The Astrophysical Journal, 611:568–574, 2004 August 10 # 2004. The American Astronomical Society. All rights reserved. Printed in U.S.A. LIMITS TO PENETRATION OF MERIDIONAL CIRCULATION BELOW THE SOLAR CONVECTION ZONE Peter A. Gilman and Mark S. Miesch High Altitude Observatory, National Center for Atmospheric Research,1 Boulder, CO 80307-3000; [email protected], [email protected] Received 2004 February 4; accepted 2004 April 14 ABSTRACT We show that meridional circulation, such as that observed at the top of and in the solar convection zone (CZ) by direct doppler and helioseismic techniques, cannot penetrate significantly below the bottom (0.7 R) of the overshoot layer at the bottom of the CZ. Therefore, solar dynamo models that rely on penetration as deep as to 0.6 R are ruled out. The analysis we carried out to reach this conclusion elucidates two boundary layers, one of which we have not seen applied to astrophysical problems before. This analysis should be relevant to understanding interfaces between convective and radiative zones in stellar interiors generally. Subject headingg s: hydrodynamics — stars: interiors — stars: rotation — Sun: interior — Sun: rotation 1. INTRODUCTION The existence of meridional circulation in the solar convection zone has recently been demonstrated to be very important for understanding the solar dynamo. The most successful current solar dynamo models, the so-called flux transport models (Dikpati & Charbonneau 1999; Dikpati & Gilman 2001; Dikpati et al. 2002), depend on the presence of poleward flow in the upper part of the convection zone, coupled with weaker equatorward flow near the bottom, to produce a number of dynamo features that compare well to properties of the solar cycle. Perhaps the most significant feature is the solar cycle period, which has been shown to be produced by equatorward flow near the bottom of the convection zone, and which is consistent with the observed 20 m s1 poleward flow in the photosphere and the upper convection zone (Dikpati & Charbonneau 1999). The period of these dynamos is much more sensitive to the meridional flow speed than it is to assumed magnetic diffusivities or even to the details of the differential rotation profile (Dikpati & Charbonneau 1999). Previous mean-field dynamo models had to rely much more on ‘‘tuning’’ the diffusivity and /or differential rotation to get the correct period. Nandy & Choudhuri (2002) have recently demanded even more of the meridional circulation, namely that it penetrate well below the bottom of the convection zone, down to a depth of perhaps 0.6 R. This requirement is dictated by their particular solutions to the flux-transport dynamo equations, in order to produce the best agreement with observations of the latitudinal distribution of sunspots (the butterfly diagram). There is no agreement among dynamo modelers at this time that such a deep meridional circulation is in fact necessary. We do not pursue that question further here, but instead focus on the question of whether it is physically realistic that such a deep penetration should occur. Since the dynamo models in question are kinematic, the meridional circulation can be simply assumed to reach down this far. But when the more realistic problem is solved, with all the relevant forces included in the equations of motion, even without a magnetic field present, can such penetration be achieved? We show that the answer is no for all reasonable assumptions. There are additional problems with a meridional circulation that penetrates as deep as 0.6 R. Such a flow would transport light elements such as lithium and berylium from the convection zone to deeper layers, where they would be destroyed Recently, both surface doppler and helioseismic measurements have demonstrated the existence of a meridional flow on the Sun (Hathaway 1996; Giles et al. 1997; Braun & Fan 1998; Schou & Bogart 1998; Haber et al. 2002; Basu & Antia 2003). This flow is predominantly toward the poles at all observed latitudes in both hemispheres in and near the photosphere, although since 1996, when this flow has been detected by multiple means, there have been years in which an equatorward flow has been seen in high latitudes in the northern hemisphere (Haber et al. 2002). The magnitude of the poleward flow is 20 m s1, but fluctuates significantly about this value. Since mass is not observed to be piling up near the solar poles over time, there must be a return flow somewhere inside the Sun. Helioseismic measures have so far failed to reveal this return flow, down to a depth of 0.8 R (Braun & Fan 1998). The deeper the flow is, the smaller it needs to be, because