The Astrophysical Journal 00000 (14pp) june 17 2020Preprint typeset using LATEX style emulateapj v 012315

REVISITING THE SUNrsquoS STRONG DIFFERENTIAL ROTATION ALONG RADIAL LINES

Loren I Matilsky1 Bradley W Hindman and Juri ToomreJILA amp Department of Astrophysical and Planetary Sciences University of Colorado Boulder CO 80309-0440 USA

The Astrophysical Journal 00000 (14pp) june 17 2020

ABSTRACT

Current state-of-the-art models of the solar convection zone consist of solutions to the Navier-Stokesequations in rotating 3D spherical shells Such models are highly sensitive to the choice of boundaryconditions Here we present two suites of simulations differing only in their outer thermal boundarycondition which is either one of fixed-entropy or fixed-entropy-gradient We find that the resultingdifferential rotation is markedly different between the two sets The fixed-entropy-gradient simulationshave strong differential rotation contrast and isocontours tilted along radial lines (in good agreementwith the Sunrsquos interior rotation revealed by helioseismology) whereas the fixed-entropy simulationshave weaker contrast and contours tilted in the opposite sense We examine in detail the force balancesin our models and find that the poleward transport of heat by Busse columns drives a thermal windresponsible for the different rotation profiles We conclude that the Sunrsquos strong differential rotationalong radial lines may result from the solar emissivity being invariant with latitude (which is similarto the fixed-entropy-gradient condition in our models) and the poleward transport of heat by Bussecolumns In future work on convection in the solar context we strongly advise modelers to use afixed-gradient outer boundary conditionKeywords convection mdash turbulence mdash Sun interior mdash Sun rotation mdash Sun kinematics and dy-

namics

1 INTRODUCTION

Helioseismology has revealed in detail the internal ro-tation profile of the solar convection zone (CZ egThompson et al 2003 Howe et al 2005) as shownin Figure 1 The most notable properties of the rotationrate are that the equator rotates significantly faster thanthe high-latitude regions and that the isorotation con-tours are tilted significantly with respect to the rotationaxis falling largely along radial lines Furthermore thereare two shear layers at the top and bottom of the CZ atthe top the contours bend toward the equator in a regionknown as the near-surface shear layer (NSSL) and at thebottom the differential rotation in the CZ transitions tosolid-body rotation over a narrow boundary layer calledthe tachocline Prior to helioseismic probing most theo-reticians had assumed that the differential rotation thatis observed directly at the surface would imprint into theinterior along isosurfaces parallel to the rotation axishence satisfying the Taylor-Proudman theorem The he-lioseismic observations have clearly demonstrated thatthis theoretical supposition was wrong

For the last several decades global 3D supercom-puter simulations of hydrodynamic convection in rotat-ing spherical shells have succeeded in achieving rotationprofiles that are fast at the equator and slow at the polesHowever simulations generally have a weaker overall dif-ferential rotation contrast than that of the Sun If thecontrast is defined to be the difference in rotation ratebetween the equator and 60 latitude expressed as a per-centage of the ldquoframerdquo rotation rate then for the Sunthis magnitude is sim20 Most simulations on the otherhand have rotation-contrast magnitudes of sim10 al-though there are some notable exceptions (eg Brun ampToomre 2002 Brown et al 2010 Matilsky amp Toomre

1 lorenmatilskycoloradoedu

Figure 1 Contour map in the upper meridional plane of theinternal rotation profile of the Sun in and below the CZ averagedin longitude and time from 1995ndash2009 The rotation rate has beenobtained using a regularized least-squares (RLS) inversion whichis sensitive only to the equatorially symmetric part of the rotationThe dashed lines are at a 25 angle to the rotation axis and alignwith the isorotation contours at mid-latitudes Image credit Howeet al (2005) extended to include GONG data until 2009

2020) To date however there is no systematic physi-cal explanation for why these particular simulations havehigh rotation contrast

Simulations have also struggled to achieve rotationcontours in the bulk of the CZ that are significantly tiltedfrom the axis as seen in Figure 1 With some exceptions(eg Elliot et al 2000 Miesch et al 2006) the simu-lations generally have cylindrically-aligned contours Inthe case of Miesch et al (2006) the contours were tiltedsystematically by imposing a modest latitudinal entropygradient at the base of the CZ By modifying the mag-nitude of this entropy gradient Miesch et al (2006)

arX

iv2

004

0020

8v2

[as

tro-

phS

R]

16

Jun

2020

2 Matilsky et al

controlled the magnitude of the contour tilt Imposingthis boundary condition was inspired by thermal-windbalance in the tachocline imprinting up into the CZ

There is also good evidence for thermal-wind balanceoperating in the deep CZ from analytical models Bal-bus (2009a) Balbus et al (2009b) assumed a simplefunctional relationship between the entropy and angularvelocity and solved explicitly for the combined isorota-tional and isentropic contours as characteristics of thethermal-wind equation This yielded tilted contours ingood agreement with helioseismology for the bulk of theCZ (away from the boundary layers of shear at the topand bottom) Such analytical workmdashin addition to lend-ing strong support for the presence of a thermal windin the solar interiormdashpoints to the need to better un-derstand the dynamical coupling between entropy andangular velocity in global numerical simulations some-thing we address in the present paper

In this work we explore the role that the outer ther-mal boundary condition plays in conjunction with aninterior thermal wind in modifying the resulting differ-ential rotation We consider the two most commonlyused options fixed-entropy (FE) in which the entropyat the outer boundary is fixed to a constant value andfixed-flux (FF) in which the radial entropy gradient atthe outer boundary is fixed therefore implying an out-ward conductive flux that is independent of latitude SeeHurle et al (1966) and Edwards (1990) for descriptionsof these boundary conditions in the context of linear the-ory and Anders et al (2020) for a detailed analysisof the boundary conditions in nonlinear simulations ofRayleigh-Benard convection The FF condition is moreappropriate for the Sun since the radiant flux from thesolar photosphere does not appear to vary substantiallywith latitude Rast et al (2008) analyzed full-disk im-ages from the Precision Solar Photometric Telescope atthe Mauna Loa Solar Observatory and calculated non-magnetic contributions to the solar photospheric inten-sity In both continuum and Ca II K intensity distribu-tions only a sim01ndash02 variation was observed corre-sponding to a solar pole that is at most sim25 K warmer(in terms of effective temperature) than the equator Thededuced near-spherical-symmetry of the solar emissiveflux is a significant observation since a thermal windstrong enough to drive the observed differential rotationwould require greater differences in the interior temper-ature between equator and pole We return to this pointin our concluding remarks

We do not address the dynamics of the near-surfacenor tachocline shear layers in this work In particularour models have an impenetrable lower boundary thatdoes not allow for the convective overshoot of downflowplumes into the stable region that may play a role in theorigin of the tachocline The dynamical maintenance ofthe NSSL is still an open question as discussed in Hottaet al (2015) and Matilsky et al (2019) and modelstend to only display signs of near-surface shear if theyhave high density contrast (gtsim100) across the CZ Toavoid high computational cost the models in this workhave a smaller density contrast of sim20

Solar-like differential rotation (fast equator and slowpoles) in spherical-shell convection is thought to be dueto the outward transport of angular momentum by Bussecolumns (also called ldquoTaylor columnsrdquo and ldquobanana

cellsrdquo eg Busse 2002 Jones et al 2011) In thiswork we show that Busse columns also transport heatpoleward and equivalently drive a solar-like differentialrotation through a thermal wind We find that in ourFF cases the thermal wind drives stronger differentialrotation magnitudes and achieves more significant tilt inthe rotation contours than the corresponding FE casesElliot et al (2000) noted this effect for one simulationbut did not explore the underlying mechanism

