textit {Ra}}}_C({{\textit {Bi}}}=0)})$ as a function of ${{\textit {Bi}}}$ for various velocity boundary conditions at the top of the sphere and values of ${{\textit {Ek}}}$, ${{\textit {Pr}}}=3$. It illustrates the transition of ${{\textit {Ra}}}_C$ from FF values ($\varGamma = 0$) to FT ($\varGamma = 1$) values. If a mechanical boundary condition is mentioned, this one is imposed at both boundaries, if not, the usual set-up (free slip at the top, no slip at the bottom) is used; see § 2.1. The axis of abscissas uses a hybrid linear-logarithmic scale.

Figure 7

Figure 7.Phase diagram of the solution structure at the linear onset of convection as a function of ${{\textit {Ek}}}$ and ${{\textit {Bi}}}$, typical structures of temperature anomalies and dependence on ${{\textit {Bi}}}$ or ${{\textit {Ek}}}$ is mentioned for each area. (I) Fixed temperature solution independent of rotation with a constant value of $m$, (II) FF solution independent of rotation with a constant value of $m$, (III) FT solution affected by rotation ($m$ increases with rotation), (IV) intermediate structure, (V) FF solution affected by rotation, (VI) common structure at high rotation rate.

Figure 8

Figure 8.Plot of $\varGamma = ({{{\textit {Ra}}}_C({{\textit {Bi}}})-{{\textit {Ra}}}_C({{\textit {Bi}}}=0)})/({{{\textit {Ra}}}_C({{\textit {Bi}}}=\infty )-{{\textit {Ra}}}_C({{\textit {Bi}}}=0)})$ as a function of ${{\textit {Bi}}}$ with variable ${{\textit {Pr}}}$ and ${{\textit {Ek}}}=+\infty$. Here $\varGamma$ goes to 0 (respectively 1) when the boundary condition goes to FF, i.e. ${{\textit {Bi}}}=0$ (respectively FT, i.e. ${{\textit {Bi}}}\rightarrow \infty$). The axis of abscissas uses a hybrid linear-logarithmic scale.

Figure 9

Figure 9.(a) Schematic of the tangent cylinder (grey) as part of a spherical shell rotating about an axis. (b) Quarter of meridional slices representing the radial cylindrical component of the velocity, $u_{s}$, at the onset of convection (linear mode, arbitrary amplitude) for FF (iii,iv) and FT (i,ii) top boundary conditions and ${{\textit {Pr}}}=0.3$ (ii,iv) or $0.03$ (i,iii), ${{\textit {Ek}}}=10^{-4}$. We observe an effect of ${{\textit {Pr}}}$ on the flow, which appears outside of the tangent cylinder for ${{\textit {Pr}}}=0.3$ and inside it for ${{\textit {Pr}}}=0.03$.

Figure 10

Figure 10.Nusselt number, ${{\textit {Nu}}}$, as a function of ${{\textit {Ra}}}$ for ${{\textit {Pr}}}=0.3$ (open symbols) or $3$ (filled symbols) with (${{\textit {Ek}}}=10^{-4}$: squares and ${{\textit {Ek}}}=10^{-2}$: circles) or without (${{\textit {Ek}}}=\infty$ circles) rotation and with FF (red points) or FT (blue ones) at the upper boundary. The black dotted line represents the scaling ${{\textit {Nu}}} \propto {{\textit {Ra}}}^{2/7}$ relevant to model FT configurations while the ${{\textit {Nu}}} \propto {{\textit {Ra}}}^{2/9}$ – green dotted line – models the FF ones.

Figure 11

Figure 11.Nusselt number, ${{\textit {Nu}}}$, as a function of ${{\textit {Ra}}}_{eff}$ (4.2) for the same data as in figure 10. For a given ${{\textit {Pr}}}$, FF and FT points follow the same behaviour except near the threshold for non-rotating convection. Here ${{\textit {Nu}}}\propto {{\textit {Ra}}}^{2/7}$ is plotted with a blue dotted line.