of the substantial rise of density through the solar convection zone. In fact, the poleward mass flux seen in the photosphere is tiny compared to that occurring down to the depth of helioseismic measurements, for the same reason. There is currently no satisfactory theory for this meridional flow, although all global convection models for solar differential rotation produce such flows. This is because these models tend to produce much more structured meridional flows, with typically several ‘‘cells’’ between the equator and poles, probably because of the larger influence of rotation on the flow in these models than is apparently occurring on the Sun (Miesch et al. 2000; Elliot et al. 2000; Brun & Toomre 2002). Also, since the meridional flow predicted is inevitably the result of small imbalances among large meridional and radial forces, namely coriolis, pressure gradient, and buoyancy, as well as turbulent stresses, high accuracy is needed to get a reliable prediction. In any case, the driving mechanisms for meridional flows in the solar convection zone are likely to be far stronger than those in the radiative zone, be they of thermal or mechanical origin, or both (see x 5). 1 The National Center for Atmospheric Research is operated by the University Corporation for Atmospheric Research under sponsorship of the National Science Foundation. 568 SOLAR MERIDIONAL CIRCULATION PENETRATION by nuclear burning. Observations of elemental abundances in the solar photosphere and helioseismic sound speed inversions rule out such deep mixing (e.g., Michaud & Charbonneau 1991; Chaboyer et al. 1995; Pinsonneault 1997; Brun et al. 2002). We show that fluid dynamical effects should limit the penetration into the radiative interior to such an extent that there is no conflict with this compositional constraint. In qualitative terms, penetration will be limited to the depth of the so-called overshoot layer immediately below the convection zone, which layer contains at least part of the solar tachocline. We show that there is effectively no penetration into the radiative domain below. If the only effect to consider in estimating the extent of penetration were how far the kinetic energy of radial motion (1 m s1 or less) in the meridional circulation would produce work done against the subadiabatic stratification of the overshoot layer, or below it the radiative layer, the penetration would be extremely short indeed: at most a kilometer in the overshoot layer, and a meter in the radiative layer! Of course, the depth of the overshoot layer itself is reasonably estimated by converting the kinetic energy of downward plumes into such work. But for such plumes the velocities are probably at least 2 orders of magnitude larger, leading to 104 times as much kinetic energy available to do the work associated with penetration. Even so, the overshoot layer thickness estimated by this means is no more than a few percent of the solar radius (Zahn 1991; Stix 2002; Rempel 2004; the latter shows how the thickness varies with penetration model, and lists many earlier references). That might be the end of the argument about the feasibility of penetration to depths of 0.6 R but for the fact that the equatorward meridional flow that must occur at or near the bottom of the convection zone can in principle drag the stably stratified material below along with it, through turbulent transfer of momentum or wave interactions, or perhaps through other mechanisms. This drag can provide an additional form of work to be used to achieve further penetration, since in the confined domain between the equator and the pole some radial motion is inevitably also caused. Therefore, it is necessary to examine a more detailed model for this part of the Sun, with which we solve for the flow below the convection zone that arises in response to an imposed equatorward meridional flow at the interface. This is what we do in the sections that follow. In carrying out this calculation we encounter two boundary layers that have been extensively studied in the geophysical literature, the Ekman layer (e.g., Pedlosky 1987) and a buoyancy layer (Barcilon & Pedlosky 1967; Veronis 1967), although not with the particular boundary layer placement seen here. Our result is therefore of more general fluid dynamical interest than just its application to the Sun. In addition, of course, if the results apply to the Sun they should apply to the bottoms of most, if not all, stellar convective envelopes. 