In Section 2 we describe the parameter space exploredby our simulation suite as well as the mathematical de-tails of the FF and FE boundary conditions In Section3 we describe the basic results of our experiment fo-cusing on the achieved differential rotation In Section4 we quantify the force balance achieved in our modelswhich for the radial and latitudinal directions consistsof a thermal wind in spherical geometry In Section 5we examine the latitudinal transport of energy by Bussecolumns that is responsible for the thermal wind InSection 6 we discuss how the effects of the thermal windare modified by the outer thermal boundary conditionIn Section 7 we discuss our simulation results in the con-text of the Sun

2 NUMERICAL EXPERIMENT

We consider time-dependent 3D simulations of a ro-tating stratified spherical shell of fluid representativeof the solar CZ We use the open-source code Rayleigh091 (Featherstone amp Hindman 2016a Matsui et al2016 Featherstone 2018) which solves the equationsof hydrodynamics in spherical geometry Our domain isa spherical shell with inner radius ri and outer radiusro We describe this domain using the spherical coor-dinates (r θ φ) and corresponding unit vectors (er eθeφ) Here r is the spherical radius θ the colatitude andφ the azimuthal coordinate When convenient we alsouse cylindrical coordinates (λ φ z) = (r sin θ φ r cos θ)and unit vectors (eλ eφ ez) where λ is the cylindricalradius and z the axial coordinate perpendicular to theequatorial plane

The thermodynamic reference state is chosen to betemporally steady and spherically symmetric with adi-abatic stratification (see Jones et al 2011 for a com-plete description) The density varies by a factor of sim20(three density scale heights) across the layer We de-note the pressure density temperature and entropy byP ρ T and S respectively using overbars to indicatethe fixed reference state and the absence of overbars toindicate deviations from the reference state

Rayleigh uses an anelastic approximation (eg Gough1969 Gilman amp Glatzmaier 1981 Jones et al 2011)

which removes sound waves from the system making themaximum allowable timestep much larger since it is lim-ited by the flow velocity rather than the sound speed Ina frame rotating with angular velocity Ω0 = Ω0ez thefluid equations are then given by (eg Featherstone ampHindman 2016a)

nabla middot (ρv) = 0 (1)

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 3

ρ

[partv

partt+ (v middot nabla)v

]=minus 2ρΩ0 times v

minus ρnabla

(P

ρ

)minus ρS

cpg +nabla middotD (2)

and

ρT

[partS

partt+ v middot nablaS

]= nabla middot

[κρTnablaS

]+Q+ D nablav (3)

Here v = (vr vθ vφ) is the fluid velocity in the rotatingframe cp is the constant-pressure specific heat g is thegravitational acceleration due to a solar mass M locatedat the center of the spherical shell D = ρν[nablav +nablavT minus(23)(nablamiddotv)I] is the Newtonian viscous-stress tensor I isthe identity tensor ν is the kinematic viscosity and κ isthe thermal diffusivity Because it is not computationallypossible to resolve convection in the solar regime downto the turbulent microscale ν and κ must be regardedas ldquoeddyrdquo diffusivities which for simplicity we choose tobe spatially constant and equal such that the Prandtlnumber is unity The internal heating function whichphysically represents heating due to radiation is chosento have the fixed radial profile Q(r) = α[P (r) minus P (ro)]with the normalization constant α chosen such that asolar luminosity L is forced through the domain

The heating function Q is designed such that most ofthe energy is deposited in the bottom sim13 of the do-main (see Featherstone amp Hindman 2016a for the radialprofile of the heat flux associated with Q) In the Sunthe internal heating arises from high-opacity metals inthe CZ absorbing the radiation streaming in from theinterior This internal heating is quite distributed andthe heat flux associated with our imposed Q agrees wellwith the radiative flux inferred from standard solar mod-els (for example model S in Christensen-Dalsgaard etal 1996) The distributed nature of the heating makesthe convection problem in the solar CZ slightly differentfrom a standard Rayleigh-Benard setup in which thefluid layer is heated from below by imposing a conduc-tive flux through the lower boundary

The equation of state for the system is that of a perfectgas subject to small thermodynamic perturbations aboutthe reference state

ρ

ρ=P

Pminus T

T=

P

γPminus S

cp (4)

where γ = 53 is the ratio of specific heatsWe adopt stress-free and impenetrable boundary con-

ditions to conserve angular momentum and mass

vr =part

partr

(vθr

)=

part

partr

(vφr

)= 0 at r = ri and ro (5)

In all cases the inner thermal boundary condition al-lows no flux of energy into the system through the lowerboundary

partS

partr= 0 at r = ri (6)

Input parameters common to all the simulations exploredhere are shown in Table 1

Table 1Common input-parameter values for all simulations

ri 500 times1010 cm = 0719 Rro 659 times1010 cm = 0947 Rcp 350 times108 erg Kminus1 gminus1

γ 167ρi 0181 g cmminus3

L 385 times1033 erg sminus1

M 199 times1033 gR 696 times1010 cm

Pr equiv νκ 100

21 Outer thermal boundary condition

The main purpose of this work is to characterize theinfluence of the outer thermal boundary condition on thebehavior of the resulting differential rotation We con-sider models with different background rotation rates Ω0

and diffusion values (ν = κ) and for each model analyzetwo sub-cases

S = 0 at r = ro (fixed entropy or FE) (7)

and

partS

partr= minus L

4πr2oρTκat r = ro (fixed flux or FF) (8)

The solar luminosity that is injected into the systemvia internal heating is ultimately carried out through theouter surface via thermal conduction which in our mod-els arises from entropy gradients (see Equation (3)) Forthe fixed-entropy condition (7) the interior is initiallyheated (leading to S gt 0 in the lower parts of the CZ)while the entropy at the outer surface is ldquopinnedrdquo tozero This naturally establishes a thermal boundary layer(sharp entropy gradients partSpartr lt 0) just below the outersurface The steepness of the gradient (ie the strengthof the outward conductive loss of energy) is allowed tovary with latitude

For the fixed-flux condition (8) the outer thermalboundary layer is present from the beginning of the sim-ulation The steepness of the entropy gradient (and thusthe energy loss) at the outer surface is independent oflatitude by construction and is forced to have exactlythe value needed to carry out a solar luminosity Thefixed-flux condition is thus more ldquosolar-likerdquo since inthe Sun there is no observed latitudinal dependence ofthe emergent intensity which is equal to the energy lostvia radiative cooling at the photosphere

For both the FE and FF cases the thermal conductiveboundary layer stands in contrast to the real solar photo-sphere in which radiative cooling removes the heat froma very thin (sim100 km) outer layer The cooling drivesvery small temporal and spatial scales of motion com-pared to the deep interior (such as granulation and super-granulation) making its direct inclusion in global mod-els problematic Researchers have sought to address thisdifficulty in various ways Nelson et al (2018) imple-mented stochastic driving of convection by near-surfaceplumes designed to mimic the effects of supergranulationfinding that that the flow structures and transport prop-erties were significantly altered in the deep CZ Hotta etal (2019) simulated the whole CZ with no rotation ormagnetic field coupling a global spherical shell that cap-tured large-scale flows in the deep interior to a Cartesian

4 Matilsky et al

box that solved the equations of radiative transfer in thephotosphere They found that the near-surface motionshad a weak influence on the deep interior Regardlessof its relevance to interior flow structures correctly cap-turing the small-scale near-surface flows in global modelsis currently prohibitively expensive computationally Inorder to explore a wider range of parameter space wethus only consider the FE and FF boundary conditionshere

3 SIMULATION RESULTS

We label simulations with a prefix that signifies theouter boundary condition (ldquoFErdquo for Equation (7) andldquoFFrdquo for Equation (8)) followed by the value of the dif-fusion constant ν = κ (in units of 1012 cm2 sminus1) followedby the value of the rotation rate (in units of the siderealCarrington value for the Sun Ω equiv 287times 10minus6 rad sminus1or Ω2π equiv 456 nHz) For example ldquocase FE4-3rdquo refersto a simulation with an FE outer boundary for whichν = κ = 4 times 1012 cm2 sminus1 throughout the domain andΩ0 = 3Ω