Figure 12

Figure 12.Péclet number ${{\textit {Pe}}}$ as a function of ${{\textit {Ra}}}_{eff}$ for the same data as plotted in figure 11. Except for ${{\textit {Ra}}}_{eff} \simeq {{\textit {Ra}}}_C$, FF and FT configurations follow the same behaviour at a given ${{\textit {Ek}}}, {{\textit {Pr}}}$. For large enough ${{\textit {Ra}}}_{eff}$, ${{\textit {Pe}}}$ becomes independent of ${{\textit {Ek}}}$. Solid and broken lines are $1/2$ slope scalings for ${{\textit {Pr}}}=3$ and ${{\textit {Pr}}}=0.3$.

Figure 13

Figure 13.Non-zonal component of the Péclet number, ${{\textit {Pe}}}_{NZ}$, as a function of ${{\textit {Ra}}}_{eff}$ for ${{\textit {Ek}}}=10^{-4}$ cases. Similar behaviour is observed regardless of $({{\textit {Pr}}},{{\textit {Bi}}})$: ${{\textit {Pe}}}_{NZ}$ values approaches the non-rotating limit when ${{\textit {Ra}}}_{eff}$ is increased. The overshoot observed in figure 12 is suppressed with the subtraction of the zonal component.

Figure 14

Figure 14.Radial temperature profiles averaged in time and normalised by the mean temperature difference across the shell, $({\bar {T}-T_{min}})/({T_{max}-T_{min}})$. Reference, FF and FT profiles are compared, ${{\textit {Ek}}}=\infty ,{{\textit {Pr}}}=3,{{\textit {Ra}}}_{eff} \approx 5.6\times 10^{6}$. For large enough ${{\textit {Ra}}}_{eff}$, equal ${{\textit {Ra}}}_{eff}$ configurations show similar profiles independently of the choice of boundary condition.

Figure 15

Figure 15.Meridional (a,b) and equatorial (c,d) maps of radial velocity at ${{\textit {Ek}}}=10^{-4}$, ${{\textit {Pr}}}=3$ and ${{\textit {Bi}}}=\infty$; (a,c) ${{\textit {Ra}}}=9\times 10^{6}$, (b,d) ${{\textit {Ra}}}=9\times 10^{7}$. (a) The flow is only active outside the tangent cylinder, (b) the flow is active in the whole shell, (c,d) the flow loses its lateral organization when ${{\textit {Ra}}}$ is increased by a factor 10.

Figure 16

Figure 16.Equatorial maps of temperature anomalies ($l=0$ component is subtracted, ${{\textit {Pr}}}=3$ for all cases); (a,c,e,g) ${{\textit {Bi}}}=0$ (FF), (b,d,f,h) ${{\textit {Bi}}}=\infty$ (FT); (c,d,g,h) rotating cases (${{\textit {Ek}}}=10^{-4}$), (a,b,e,f) non-rotating ones; (a,b,c,d) ${{\textit {Ra}}}/ {{\textit {Ra}}}_{C} \simeq 1.1$, (e,f,g,h) ${{\textit {Ra}}}_{eff}=3.10^{8}$. Except for differences in the thermal upper boundary layer, structures of FF and FT thermal anomalies are similar for each line.

Figure 17

Figure 17.Nusselt number ${{\textit {Nu}}}$ as a function of ${{\textit {Ra}}}_{eff}$ for the same representation code as in figure 10. Intermediate ${{\textit {Bi}}}$ cases have been added to the figure 10 data. They are plotted with big symbols, coloured according to their ${{\textit {Bi}}}$ value. As in figure 10, points are distributed along two main branches determined by the value of ${{\textit {Ek}}}$. Except for (${{\textit {Ek}}}= \infty$, ${{\textit {Ra}}} \approx {{\textit {Ra}}}_{C}$) points, intermediate ${{\textit {Bi}}}$ points are located on the same branches as extremal ones and then follow the same scalings as described in figure 10.