2. MODEL FORMULATION The solar tachocline may be no more than 3% of the solar radius, so it may be accurately treated as a thin shell (Charbonneau et al. 1999). The overshoot layer of the convection zone, partially congruent with the tachocline, is similarly thin. Therefore, to a very good approximation the radius of the shell we consider can be taken to be constant, with the radial coordinate replaced by a local vertical coordinate z. In addition, the shell is thin compared to scale heights of all the 569 Fig. 1.—Schematic showing the geometry of our domain as a rotating, stably stratified Cartesian slab that extends from the equator to the north pole in latitude. A meridional circulation (v; w) is imposed at the upper surface, which enters the domain at high latitudes, moves equatorward, and exits at low latitudes. The latitudinal boundary conditions prohibit the flow from crossing the equator or pole. thermodynamic variables, so it can be treated as a gas in the so-called Boussinesq approximation (Spiegel & Veronis 1960; Lantz & Fan 1999; Cally 2003). The meridional circulation flowing equatorward is almost certainly of very large latitudinal scale compared to the thickness of the shell, so this class of motion can be taken to be in hydrostatic balance in the radial direction. While in the Sun this shell contains both molecular and turbulent viscous and thermal and radiative diffusion, we limit consideration to a fluid of constant viscous and thermal diffusivities and respectively, which is sufficient to do plausible boundary layer physics. Since the vertical scale of the motion and thermodynamic variables is very small compared to the horizontal scale, only diffusion in the vertical direction need be included. The real Sun is also spherical, but we first consider instead a shallow channel of fluid in Cartesian geometry, periodic or infinitely long in the x direction (corresponding to solar longitude) but bounded in y (corresponding to solar latitude) by lateral, stress-free thermally insulating walls, meant to correspond to the pole and the equator of the shell. The bottom of the channel is rigid and stress free (but may be taken to great depth compared to the boundary layers we find), while the top boundary is allowed to exchange momentum and thermal energy as well as mass with the fluid below. The Cartesian geometry allows us to highlight the relevant boundary layer thicknesses, which also apply to the spherical case, as we show in x 4. With the forcing of equatorward ( y) flow at the top, which is assumed to be zero at both lateral boundaries, latitudinal flow will in general be induced within the channel and accompanied by vertical flow for mass conservation, including usually flow into the channel from above at ‘‘high latitudes’’ and flow out again at ‘‘low latitudes.’’ The channel is allowed to rotate about the z-axis at a rate . Magnetic fields are omitted in this first study. The channel fluid is always subadiabatically stratified, with the difference between the actual temperature gradient and the adiabatic gradient denoted by . A schematic diagram of the domain we have defined is shown in Figure 1. We look then for steady, x-independent solutions for the flow in the interior of this channel forced by latitudinal flow at the top. The variables include u; v; and w, the velocities in the x; y; and z directions, respectively; p, the perturbation pressure divided by the mean density; and T, the perturbation temperature. Perturbations are defined as departures from a 570 GILMAN & MIESCH horizontally uniform background state that is in hydrostatic balance. Given all of the assumptions above, the linearized dimensional equations that govern this problem are given by mass continuity @v @w þ ¼ 0; @y @z ð1Þ the horizontal equations of motion 2 v þ @2u ¼ 0; @z2 2 u @p @2v þ 2 ¼ 0; @y @z ð2Þ vertical hydrostatic balance @p ¼ g T @z 1 ¼ T ð3Þ (where T is the mean temperature), and the thermodynamic equation w @2T ¼ 0: @z2 ð4Þ In the above, we have eliminated the perturbation density from the hydrostatic relation by expanding the equation of state and linearizing and subtracting out the equation of state for the background, horizontally uniform thermodynamic variables. If we eliminate all dependent variables in favor of the meridional flow v that we are forcing at the upper boundary, we get the single equation @ 6 v 4 2 @ 2 v g @ 2 v þ 2 þ ¼ 0: @z6 @y2 @z2 ð5Þ For this equation to be dimensionally correct, the coefficients of the second derivative terms in equation (5) must both have dimensions of length to the negative fourth power. These lengths define the two boundary layer thicknesses of the problem, namely dE ¼ (=2 )1=2, the classical Ekman depth, and dBD ¼ (=g )1=4 , a thickness for what we call a buoyancy-diffusion layer, present in the