Table 2 (in the Appendix) contains the values of thenon-dimensional parameters as well as the grid resolu-tion for each of the 18 simulations considered in thiswork Table 2 has four groupings according to FE or FFat two different rotation rates Following the notationof Featherstone amp Hindman (2016b) we parameterizethe strength of the imposed driving in each simulationthrough a bulk ldquoflux Rayleigh numberrdquo (imposed a pri-ori)

Ra equiv gFH4

cpρT ν3(9)

and the level of turbulence through bulk Reynolds orPeclet numbers (calculated a posteriori)

Re =Pe

Pr=vprimeH

ν (10)

Since the Prandtl number for all models is unity thePeclet number Pe = vprimeHκminus1 equals the Reynolds num-ber

Similarly we parameterize the influence of rotationthrough an Ekman number (imposed a priori)

Ek equiv ν

2Ω0H2(11)

and a bulk Rossby number (calculated a posteriori)

Ro equiv vprime

2Ω0H (12)

In the preceding equations the length scale H is takento be the shell depth ro minus ri the tildes refer to volumeaverages over the full spherical shell and F refers to theenergy flux associated with conduction and convectionin equilibrium (see Featherstone amp Hindman 2016b)The typical convective velocity amplitude vprime refers to therms of the velocity with the longitudinally averaged partsubtracted the mean being taken over time and over thefull volume of the shell Throughout this work temporalaverages are taken during the latter portion of run timefor which there is statistical equilibriummdashgenerally sim34of the total run time listed in Table 2

Before discussing our results in detail we note thatall our models have fairly high levels of thermal and vis-cous diffusion Furthermore all our models rotate at ei-ther two or three times the solar Carrington rate Thesechoices which stand in contrast to the physics of thesolar interior ensure that our models have low enoughRossby numbers to avoid antisolar differential rotation(fast poles slow equator) All global spherical-shell con-vection codes produce high velocities at large scales inthe solar context when sufficiently turbulent The in-fluence of rotation on the large scales is therefore weakwhich causes less coherence in the outward angular mo-mentum transport by the convection and ultimately lessangular momentum in the outer layers than the inner lay-ers (ie an antisolar differential rotation) The overallproblemmdashthat increasing the turbulence in simulationsleads to antisolar statesmdashis now called the ldquoconvectiveconundrumrdquo (OrsquoMara et al 2016) The antisolar statescan be avoided by raising the rotation rate raising thediffusions or lowering the luminosity We choose a com-bination of the former two which requires that our mod-els are only moderately turbulent Nonetheless the vis-cous force and heat flux are small to leading order in theprimary dynamical balances

Returning to our simulation results we quantify themagnitude of the overall differential rotation contrast asthe difference in the outer-surface rotation rate betweenthe equator and 60-latitude normalized by the framerotation rate

∆Ω

Ω0equiv Ω(ro π2)minus Ω(ro π6)

Ω0 (13)

where

Ω(r θ) equiv Ω0 +〈vφ〉r sin θ

(14)

is the rotation rate of the fluid as a function of r andθ and the angular brackets denote a combined temporaland longitudinal average From Table 2 the FF caseshave differential rotation contrasts ∆ΩΩ0 that are sig-nificantly greatermdashon average by sim40mdashthan those ofthe FE cases For comparison the solar value of the ro-tation contrast is substantially higher than in any of ourmodels ∆ΩΩ = 020 (see Howe et al 2000 Fig-ure 1) For the solar estimate we have taken Ω to bethe sidereal Carrington rate and ro to lie just below thenear-surface shear layer

Figure 2 shows how the differential rotation fractionscales with the reduced Rayleigh number which accountsfor the increase to the critical Rayleigh number for con-vective onset caused by rotation (Chandrasekhar 1961)The reduced Rayleigh number thus serves as a better pa-rameterization of the supercriticality of the system thansimply the Rayleigh number From Figure 2 each type ofboundary condition yields a similar scaling with the re-duced Rayleigh number R For the Ω0 = 3Ω cases (cir-cles in Figure 2) the rotation contrast increases mono-tonically (but with decreasing slope) with increasing Rso that the curves connected by circles in Figure 2 areconcave-down For the Ω0 = 2Ω cases (triangles in Fig-ure 2) the curves ldquooverturnrdquo so that a peak value of therotation contrast (at around R sim 27) is achieved Thisbehavior (concisely described in Gastine et al 2013) is asymptom of the convective conundrum as models grow

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 5

Figure 2 Rotation contrast achieved by the suite of simula-tions plotted as a function of the reduced Rayleigh number R equivRa Ek43 Circles mark the Ω0 = 3Ω cases and triangles theΩ = 2Ω cases

Figure 3 Temporally and longitudinally averaged rotation rate(normalized by the frame rate Ω0) at the outer surface for three ofthe cases rotating at Ω0 = 3Ω plotted versus latitude Dashedlines correspond to the FE cases and solid lines to the FF casesIn the legend ν = κ is given in units of 1012 cm2 sminus1

more turbulent the rotation contrast increases at firstbut then decreases and eventually becomes negative (ieantisolar)

To illustrate exactly where the ldquoextrardquo rotation con-trast in the FF cases is located we plot the rotationrate at the outer surface for three of the cases rotatingat 3Ω in Figure 3 Most of the additional contrast oc-curs at high latitudes where the polar regions in the FFcases rotate significantly more slowly than in their FEcounterparts Additionally the equator in the FF casesrotates slightly faster than in the FE cases For all sim-ulation pairs the difference in contrast between the FEcase and the FF case is greater the smaller the value ofthe diffusion (or the higher the level of turbulence)

Figure 4 shows contour plots of rotation rate in theupper meridional plane for some of the simulation suiteClearly there is a striking difference between the tilts ofthe rotation contours in the FE and FF simulations

In this paper we define all rotation-contour tilt angles(or simply tilts) with respect to the rotation axis withzero tilt corresponding to alignment of the contour withthe rotation axis We use the sign convention for tiltangle illustrated in Figure 5 Under this convention thesolar rotation contours have positive tilts at all latitudesin the bulk of the CZ (above the tachocline and belowthe NSSL as shown in Figure 1) We thus define the

tilt angle of a rotation contour at a given point in themeridional plane as

tilt equiv minus tanminus1[partΩpartz

partΩpartλ

] (15)

which is consistent with the sign convention shown inFigure 5 for solar-like differential rotation in which thecontours further from the rotation axis correspond to ahigher rotation rate

Describing the solar rotation contours as ldquotilted alongradial linesrdquo as is often done is technically misleadingRadial tilt implies a specific dependence of the contourtilt angle with latitude namely contours that fan radi-ally outward from the center of the Sun In Figure 1 bycontrast the bulk-CZ tilts are roughly constant at sim25

for mid-latitudes are smaller than sim25 for low lati-tudes (where radially-aligned tilts would be greater) andare greater than sim25 for high latitudes (where radially-aligned tilts would be smaller) To avoid confusion wewill henceforth not use the term ldquoradial tiltrdquo and insteaddescribe the rotation-contour tilt (in the Sun and in oursimulations) simply as ldquopositiverdquo or ldquonegativerdquo using thesign convention illustrated in Figure 5

In Figure 6 we show the values of the rotation-contourtilt angle at mid-depth for a subset of our models andfor the Sun The positive tilt for the FF cases is obviouswith the maximum tilt angle being about +15 for thehighest value of the diffusion (ν = 4 times 1012 cm2 sminus1)This is still substantially lower than the solar value forcontour tilt which attains a maximum value of sim25 inthe middle of the solar CZ The contours in the FE casesall have positive tilt at low latitudes At high latitudeshowever they have negative tilt and are tilted the most(with a tilt angle of about minus10) for the lowest value ofthe diffusion (ν = 2times 1012 cm2 sminus1)