Figure 18

Figure 18.Normalised and time-averaged radial temperature profiles for quasi constant ${{\textit {Ra}}}_{eff} \simeq 2.2 \times 10^{6}$ but variable $({{\textit {Ra}}},{{\textit {Bi}}})$, (${{\textit {Ek}}}=\infty ,{{\textit {Pr}}}=0.3$). Similar profiles are obtained, nevertheless, three different behaviours can be distinguished for large (FT), small (FF) and intermediate ${{\textit {Bi}}}$ values.

Figure 19

Figure 19.Time-averaged energy spectrum of the velocity (sum of toroidal and poloidal components over the whole shell) for roughly constant ${{\textit {Ra}}}_{eff}\simeq 2.2 \times 10^{6}$, ${{\textit {Ek}}}=\infty$, ${{\textit {Pr}}}=3$ but variable ${{\textit {Bi}}}$. For ${{\textit {Bi}}}\le 0.1$, $l=1$ is the dominant mode, intermediate values ($1/3\le {{\textit {Bi}}} < 10/3$) favour $l=2$, and $l=3$ dominates at high ${{\textit {Bi}}}$ as in the FT configuration. The total kinetic energy in all these configurations varies by less than 10$\%$.

Figure 20

Figure 20.Non-zonal component of ${{\textit {Pe}}}$ as a function of ${{\textit {Nu}}}-1$ for various ${{\textit {Ek}}},{{\textit {Pr}}},{{\textit {Bi}}}$ coded by shape, filling and colour of the symbols, respectively (as in figures 10 and 17). We distinguish two main regimes: the weakly nonlinear regimes at low ${{\textit {Pe}}}$ and ${{\textit {Nu}}}\simeq 1$ following the two broken-line models and the weakly or non-rotating regimes represented by the dotted line.

Figure 21

Figure 21.Kinetic energy ratio ${E}/{E_{NZ}}$ as a function of the convective Rossby number $R_{oc}$ for various (${{\textit {Ek}}},{{\textit {Pr}}},{{\textit {Bi}}}$). Legend is similar to figure 22. Zonal flows are mainly active for small ${{\textit {Ek}}}$ and $R_{oc}\in [0.1,1]$.

Figure 22

Figure 22.Nusselt Number as a function of the Biot Number for various $({{\textit {Ra}}},{{\textit {Ek}}},{{\textit {Pr}}}=3)$ simulations. Each data subset defined by $({{\textit {Ra}}},{{\textit {Ek}}})$ is plotted by symbols of the same colour and shape. Here FF (${{\textit {Bi}}}=0$) and FT (${{\textit {Bi}}}\rightarrow \infty$) limits are represented respectively by solid and dotted lines of the same colour.

Figure 23

Figure 23.Plot of $\varSigma$ transition parameter (cf. (B1), computed with $X={{\textit {Nu}}}-1$) from FF (${{\textit {Bi}}}=0$) to FT (${{\textit {Bi}}}=\infty$) as a function of the Biot Number for various $({{\textit {Ra}}},{{\textit {Ek}}},{{\textit {Pr}}}=3)$ simulations. Here $\varSigma _{0}$ trends (cf. (B2)) are plotted for three transition values: ${{\textit {Bi}}}_{T} =1,\,5/3,\,10/3$. Some cases show a large deviation from the trends, which is explained in the main body of this appendix.

Figure 24

Figure 24.Three-dimensional representation of the temperature anomaly ($l=0$ subtracted) inside a shell of radial limits 0.7 and 0.95 with ${{\textit {Pr}}}=3$. Rotational axis is $z$. (a,c,e) FF, (b,d,f) FT. For each line, (${{\textit {Ra}}}_{eff},{{\textit {Ek}}}$) is kept approximately constant; (a,b) ($Ra_{eff} =3\times 10^{8},Ek =\infty$). (c,d) ($Ra_{eff} =3\times 10^{8}, Ek = 1 \times 10^{-4}$). (e,f) ($Ra_{eff} = 1 \times 10^{7}, Ek = 1 \times 10^{-4}$).