nonrotating case and discussed in a geophysical context in Barcilon & Pedlosky (1967) and Veronis (1967). With these definitions, equation (5) becomes @6v 1 @2v 1 @2v þ 4 2þ 4 ¼ 0: 6 @z dE @z dBD @y2 ð6Þ 3. SOLUTIONS Having all constant coefficients, equation (6) clearly separates in y and z. The simplest forcing we can impose at the top that is consistent with our constraints is of the form v 0 sinðy=LÞ, in which L is the actual width of the channel. From equation (6), this forcing leads to the same functional form in y for all depths. So if we assume a separated solution of the form v ¼ v 0 e kz sin y ; L ð7Þ Vol. 611 then substitution of equation (7) into equation (6) yields the following equation for k: k 6 þ dE4 k 2 2 L 4 dBD ¼ 0; ð8Þ a bicubic equation already in standard form to be solved according to formulas in Burington (1962). Solving the bicubic equation for k 2 yields one real root and two roots that are complex conjugates of one another. The six roots of equation (8) are then obtained by taking the positive and negative square roots of these three solutions for k 2, yielding two real and four complex roots. Since there are six roots, a total of six boundary conditions at top and bottom can be satisfied, and so far we have applied only one. Three of the six solutions have a positive real part, implying a boundary layer structure with profiles exponentially declining away from the upper boundary in distances determined by combinations of the two lengths dE and dBD . A linear superposition of these three decaying roots (one real, two complex) yields the general solution for a semi-infinite domain. If a bottom boundary is included, all six solutions are required and a similar boundary layer will form near the bottom surface. Since four of the six roots are complex, the general solution will exhibit periodic structure (multiple cells) as well as exponential envelopes. Two limiting cases are particularly useful. When ! 0, dE ! 1, and equation (8) gives solutions of the form 1=3 2=3 dBD ; n ¼ 0; 1; 2; 3; 4; 5: ð9Þ k ¼ e in=3 L There is some real part to k for all , so all solutions have some exponential envelope. Except for numerical factors of order unity, the decay length k 1 of a typical solution is 1=6 1 1=3 2=3 1=3 : ð10Þ k L dBD L g It is instructive to evaluate equation (10) in the neighborhood of the solar tachocline. At tachocline depths, L 7:7 ; 1010 cm for the distance between the equator and a pole. The 1/6 power makes the boundary layer depth rather insensitive to all the other parameters. For the overshoot layer, the largest ; we can possibly assume is ~1012 cm2 s1, which is not much less than in the convection zone, but ; could be as low as 109 cm2 s1 each. Here g is related to the effective gravity of the layer, which is determined by the fraction that the actual subadiabatic temperature gradient is of the adiabatic gradient. For the overshoot layer, this fraction f is probably no larger than 104 and might be as small as 105 or even 106. Here g at tachocline depths is 5 ; 104 cm s2. In more detail, g g f ð 1= Þ=Hp , in which Hp is the pressure scale height at tachocline depths and is the ratio of specific heats. With these numbers, the range of estimates for k 1 is 1600 k 1 3:4 ; 104 km (overshoot layer). By these estimates, then, the penetration ignoring rotation effects will be between 0.1% of the solar radius and the depth of the overshoot layer, which is very unlikely to be as large as 3:4 ; 104 km (5% of the solar radius). Various estimates (Rempel 2004, and references therein) put it closer to 1% of the solar radius. In the radiative layer below the overshooting, we expect ; to be much smaller and the effective gravity to be much larger, so the penetration into this No. 1, 2004 SOLAR MERIDIONAL CIRCULATION PENETRATION 571 Fig. 2.—General solution of eq. (8) for a semi-infinite layer in solar parameter regimes. Of the three roots that decay with depth (positive real part), one is real (k1 ) and two are complex conjugates (k2 , k3 ). The left panel shows the amplitude of k1 as a function of the two parameters dE and dBD (the influence of rotation increases downward and the influence of stratification increases toward the left). The center and right panels show similarly the amplitude of the real and imaginary parts of k2 (and, in turn, k3 ). The values that correspond to the colors and contour levels are indicated by the color bar at right. layer is much smaller still. For f 101 and ; 106 cm2 s1, we get k 1 50 km! Clearly, from this boundary layer process no effective penetration is possible below the overshoot layer. From helioseismic inferences, the bottom