4 THERMAL WIND BALANCE

We find that to leading order the longitudinally andtemporally averaged force balance in the meridional di-rections r and θ (or λ and z) is dominated by the Coriolispressure and buoyancy forces for each simulation in thiswork

minusnabla(〈P 〉ρ

)+〈S〉cp

g(r)er + 2〈vφ〉eφ timesΩ0 asymp 0 (16)

Here the angular brackets denote a combined temporaland longitudinal average

In the Earthrsquos atmosphere a ldquothermal windrdquo describesa situation in which geostrophic balance (pressure bal-ancing the Coriolis force) holds in the horizontal direc-tions and hydrostatic balance (pressure balancing grav-ity) holds in the vertical direction Equation (16) thusrepresents the generalization of a thermal wind to the so-lar geometry in which the vertical (radial) and horizontalextents of the flow structures are comparable (unlike inthe Earthrsquos atmosphere where the vertical extent is verysmall) Furthermore the flows in the solar geometrygenerally have a vertical component unlike in a classicalthermal wind for which the flows are purely horizontal

The colatitudinal component of Equation (16) may be

6 Matilsky et al

Figure 4 Temporally and longitudinally averaged rotation rate in the meridional plane for some of the simulations in the suite with thetwo hemispheres averaged assuming even symmetry about the equator In the titles at the top ν = κ is given of units of 1012 cm2 sminus1and Ω0 in units of Ω

Figure 5 Schematic for our definition of contour tilts showing(a) positive contour tilt (all contours tilted at a constant +25)(b) zero contour tilt and (c) negative contour tilt (all contourstilted at a constant minus25)

rearranged to yield

Ω(r θ) asymp Ω0 +1

Ω0ρr2 sin 2θ

langpartP

partθ

rang (17)

which is a purely geostrophic equation since the buoy-ancy force is radial Figure 7 shows a representative ex-ample of geostrophic balance for the FE2-3FF2-3 pairClearly Equation (17) is very well satisfied for both caseswith deviations from geostrophy being no more than 1part in 103 in the bulk of the meridional plane and 1part in 102 at isolated regions by the equator and poleThe same is true for all the cases considered in this workindicating that the differential rotation profile in our sim-ulations is almost completely determined by the pressureprofile and vice versa The fact that the differential rota-tion magnitudes are sim40 greater in the FF cases com-pared to the FE cases is thus a consequence of greaterlatitudinal pressure gradients Figure 7 also indicatesthat viscosity plays a relatively insignificant role in theforce balance at large scales

Figure 6 Tilt angle of Equation (15) shown as a function oflatitude for three of the cases rotating at 3Ω and for the SunThe profiles are taken at the middle of the shell for our modelsand the middle of the CZ for the Sun The northern and southernhemispheres have been averaged assuming odd symmetry for tiltangle Dashed lines correspond to the FE cases and solid lines tothe FF cases For the solar tilt angle we use the inversion fromGONG data 1995ndash2004 as reported in Howe et al (2005) andshown in Figure 1

To assess why there are opposite signs of tilt for therotation contours in the FF and FE simulations we dif-ferentiate Equation (17) with respect to the axial coordi-nate z and use the radial component of Equation (16) toeliminate terms (or equivalently take the φ-componentof the curl of Equation (16)) yielding

partΩ

partzasymp g

2Ω0r2 sin θcp

langpartS

partθ

rang (18)

The tilt of the rotation contours is thus determinedby the entropy distribution in the final thermodynamicstate

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

2 Matilsky et al

controlled the magnitude of the contour tilt Imposingthis boundary condition was inspired by thermal-windbalance in the tachocline imprinting up into the CZ

There is also good evidence for thermal-wind balanceoperating in the deep CZ from analytical models Bal-bus (2009a) Balbus et al (2009b) assumed a simplefunctional relationship between the entropy and angularvelocity and solved explicitly for the combined isorota-tional and isentropic contours as characteristics of thethermal-wind equation This yielded tilted contours ingood agreement with helioseismology for the bulk of theCZ (away from the boundary layers of shear at the topand bottom) Such analytical workmdashin addition to lend-ing strong support for the presence of a thermal windin the solar interiormdashpoints to the need to better un-derstand the dynamical coupling between entropy andangular velocity in global numerical simulations some-thing we address in the present paper

In this work we explore the role that the outer ther-mal boundary condition plays in conjunction with aninterior thermal wind in modifying the resulting differ-ential rotation We consider the two most commonlyused options fixed-entropy (FE) in which the entropyat the outer boundary is fixed to a constant value andfixed-flux (FF) in which the radial entropy gradient atthe outer boundary is fixed therefore implying an out-ward conductive flux that is independent of latitude SeeHurle et al (1966) and Edwards (1990) for descriptionsof these boundary conditions in the context of linear the-ory and Anders et al (2020) for a detailed analysisof the boundary conditions in nonlinear simulations ofRayleigh-Benard convection The FF condition is moreappropriate for the Sun since the radiant flux from thesolar photosphere does not appear to vary substantiallywith latitude Rast et al (2008) analyzed full-disk im-ages from the Precision Solar Photometric Telescope atthe Mauna Loa Solar Observatory and calculated non-magnetic contributions to the solar photospheric inten-sity In both continuum and Ca II K intensity distribu-tions only a sim01ndash02 variation was observed corre-sponding to a solar pole that is at most sim25 K warmer(in terms of effective temperature) than the equator Thededuced near-spherical-symmetry of the solar emissiveflux is a significant observation since a thermal windstrong enough to drive the observed differential rotationwould require greater differences in the interior temper-ature between equator and pole We return to this pointin our concluding remarks

We do not address the dynamics of the near-surfacenor tachocline shear layers in this work In particularour models have an impenetrable lower boundary thatdoes not allow for the convective overshoot of downflowplumes into the stable region that may play a role in theorigin of the tachocline The dynamical maintenance ofthe NSSL is still an open question as discussed in Hottaet al (2015) and Matilsky et al (2019) and modelstend to only display signs of near-surface shear if theyhave high density contrast (gtsim100) across the CZ Toavoid high computational cost the models in this workhave a smaller density contrast of sim20

Solar-like differential rotation (fast equator and slowpoles) in spherical-shell convection is thought to be dueto the outward transport of angular momentum by Bussecolumns (also called ldquoTaylor columnsrdquo and ldquobanana

cellsrdquo eg Busse 2002 Jones et al 2011) In thiswork we show that Busse columns also transport heatpoleward and equivalently drive a solar-like differentialrotation through a thermal wind We find that in ourFF cases the thermal wind drives stronger differentialrotation magnitudes and achieves more significant tilt inthe rotation contours than the corresponding FE casesElliot et al (2000) noted this effect for one simulationbut did not explore the underlying mechanism

In Section 2 we describe the parameter space exploredby our simulation suite as well as the mathematical de-tails of the FF and FE boundary conditions In Section3 we describe the basic results of our experiment fo-cusing on the achieved differential rotation In Section4 we quantify the force balance achieved in our modelswhich for the radial and latitudinal directions consistsof a thermal wind in spherical geometry In Section 5we examine the latitudinal transport of energy by Bussecolumns that is responsible for the thermal wind InSection 6 we discuss how the effects of the thermal windare modified by the outer thermal boundary conditionIn Section 7 we discuss our simulation results in the con-text of the Sun