of the nearly adiabatic layer associated with the convection zone is no further in than 0.71 R (Christensen-Dalsgaard et al. 1991; Basu & Antia 1997; Charbonneau et al. 1999). Thus it is not nearly deep enough to allow the meridional circulation to reach down to 0.6 R, as Nandy & Choudhuri (2002) require. The boundary layer characterized by equation (10) has been discussed in, e.g., Barcilon & Pedlosky (1967) and Veronis (1967), but there the boundary layer occurs on the side wall of the fluid container, rather than its top. Steady boundary-layer type solutions are possible because, even without rotation, vertical diffusion of temperature is balanced by vertical advection along the subadiabatic temperature gradient, while the viscous forces are balanced by pressure work, which links back to the stable stratification also. In Lagrangian terms, steady flow occurs because the residence time of every fluid particle that enters the domain at high latitudes and exits at low latitudes is finite. The boundary layer is thicker for larger and/or k because more downward diffusion can take place during this residence time. The boundary layer is thinner when the stratification is more subadiabatic because this combats downward diffusion of temperature and it requires more work to alter the pressure contours to drive the flow. How does the presence of rotation change the amount of penetration? When rotation dominates, the term involving dBD in equation (8) can be discarded, and we reduce the problem to four solutions of the form k ¼ e i=2 dE1 ; ð11Þ which is the classical Ekman boundary layer problem (Pedlosky 1987), for which the boundary layer depth k 1 dE . For the solar tachocline, 2:6 ; 106 s1 and 109 1012 cm2 s1, as we assumed earlier, the range for this depth is 140 k 1 4400 km. Thus, the Ekman depth is generally smaller than in the buoyancy-diffusion depth in the overshoot layer, so penetration is only constrained further. The nature of the boundary layer is governed by the product PrBu, where Pr is the Prandtl number, Pr ¼ =, and Bu is the Burger number, Bu ¼ (2 L=ND)2 , which measures the relative influence of rotation and stratification (D is the thickness of the layer and N is the Brunt-Vaisala frequency; N 2 ¼ g ). If PrBu 3 1 the solution will be dominated by the buoyancy-diffusion layer, but if PrBuT1 then the Ekman layer will prevail. In the solar interior below the convective envelope, PrBu is estimated to be between 104 and 101, so the Ekman layer is expected to play the biggest role, except near the equator where the vertical component of the rotation vector vanishes. As with the buoyancy-diffusion layer associated with dBD , the Ekman depth in the radiative interior is much less than in the overshoot layer. If we use ¼ 106 cm2 s1 as earlier, we get k 1 5 km there. So again, there is no effective penetration into the radiative layer. Results for the general case when both boundary layers are included are shown in Figure 2. These show the form and size of the vertical attenuation ‘‘rate’’ k from equation (8) for the three solutions k1 ; k2 ; k3 that decay as z ! 1 as functions of the two boundary layer lengths. The log scale range for k is 0.01–1000 Mm1. The units on the axes are Mm. We can see that the domains in which the two boundary layers dominate are clearly separated. Where the k contours are horizontal, the Ekman layer dominates, root k1 vanishes, and k2 and k3 are complex conjugates of each other. Where the contours are vertical, the buoyancy-diffusion layer dominates, and the decay lengths represented by the real parts of k1 , k2 , k3 differ only by trigonometric factors of order unity (see eq. [9]). The purely real root k1 disappears (black) when the subadiabaticity ! zero, because the thermodynamics becomes decoupled from the equations of motion and the effect of temperature diffusion is lost. In that limit, the original equations yield a fourth-order rather than a sixth-order equation for k. We can see from Figure 2 that the combined boundary layer thicknesses, as defined by R(k)1, are not that different from the asymptotic limits. Therefore, the ranges of k 1 estimated above apply in the general case quite well. Therefore, our conclusion that there is no significant penetration of meridional circulation below the overshoot layer applies equally 572 GILMAN & MIESCH Vol. 611 well when both boundary layer depths must be taken into account. 