2 NUMERICAL EXPERIMENT

We consider time-dependent 3D simulations of a ro-tating stratified spherical shell of fluid representativeof the solar CZ We use the open-source code Rayleigh091 (Featherstone amp Hindman 2016a Matsui et al2016 Featherstone 2018) which solves the equationsof hydrodynamics in spherical geometry Our domain isa spherical shell with inner radius ri and outer radiusro We describe this domain using the spherical coor-dinates (r θ φ) and corresponding unit vectors (er eθeφ) Here r is the spherical radius θ the colatitude andφ the azimuthal coordinate When convenient we alsouse cylindrical coordinates (λ φ z) = (r sin θ φ r cos θ)and unit vectors (eλ eφ ez) where λ is the cylindricalradius and z the axial coordinate perpendicular to theequatorial plane

The thermodynamic reference state is chosen to betemporally steady and spherically symmetric with adi-abatic stratification (see Jones et al 2011 for a com-plete description) The density varies by a factor of sim20(three density scale heights) across the layer We de-note the pressure density temperature and entropy byP ρ T and S respectively using overbars to indicatethe fixed reference state and the absence of overbars toindicate deviations from the reference state

Rayleigh uses an anelastic approximation (eg Gough1969 Gilman amp Glatzmaier 1981 Jones et al 2011)

which removes sound waves from the system making themaximum allowable timestep much larger since it is lim-ited by the flow velocity rather than the sound speed Ina frame rotating with angular velocity Ω0 = Ω0ez thefluid equations are then given by (eg Featherstone ampHindman 2016a)

nabla middot (ρv) = 0 (1)

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 3

ρ

[partv

partt+ (v middot nabla)v

]=minus 2ρΩ0 times v

minus ρnabla

(P

ρ

)minus ρS

cpg +nabla middotD (2)

and

ρT

[partS

partt+ v middot nablaS

]= nabla middot

[κρTnablaS

]+Q+ D nablav (3)

Here v = (vr vθ vφ) is the fluid velocity in the rotatingframe cp is the constant-pressure specific heat g is thegravitational acceleration due to a solar mass M locatedat the center of the spherical shell D = ρν[nablav +nablavT minus(23)(nablamiddotv)I] is the Newtonian viscous-stress tensor I isthe identity tensor ν is the kinematic viscosity and κ isthe thermal diffusivity Because it is not computationallypossible to resolve convection in the solar regime downto the turbulent microscale ν and κ must be regardedas ldquoeddyrdquo diffusivities which for simplicity we choose tobe spatially constant and equal such that the Prandtlnumber is unity The internal heating function whichphysically represents heating due to radiation is chosento have the fixed radial profile Q(r) = α[P (r) minus P (ro)]with the normalization constant α chosen such that asolar luminosity L is forced through the domain

The heating function Q is designed such that most ofthe energy is deposited in the bottom sim13 of the do-main (see Featherstone amp Hindman 2016a for the radialprofile of the heat flux associated with Q) In the Sunthe internal heating arises from high-opacity metals inthe CZ absorbing the radiation streaming in from theinterior This internal heating is quite distributed andthe heat flux associated with our imposed Q agrees wellwith the radiative flux inferred from standard solar mod-els (for example model S in Christensen-Dalsgaard etal 1996) The distributed nature of the heating makesthe convection problem in the solar CZ slightly differentfrom a standard Rayleigh-Benard setup in which thefluid layer is heated from below by imposing a conduc-tive flux through the lower boundary

The equation of state for the system is that of a perfectgas subject to small thermodynamic perturbations aboutthe reference state

ρ

ρ=P

Pminus T

T=

P

γPminus S

cp (4)

where γ = 53 is the ratio of specific heatsWe adopt stress-free and impenetrable boundary con-

ditions to conserve angular momentum and mass

vr =part

partr

(vθr

)=

part

partr

(vφr

)= 0 at r = ri and ro (5)

In all cases the inner thermal boundary condition al-lows no flux of energy into the system through the lowerboundary

partS

partr= 0 at r = ri (6)

Input parameters common to all the simulations exploredhere are shown in Table 1

Table 1Common input-parameter values for all simulations

ri 500 times1010 cm = 0719 Rro 659 times1010 cm = 0947 Rcp 350 times108 erg Kminus1 gminus1

γ 167ρi 0181 g cmminus3

L 385 times1033 erg sminus1

M 199 times1033 gR 696 times1010 cm

Pr equiv νκ 100

21 Outer thermal boundary condition

The main purpose of this work is to characterize theinfluence of the outer thermal boundary condition on thebehavior of the resulting differential rotation We con-sider models with different background rotation rates Ω0

and diffusion values (ν = κ) and for each model analyzetwo sub-cases

S = 0 at r = ro (fixed entropy or FE) (7)

and

partS

partr= minus L

4πr2oρTκat r = ro (fixed flux or FF) (8)

The solar luminosity that is injected into the systemvia internal heating is ultimately carried out through theouter surface via thermal conduction which in our mod-els arises from entropy gradients (see Equation (3)) Forthe fixed-entropy condition (7) the interior is initiallyheated (leading to S gt 0 in the lower parts of the CZ)while the entropy at the outer surface is ldquopinnedrdquo tozero This naturally establishes a thermal boundary layer(sharp entropy gradients partSpartr lt 0) just below the outersurface The steepness of the gradient (ie the strengthof the outward conductive loss of energy) is allowed tovary with latitude

For the fixed-flux condition (8) the outer thermalboundary layer is present from the beginning of the sim-ulation The steepness of the entropy gradient (and thusthe energy loss) at the outer surface is independent oflatitude by construction and is forced to have exactlythe value needed to carry out a solar luminosity Thefixed-flux condition is thus more ldquosolar-likerdquo since inthe Sun there is no observed latitudinal dependence ofthe emergent intensity which is equal to the energy lostvia radiative cooling at the photosphere

For both the FE and FF cases the thermal conductiveboundary layer stands in contrast to the real solar photo-sphere in which radiative cooling removes the heat froma very thin (sim100 km) outer layer The cooling drivesvery small temporal and spatial scales of motion com-pared to the deep interior (such as granulation and super-granulation) making its direct inclusion in global mod-els problematic Researchers have sought to address thisdifficulty in various ways Nelson et al (2018) imple-mented stochastic driving of convection by near-surfaceplumes designed to mimic the effects of supergranulationfinding that that the flow structures and transport prop-erties were significantly altered in the deep CZ Hotta etal (2019) simulated the whole CZ with no rotation ormagnetic field coupling a global spherical shell that cap-tured large-scale flows in the deep interior to a Cartesian

4 Matilsky et al

box that solved the equations of radiative transfer in thephotosphere They found that the near-surface motionshad a weak influence on the deep interior Regardlessof its relevance to interior flow structures correctly cap-turing the small-scale near-surface flows in global modelsis currently prohibitively expensive computationally Inorder to explore a wider range of parameter space wethus only consider the FE and FF boundary conditionshere

3 SIMULATION RESULTS

We label simulations with a prefix that signifies theouter boundary condition (ldquoFErdquo for Equation (7) andldquoFFrdquo for Equation (8)) followed by the value of the dif-fusion constant ν = κ (in units of 1012 cm2 sminus1) followedby the value of the rotation rate (in units of the siderealCarrington value for the Sun Ω equiv 287times 10minus6 rad sminus1or Ω2π equiv 456 nHz) For example ldquocase FE4-3rdquo refersto a simulation with an FE outer boundary for whichν = κ = 4 times 1012 cm2 sminus1 throughout the domain andΩ0 = 3Ω

Table 2 (in the Appendix) contains the values of thenon-dimensional parameters as well as the grid resolu-tion for each of the 18 simulations considered in thiswork Table 2 has four groupings according to FE or FFat two different rotation rates Following the notationof Featherstone amp Hindman (2016b) we parameterizethe strength of the imposed driving in each simulationthrough a bulk ldquoflux Rayleigh numberrdquo (imposed a pri-ori)

Ra equiv gFH4

cpρT ν3(9)

and the level of turbulence through bulk Reynolds orPeclet numbers (calculated a posteriori)