4. IMPROVEMENTS FOR GREATER REALISM One obvious improvement we can make for solar applications is to go to spherical geometry. When we do that, we can reduce the problem to a single equation for v, completely analogous to equation (6), of the form @6v cos2 @ 2 v 1 @ 1 @ (v sin ) ¼ 0; þ þ 4 @ sin @ @z6 dE4 @z2 rt2 dBD ð12Þ in which rt is the radius of the spherical shell at the tachocline and is the colatitude. Thus the same two boundary layer lengths appear, but the coefficients of the equation are no longer constant; rather, they depend on the colatitude . However, the coefficients are still independent of z, so we still expect solutions e kz as in the Cartesian problem. Equation (12) can be solved analytically in the nonrotating case, where the problem reduces exactly to the Cartesian problem in z if we assume a top forcing for v sin cos . This leads to a form for k 1 similar to equation (9) for Cartesian geometry, with the factor (=L)1=3 replaced by (6=rt )1=3. The nonrotating case is interesting from a mathematical standpoint because it admits analytic solutions that are directly analogous to the Cartesian case. However, it is not very realistic because the Sun and other stars do in fact rotate and this profoundly effects the structure of the meridional circulation. Thus, in the remainder of this section we consider the rotating, spherical system expressed by equation (12). For this we need a series of Legendre polynomials in order to describe the structure of the resulting circulation. From equation (12), we should expect that the boundary layer depth dBD dominates near the equator, and the boundary layer thickness should vary with latitude. We show in Figure 3 a typical example of the flow found in a spherical shell as a solution to equation (12), obtained using a code that solves the axisymmetric, nonmagnetic version of the thin-shell equations derived by Miesch & Gilman (2004), evolving the flow until it reaches a steady state. Here we have imposed a latitudinal flow at the top of the form sin cos . The thickness of the layer corresponds to ~15 Mm in the overshoot layer of the Sun and ~ 0.2 Mm in the radiative zone. The dashed white curve gives the Ekman depth dE assumed and the dotted straight line shows the buoyancy-diffusion layer depth dBD taken. The top frame depicts the latitudinal flow, v, and the bottom frame the radial motion, w. The boundary layer, as expected, is thicker near the equator than near the poles, but is at all latitudes close to dE and dBD . Penetration is only a fraction of the depth shown. There is also an induced differential rotation (not shown) inside the shell, even though none is imposed at the top. This arises from the action of coriolis forces on the meridional circulation, leading to the so-called thermal wind. In the linear approximation, however, this has no effect on the penetration depth of the meridional circulation. Thus the spherical system, while not as simple as the Cartesian one analyzed above, gives very similar results. The spherical problem is also straightforward to solve when a differential rotation, even a large one, is imposed on the top boundary, such as the solar convection zone imposes on the solar tachocline. To solve this problem requires retaining more terms in the equations of motion, and we save this for a later paper. But adding this differential rotation, Fig. 3.—Steady-state meridional circulation (satisfying eq. [12]) for a thin shell subject to a radial and latitudinal velocity imposed on the upper surface. The upper and lower frames display the latitudinal and vertical velocity, respectively, with contour lines at v ¼ 0 and w ¼ 0 in order to illustrate the multicellular structure. Yellow/orange tones denote southward and outward flow while blue/black tones pffiffiffidenote northward and2=3inward flow. The solution corresponds to dE ¼ (0:1= 2) D at the poles and d BD L1=3 ¼ 0:17D, where D is the thickness of the layer. The local values of dE and dBD obtained from the boundary layer analysis are shown on the upper frame as dashed and dotted curves respectively. or thermal wind as it is often called in geophysical fluid dynamics, is unlikely to change our results qualitatively. If the amplitude of the imposed differential rotation is small, it can be treated linearly and then it has no effect on the meridional circulation. A larger differential rotation can alter the circulation by modifying the centrifugal forces, but this effect has to compete with the negative buoyancy force associated with the subadiabatic stratification. The effect of rotation compared to gravity on the Sun is measured by its oblateness, and so is P104. The part of that from differential rotation would therefore be 105. This effect would therefore be totally negligible in the radiative zone, where the effective gravity associated with the subadiabatic stratification