Re =Pe

Pr=vprimeH

ν (10)

Since the Prandtl number for all models is unity thePeclet number Pe = vprimeHκminus1 equals the Reynolds num-ber

Similarly we parameterize the influence of rotationthrough an Ekman number (imposed a priori)

Ek equiv ν

2Ω0H2(11)

and a bulk Rossby number (calculated a posteriori)

Ro equiv vprime

2Ω0H (12)

In the preceding equations the length scale H is takento be the shell depth ro minus ri the tildes refer to volumeaverages over the full spherical shell and F refers to theenergy flux associated with conduction and convectionin equilibrium (see Featherstone amp Hindman 2016b)The typical convective velocity amplitude vprime refers to therms of the velocity with the longitudinally averaged partsubtracted the mean being taken over time and over thefull volume of the shell Throughout this work temporalaverages are taken during the latter portion of run timefor which there is statistical equilibriummdashgenerally sim34of the total run time listed in Table 2

Before discussing our results in detail we note thatall our models have fairly high levels of thermal and vis-cous diffusion Furthermore all our models rotate at ei-ther two or three times the solar Carrington rate Thesechoices which stand in contrast to the physics of thesolar interior ensure that our models have low enoughRossby numbers to avoid antisolar differential rotation(fast poles slow equator) All global spherical-shell con-vection codes produce high velocities at large scales inthe solar context when sufficiently turbulent The in-fluence of rotation on the large scales is therefore weakwhich causes less coherence in the outward angular mo-mentum transport by the convection and ultimately lessangular momentum in the outer layers than the inner lay-ers (ie an antisolar differential rotation) The overallproblemmdashthat increasing the turbulence in simulationsleads to antisolar statesmdashis now called the ldquoconvectiveconundrumrdquo (OrsquoMara et al 2016) The antisolar statescan be avoided by raising the rotation rate raising thediffusions or lowering the luminosity We choose a com-bination of the former two which requires that our mod-els are only moderately turbulent Nonetheless the vis-cous force and heat flux are small to leading order in theprimary dynamical balances

Returning to our simulation results we quantify themagnitude of the overall differential rotation contrast asthe difference in the outer-surface rotation rate betweenthe equator and 60-latitude normalized by the framerotation rate

∆Ω

Ω0equiv Ω(ro π2)minus Ω(ro π6)

Ω0 (13)

where

Ω(r θ) equiv Ω0 +〈vφ〉r sin θ

(14)

is the rotation rate of the fluid as a function of r andθ and the angular brackets denote a combined temporaland longitudinal average From Table 2 the FF caseshave differential rotation contrasts ∆ΩΩ0 that are sig-nificantly greatermdashon average by sim40mdashthan those ofthe FE cases For comparison the solar value of the ro-tation contrast is substantially higher than in any of ourmodels ∆ΩΩ = 020 (see Howe et al 2000 Fig-ure 1) For the solar estimate we have taken Ω to bethe sidereal Carrington rate and ro to lie just below thenear-surface shear layer

Figure 2 shows how the differential rotation fractionscales with the reduced Rayleigh number which accountsfor the increase to the critical Rayleigh number for con-vective onset caused by rotation (Chandrasekhar 1961)The reduced Rayleigh number thus serves as a better pa-rameterization of the supercriticality of the system thansimply the Rayleigh number From Figure 2 each type ofboundary condition yields a similar scaling with the re-duced Rayleigh number R For the Ω0 = 3Ω cases (cir-cles in Figure 2) the rotation contrast increases mono-tonically (but with decreasing slope) with increasing Rso that the curves connected by circles in Figure 2 areconcave-down For the Ω0 = 2Ω cases (triangles in Fig-ure 2) the curves ldquooverturnrdquo so that a peak value of therotation contrast (at around R sim 27) is achieved Thisbehavior (concisely described in Gastine et al 2013) is asymptom of the convective conundrum as models grow

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 5

Figure 2 Rotation contrast achieved by the suite of simula-tions plotted as a function of the reduced Rayleigh number R equivRa Ek43 Circles mark the Ω0 = 3Ω cases and triangles theΩ = 2Ω cases

Figure 3 Temporally and longitudinally averaged rotation rate(normalized by the frame rate Ω0) at the outer surface for three ofthe cases rotating at Ω0 = 3Ω plotted versus latitude Dashedlines correspond to the FE cases and solid lines to the FF casesIn the legend ν = κ is given in units of 1012 cm2 sminus1

more turbulent the rotation contrast increases at firstbut then decreases and eventually becomes negative (ieantisolar)

To illustrate exactly where the ldquoextrardquo rotation con-trast in the FF cases is located we plot the rotationrate at the outer surface for three of the cases rotatingat 3Ω in Figure 3 Most of the additional contrast oc-curs at high latitudes where the polar regions in the FFcases rotate significantly more slowly than in their FEcounterparts Additionally the equator in the FF casesrotates slightly faster than in the FE cases For all sim-ulation pairs the difference in contrast between the FEcase and the FF case is greater the smaller the value ofthe diffusion (or the higher the level of turbulence)

Figure 4 shows contour plots of rotation rate in theupper meridional plane for some of the simulation suiteClearly there is a striking difference between the tilts ofthe rotation contours in the FE and FF simulations

In this paper we define all rotation-contour tilt angles(or simply tilts) with respect to the rotation axis withzero tilt corresponding to alignment of the contour withthe rotation axis We use the sign convention for tiltangle illustrated in Figure 5 Under this convention thesolar rotation contours have positive tilts at all latitudesin the bulk of the CZ (above the tachocline and belowthe NSSL as shown in Figure 1) We thus define the

tilt angle of a rotation contour at a given point in themeridional plane as

tilt equiv minus tanminus1[partΩpartz

partΩpartλ

] (15)

which is consistent with the sign convention shown inFigure 5 for solar-like differential rotation in which thecontours further from the rotation axis correspond to ahigher rotation rate

Describing the solar rotation contours as ldquotilted alongradial linesrdquo as is often done is technically misleadingRadial tilt implies a specific dependence of the contourtilt angle with latitude namely contours that fan radi-ally outward from the center of the Sun In Figure 1 bycontrast the bulk-CZ tilts are roughly constant at sim25

for mid-latitudes are smaller than sim25 for low lati-tudes (where radially-aligned tilts would be greater) andare greater than sim25 for high latitudes (where radially-aligned tilts would be smaller) To avoid confusion wewill henceforth not use the term ldquoradial tiltrdquo and insteaddescribe the rotation-contour tilt (in the Sun and in oursimulations) simply as ldquopositiverdquo or ldquonegativerdquo using thesign convention illustrated in Figure 5

In Figure 6 we show the values of the rotation-contourtilt angle at mid-depth for a subset of our models andfor the Sun The positive tilt for the FF cases is obviouswith the maximum tilt angle being about +15 for thehighest value of the diffusion (ν = 4 times 1012 cm2 sminus1)This is still substantially lower than the solar value forcontour tilt which attains a maximum value of sim25 inthe middle of the solar CZ The contours in the FE casesall have positive tilt at low latitudes At high latitudeshowever they have negative tilt and are tilted the most(with a tilt angle of about minus10) for the lowest value ofthe diffusion (ν = 2times 1012 cm2 sminus1)

4 THERMAL WIND BALANCE

We find that to leading order the longitudinally andtemporally averaged force balance in the meridional di-rections r and θ (or λ and z) is dominated by the Coriolispressure and buoyancy forces for each simulation in thiswork

minusnabla(〈P 〉ρ

)+〈S〉cp

g(r)er + 2〈vφ〉eφ timesΩ0 asymp 0 (16)