there is 101 g. In the overshoot layer it might make a quantitative difference, but would certainly not dominate. We have also not included effects of Lorentz forces associated with the likely very large amplitude toroidal field in the tachocline. But this force is certain to oppose penetration, not enhance it. 5. DISCUSSION AND CONCLUSIONS We have shown here that there are two boundary layer thicknesses that govern the amount a meridional circulation originating in the convection zone can penetrate through the tachocline to the solar interior. One of these, a buoyancydiffusion layer, which had been previously reported in the geophysical fluid dynamics literature, is independent of rotation and sets an upper limit to the amount of penetration possible; hence its value in examining the nonrotating limit. The second boundary layer, the classical Ekman layer, shrinks with increasing rotation and, except near the solar equator where it becomes infinite, further constrains the amount of No. 1, 2004 SOLAR MERIDIONAL CIRCULATION PENETRATION penetration. The overall result is that for plausible values of the subadiabatic stratification in the solar tachocline, the meridional circulation cannot penetrate significantly below the bottom of the overshoot part of the tachocline, near 0.7 R . It is important to realize that even though the Cartesian analysis reduces the problem to algebra, the system is truly two-dimensional, not one-dimensional, in that a finite length L in latitude ( y) is required to generate any penetration at all. From equation (8), if L ! 1 then the problem reduces to the classical Ekman layer adjacent to an infinite plane boundary, vertical motions vanish, and the buoyancy-diffusion layer plays no role. For finite L, the y-bounded Cartesian system is made to mimic the confinement of a spherical shell such as the solar tachocline and its neighborhood. These fundamental boundary layer thicknesses are the same in Cartesian and spherical geometries, and therefore the results and conclusions are essentially the same, even though the rotating spherical case does not lend itself readily to analytical solution. Given that the combination of these two boundary layers have not to our knowledge been discussed previously in the astrophysical literature, and since they should be relevant to the interface between convective and radiative domains in the interior of stars generally, we use the Cartesian solutions to elucidate their properties in combination in the clearest possible way. The thin-shell model used here breaks down for small-scale circulations that are not necessarily in hydrostatic balance as reflected by equation (3). We have not specified a driving mechanism for the meridional flow in the convection zone, because none is required for the theory. It may well be that such meridional flow arises because the Sun is rotating, but rotation is not a requirement for the existence of meridional circulation. In any case, only meridional flows within the bulk of the convection zone and the tachocline are relevant to flux-transport dynamo theories for the Sun. Futhermore, they must have sufficient amplitude to play a role on a timescale less than a solar cycle. This amplitude is set empirically with reference to surface doppler and helioseismic measurements. In radiative regions of rotating stars, the distortion of the gravitational potential by centrifugal forces results in a thermal imbalance that tends to drive global motions often referred to as Eddington-Sweet circulations (e.g., Tassoul 1978). A variety of other mechanisms can also drive or alter meridional circulations near the base of the convection zone, including compositional gradients (e.g., Maeder & Zahn 1998), rotational and baroclinic instabilities (e.g., Zahn 1992; Maeder & Meynet 2000), gravity waves (e.g., Fritts et al. 1998), and magnetic fields (e.g., Gough & McIntyre 1998). However, these circulations are generally very weak relative to those in the convection zone, with turnover timescales of thousands or millions of years, and are therefore of less relevance to solar dynamo theory. 573 Observations of elemental abundances in the Sun reveal a depletion of lithium relative to cosmic abundances, which is attributed at least in part to nuclear burning over the Sun’s main-sequence lifetime (e.g., Michaud & Charbonneau 1991; Pinsonneault 1997). Lithium is destroyed at temperatures above 2:5 ; 106 K, which in the Sun corresponds to a radius of ~ 0.68 R , well below the convective envelope. Thus, the observed depletion has been attributed to enhanced mixing below the convection zone, which may be due to meridional circulation or to a number of alternative mechanisms such as rotational instabilities or gravity waves (e.g., Michaud & Charbonneau 1991; Pinsonneault 1997). However, the observed depletion of other light elements that are burned at higher temperatures is much less, suggesting that the mixing does not reach much deeper into the radiative interior. The most important of these is beryllium, which burns at 3:5 ; 106 K, suggesting that efficient mixing by meridional circulation or other processes cannot extend deeper than 0.55 R. More sophisticated estimates obtained from applying elemental abundance observations to one-dimensional stellar structure models favor no mixing down as far as r 0:6 R (e.g., Michaud & Charbonneau 1991; Chaboyer et al. 1995; Pinsonneault 1997; Brun et al. 2002). However, this constraint on meridional circulation is not as strict as the one we have found, which is r 0:7 R . Our fundamental conclusion from this analysis is that the physics of the solar tachocline and neighboring regions does not allow penetration of meridional circulation originating in the solar convection zone below the overshoot layer. However, some clarification is necessary here because there are several alternative ways in which the extent of the overshoot layer in the Sun may be defined. For example, the penetration depth may be determined based on the kinetic energy or enthalpy flux of the convection or on the efficiency of tracer particle transport (e.g., Zahn 1991; Brummell et al. 2002; Rempel 2004). For our purposes we define the base of the overshoot region rb as the radius above which the stratification is nearly adiabatic and below which the stratification is substantially subadiabatic. Although little is known about the detailed structure of the overshoot region at the base of the solar convection zone, rb is reasonably well established from helioseismic inversions, with a value of 0.713 R (Christensen-Dalsgaard et al. 1991; Basu 1997). Thus, our results indicate that penetration inward to 0.6 R , needed to make certain flux-transport dynamo models apply to the Sun, is ruled out. We thank Matthias Rempel and the anonymous reviewer for helpful suggestions, which have improved the manuscript. This work was partly supported by NASA through the SEC theory program, work orders W-10,177 and W-10,175. REFERENCES Barcilon, V., & Pedlosky, J. 1967, J. Fluid Mech., 29, 609 Cally, P. S. 2003, MNRAS, 339, 957 Basu, S. 1997, MNRAS, 288, 572 Chaboyer, B., Demarque, P., & Pinsonneault, M. H. 1995, ApJ, 441, 865 Basu, S., & Antia, H. M. 1997, MNRAS, 287, 189 Charbonneau, P., Christensen-Dalsgaard, J., Henning, R., Larsen, R. M., ———. 2003, ApJ, 585, 553 Schou, J., Thompson, M. J., & Tomczyk, S. 1999, ApJ, 527, 445 Braun, D. C., & Fan, Y. 1998, ApJ, 508, L105 Christensen-Dalsgaard, J., Gough, D. O., & Thompson, M. J. 1991, ApJ, Brummell, N. H., Clune, T. L., & Toomre, J. 2002, ApJ, 570, 825 378, 413 Brun, A. S., Antia, H. M., Chitre, S. M., & Zahn, J.-P. 2002, A&A, 391, 725 Dikpati, M., & Charbonneau, P. 1999, ApJ, 518, 508 Brun, A. S., & Toomre, J. 2002, ApJ, 570, 865 Dikpati, M., Corbard, T., Thompson, M. J., & Gilman, P. A. 2002, ApJ, Burington, R. S. 1962, Handbook of Mathematical Tables and Formulas 575, L41 (New York: McGraw-Hill) Dikpati, M., & Gilman, P. A. 2001, ApJ, 559, 428 574 GILMAN & MIESCH Elliott, J. R., Miesch, M. S., & Toomre, J. 2000, ApJ, 533, 546 Fritts, D. C., Vadas, S. L., & Andreassen, O. 1998, A&A, 333, 343 Giles, P. M., Duvall, T. L., Scherrer, P. H., & Bogart, R. S. 1997, Nature, 390, 52 Gough, D. O., & McIntyre, M. E. 1998, Nature, 394, 755 Haber, D. A., Hindman, B. W., Toomre, J., Bogart, R. S., Larsen, R. M., & Hill, F. 2002, ApJ, 570, 855 Hathaway, D. H. 1996, ApJ, 460, 1027 Lantz, S. R., & Fan, Y. 1999, ApJS, 121, 247 Maeder, A., & Meynet, G. 2000, ARA&A, 38, 143 Maeder, A., & Zahn, J.-P. 1998, A&A, 334, 1000 Michaud, G., & Charbonneau, P. 1991, Space Sci. Rev., 57, 1 Miesch, M. S., Elliott, J. R., Toomre, J., Clune, T. C., Glatzmaier, G. A., & Gilman, P. A. 2000, ApJ, 532, 593 Miesch, M. S., & Gilman, P. A. 2004, Sol. Phys., in press Nandy, D., & Choudhuri, A. R. 2002, Science, 296, 1671 Pedlosky, J. 1987, Geophysical Fluid Dynamics, (2nd ed.; New York: Springer) Pinsonneault, M. 1997, ARA&A, 35, 557 Rempel, M. 2004, ApJ, 607, 1046 Schou, J., & Bogart, R. S. 1998, ApJ, 504, L131 Spiegel, E. A., & Veronis, G. 1960, ApJ, 131, 442 Stix, M. 2002, The Sun (2nd ed.; Berlin: Springer) Tassoul, J.-L. 1978, Theory of Rotating Stars ( Princeton: Princeton Univ. Press) Veronis, G. 1967, Tellus, 19, 326 Zahn, J.-P. 1991, A&A, 252, 179 ———. 1992, A&A, 265, 115
© Copyright 2026 Paperzz