Here the angular brackets denote a combined temporaland longitudinal average

In the Earthrsquos atmosphere a ldquothermal windrdquo describesa situation in which geostrophic balance (pressure bal-ancing the Coriolis force) holds in the horizontal direc-tions and hydrostatic balance (pressure balancing grav-ity) holds in the vertical direction Equation (16) thusrepresents the generalization of a thermal wind to the so-lar geometry in which the vertical (radial) and horizontalextents of the flow structures are comparable (unlike inthe Earthrsquos atmosphere where the vertical extent is verysmall) Furthermore the flows in the solar geometrygenerally have a vertical component unlike in a classicalthermal wind for which the flows are purely horizontal

The colatitudinal component of Equation (16) may be

6 Matilsky et al

Figure 4 Temporally and longitudinally averaged rotation rate in the meridional plane for some of the simulations in the suite with thetwo hemispheres averaged assuming even symmetry about the equator In the titles at the top ν = κ is given of units of 1012 cm2 sminus1and Ω0 in units of Ω

Figure 5 Schematic for our definition of contour tilts showing(a) positive contour tilt (all contours tilted at a constant +25)(b) zero contour tilt and (c) negative contour tilt (all contourstilted at a constant minus25)

rearranged to yield

Ω(r θ) asymp Ω0 +1

Ω0ρr2 sin 2θ

langpartP

partθ

rang (17)

which is a purely geostrophic equation since the buoy-ancy force is radial Figure 7 shows a representative ex-ample of geostrophic balance for the FE2-3FF2-3 pairClearly Equation (17) is very well satisfied for both caseswith deviations from geostrophy being no more than 1part in 103 in the bulk of the meridional plane and 1part in 102 at isolated regions by the equator and poleThe same is true for all the cases considered in this workindicating that the differential rotation profile in our sim-ulations is almost completely determined by the pressureprofile and vice versa The fact that the differential rota-tion magnitudes are sim40 greater in the FF cases com-pared to the FE cases is thus a consequence of greaterlatitudinal pressure gradients Figure 7 also indicatesthat viscosity plays a relatively insignificant role in theforce balance at large scales

Figure 6 Tilt angle of Equation (15) shown as a function oflatitude for three of the cases rotating at 3Ω and for the SunThe profiles are taken at the middle of the shell for our modelsand the middle of the CZ for the Sun The northern and southernhemispheres have been averaged assuming odd symmetry for tiltangle Dashed lines correspond to the FE cases and solid lines tothe FF cases For the solar tilt angle we use the inversion fromGONG data 1995ndash2004 as reported in Howe et al (2005) andshown in Figure 1

To assess why there are opposite signs of tilt for therotation contours in the FF and FE simulations we dif-ferentiate Equation (17) with respect to the axial coordi-nate z and use the radial component of Equation (16) toeliminate terms (or equivalently take the φ-componentof the curl of Equation (16)) yielding

partΩ

partzasymp g

2Ω0r2 sin θcp

langpartS

partθ

rang (18)

The tilt of the rotation contours is thus determinedby the entropy distribution in the final thermodynamicstate

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 3

ρ

[partv

partt+ (v middot nabla)v

]=minus 2ρΩ0 times v

minus ρnabla

(P

ρ

)minus ρS

cpg +nabla middotD (2)

and

ρT

[partS

partt+ v middot nablaS

]= nabla middot

[κρTnablaS

]+Q+ D nablav (3)

Here v = (vr vθ vφ) is the fluid velocity in the rotatingframe cp is the constant-pressure specific heat g is thegravitational acceleration due to a solar mass M locatedat the center of the spherical shell D = ρν[nablav +nablavT minus(23)(nablamiddotv)I] is the Newtonian viscous-stress tensor I isthe identity tensor ν is the kinematic viscosity and κ isthe thermal diffusivity Because it is not computationallypossible to resolve convection in the solar regime downto the turbulent microscale ν and κ must be regardedas ldquoeddyrdquo diffusivities which for simplicity we choose tobe spatially constant and equal such that the Prandtlnumber is unity The internal heating function whichphysically represents heating due to radiation is chosento have the fixed radial profile Q(r) = α[P (r) minus P (ro)]with the normalization constant α chosen such that asolar luminosity L is forced through the domain

The heating function Q is designed such that most ofthe energy is deposited in the bottom sim13 of the do-main (see Featherstone amp Hindman 2016a for the radialprofile of the heat flux associated with Q) In the Sunthe internal heating arises from high-opacity metals inthe CZ absorbing the radiation streaming in from theinterior This internal heating is quite distributed andthe heat flux associated with our imposed Q agrees wellwith the radiative flux inferred from standard solar mod-els (for example model S in Christensen-Dalsgaard etal 1996) The distributed nature of the heating makesthe convection problem in the solar CZ slightly differentfrom a standard Rayleigh-Benard setup in which thefluid layer is heated from below by imposing a conduc-tive flux through the lower boundary

The equation of state for the system is that of a perfectgas subject to small thermodynamic perturbations aboutthe reference state

ρ

ρ=P

Pminus T

T=

P

γPminus S

cp (4)

where γ = 53 is the ratio of specific heatsWe adopt stress-free and impenetrable boundary con-

ditions to conserve angular momentum and mass

vr =part

partr

(vθr

)=

part

partr

(vφr

)= 0 at r = ri and ro (5)

In all cases the inner thermal boundary condition al-lows no flux of energy into the system through the lowerboundary

partS

partr= 0 at r = ri (6)

Input parameters common to all the simulations exploredhere are shown in Table 1

Table 1Common input-parameter values for all simulations

ri 500 times1010 cm = 0719 Rro 659 times1010 cm = 0947 Rcp 350 times108 erg Kminus1 gminus1

γ 167ρi 0181 g cmminus3

L 385 times1033 erg sminus1

M 199 times1033 gR 696 times1010 cm

Pr equiv νκ 100

21 Outer thermal boundary condition

The main purpose of this work is to characterize theinfluence of the outer thermal boundary condition on thebehavior of the resulting differential rotation We con-sider models with different background rotation rates Ω0

and diffusion values (ν = κ) and for each model analyzetwo sub-cases

S = 0 at r = ro (fixed entropy or FE) (7)

and

partS

partr= minus L

4πr2oρTκat r = ro (fixed flux or FF) (8)

The solar luminosity that is injected into the systemvia internal heating is ultimately carried out through theouter surface via thermal conduction which in our mod-els arises from entropy gradients (see Equation (3)) Forthe fixed-entropy condition (7) the interior is initiallyheated (leading to S gt 0 in the lower parts of the CZ)while the entropy at the outer surface is ldquopinnedrdquo tozero This naturally establishes a thermal boundary layer(sharp entropy gradients partSpartr lt 0) just below the outersurface The steepness of the gradient (ie the strengthof the outward conductive loss of energy) is allowed tovary with latitude

For the fixed-flux condition (8) the outer thermalboundary layer is present from the beginning of the sim-ulation The steepness of the entropy gradient (and thusthe energy loss) at the outer surface is independent oflatitude by construction and is forced to have exactlythe value needed to carry out a solar luminosity Thefixed-flux condition is thus more ldquosolar-likerdquo since inthe Sun there is no observed latitudinal dependence ofthe emergent intensity which is equal to the energy lostvia radiative cooling at the photosphere

For both the FE and FF cases the thermal conductiveboundary layer stands in contrast to the real solar photo-sphere in which radiative cooling removes the heat froma very thin (sim100 km) outer layer The cooling drivesvery small temporal and spatial scales of motion com-pared to the deep interior (such as granulation and super-granulation) making its direct inclusion in global mod-els problematic Researchers have sought to address thisdifficulty in various ways Nelson et al (2018) imple-mented stochastic driving of convection by near-surfaceplumes designed to mimic the effects of supergranulationfinding that that the flow structures and transport prop-erties were significantly altered in the deep CZ Hotta etal (2019) simulated the whole CZ with no rotation ormagnetic field coupling a global spherical shell that cap-tured large-scale flows in the deep interior to a Cartesian

4 Matilsky et al

box that solved the equations of radiative transfer in thephotosphere They found that the near-surface motionshad a weak influence on the deep interior Regardlessof its relevance to interior flow structures correctly cap-turing the small-scale near-surface flows in global modelsis currently prohibitively expensive computationally Inorder to explore a wider range of parameter space wethus only consider the FE and FF boundary conditionshere

3 SIMULATION RESULTS

We label simulations with a prefix that signifies theouter boundary condition (ldquoFErdquo for Equation (7) andldquoFFrdquo for Equation (8)) followed by the value of the dif-fusion constant ν = κ (in units of 1012 cm2 sminus1) followedby the value of the rotation rate (in units of the siderealCarrington value for the Sun Ω equiv 287times 10minus6 rad sminus1or Ω2π equiv 456 nHz) For example ldquocase FE4-3rdquo refersto a simulation with an FE outer boundary for whichν = κ = 4 times 1012 cm2 sminus1 throughout the domain andΩ0 = 3Ω

Table 2 (in the Appendix) contains the values of thenon-dimensional parameters as well as the grid resolu-tion for each of the 18 simulations considered in thiswork Table 2 has four groupings according to FE or FFat two different rotation rates Following the notationof Featherstone amp Hindman (2016b) we parameterizethe strength of the imposed driving in each simulationthrough a bulk ldquoflux Rayleigh numberrdquo (imposed a pri-ori)

Ra equiv gFH4

cpρT ν3(9)

and the level of turbulence through bulk Reynolds orPeclet numbers (calculated a posteriori)

Re =Pe

Pr=vprimeH

ν (10)

Since the Prandtl number for all models is unity thePeclet number Pe = vprimeHκminus1 equals the Reynolds num-ber

Similarly we parameterize the influence of rotationthrough an Ekman number (imposed a priori)

Ek equiv ν

2Ω0H2(11)

and a bulk Rossby number (calculated a posteriori)

Ro equiv vprime

2Ω0H (12)

In the preceding equations the length scale H is takento be the shell depth ro minus ri the tildes refer to volumeaverages over the full spherical shell and F refers to theenergy flux associated with conduction and convectionin equilibrium (see Featherstone amp Hindman 2016b)The typical convective velocity amplitude vprime refers to therms of the velocity with the longitudinally averaged partsubtracted the mean being taken over time and over thefull volume of the shell Throughout this work temporalaverages are taken during the latter portion of run timefor which there is statistical equilibriummdashgenerally sim34of the total run time listed in Table 2

Before discussing our results in detail we note thatall our models have fairly high levels of thermal and vis-cous diffusion Furthermore all our models rotate at ei-ther two or three times the solar Carrington rate Thesechoices which stand in contrast to the physics of thesolar interior ensure that our models have low enoughRossby numbers to avoid antisolar differential rotation(fast poles slow equator) All global spherical-shell con-vection codes produce high velocities at large scales inthe solar context when sufficiently turbulent The in-fluence of rotation on the large scales is therefore weakwhich causes less coherence in the outward angular mo-mentum transport by the convection and ultimately lessangular momentum in the outer layers than the inner lay-ers (ie an antisolar differential rotation) The overallproblemmdashthat increasing the turbulence in simulationsleads to antisolar statesmdashis now called the ldquoconvectiveconundrumrdquo (OrsquoMara et al 2016) The antisolar statescan be avoided by raising the rotation rate raising thediffusions or lowering the luminosity We choose a com-bination of the former two which requires that our mod-els are only moderately turbulent Nonetheless the vis-cous force and heat flux are small to leading order in theprimary dynamical balances

Returning to our simulation results we quantify themagnitude of the overall differential rotation contrast asthe difference in the outer-surface rotation rate betweenthe equator and 60-latitude normalized by the framerotation rate

∆Ω

Ω0equiv Ω(ro π2)minus Ω(ro π6)

Ω0 (13)

where

Ω(r θ) equiv Ω0 +〈vφ〉r sin θ

(14)

is the rotation rate of the fluid as a function of r andθ and the angular brackets denote a combined temporaland longitudinal average From Table 2 the FF caseshave differential rotation contrasts ∆ΩΩ0 that are sig-nificantly greatermdashon average by sim40mdashthan those ofthe FE cases For comparison the solar value of the ro-tation contrast is substantially higher than in any of ourmodels ∆ΩΩ = 020 (see Howe et al 2000 Fig-ure 1) For the solar estimate we have taken Ω to bethe sidereal Carrington rate and ro to lie just below thenear-surface shear layer

Figure 2 shows how the differential rotation fractionscales with the reduced Rayleigh number which accountsfor the increase to the critical Rayleigh number for con-vective onset caused by rotation (Chandrasekhar 1961)The reduced Rayleigh number thus serves as a better pa-rameterization of the supercriticality of the system thansimply the Rayleigh number From Figure 2 each type ofboundary condition yields a similar scaling with the re-duced Rayleigh number R For the Ω0 = 3Ω cases (cir-cles in Figure 2) the rotation contrast increases mono-tonically (but with decreasing slope) with increasing Rso that the curves connected by circles in Figure 2 areconcave-down For the Ω0 = 2Ω cases (triangles in Fig-ure 2) the curves ldquooverturnrdquo so that a peak value of therotation contrast (at around R sim 27) is achieved Thisbehavior (concisely described in Gastine et al 2013) is asymptom of the convective conundrum as models grow

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 5

Figure 2 Rotation contrast achieved by the suite of simula-tions plotted as a function of the reduced Rayleigh number R equivRa Ek43 Circles mark the Ω0 = 3Ω cases and triangles theΩ = 2Ω cases

Figure 3 Temporally and longitudinally averaged rotation rate(normalized by the frame rate Ω0) at the outer surface for three ofthe cases rotating at Ω0 = 3Ω plotted versus latitude Dashedlines correspond to the FE cases and solid lines to the FF casesIn the legend ν = κ is given in units of 1012 cm2 sminus1

more turbulent the rotation contrast increases at firstbut then decreases and eventually becomes negative (ieantisolar)

To illustrate exactly where the ldquoextrardquo rotation con-trast in the FF cases is located we plot the rotationrate at the outer surface for three of the cases rotatingat 3Ω in Figure 3 Most of the additional contrast oc-curs at high latitudes where the polar regions in the FFcases rotate significantly more slowly than in their FEcounterparts Additionally the equator in the FF casesrotates slightly faster than in the FE cases For all sim-ulation pairs the difference in contrast between the FEcase and the FF case is greater the smaller the value ofthe diffusion (or the higher the level of turbulence)

Figure 4 shows contour plots of rotation rate in theupper meridional plane for some of the simulation suiteClearly there is a striking difference between the tilts ofthe rotation contours in the FE and FF simulations

In this paper we define all rotation-contour tilt angles(or simply tilts) with respect to the rotation axis withzero tilt corresponding to alignment of the contour withthe rotation axis We use the sign convention for tiltangle illustrated in Figure 5 Under this convention thesolar rotation contours have positive tilts at all latitudesin the bulk of the CZ (above the tachocline and belowthe NSSL as shown in Figure 1) We thus define the

tilt angle of a rotation contour at a given point in themeridional plane as

tilt equiv minus tanminus1[partΩpartz

partΩpartλ

] (15)

which is consistent with the sign convention shown inFigure 5 for solar-like differential rotation in which thecontours further from the rotation axis correspond to ahigher rotation rate

Describing the solar rotation contours as ldquotilted alongradial linesrdquo as is often done is technically misleadingRadial tilt implies a specific dependence of the contourtilt angle with latitude namely contours that fan radi-ally outward from the center of the Sun In Figure 1 bycontrast the bulk-CZ tilts are roughly constant at sim25

for mid-latitudes are smaller than sim25 for low lati-tudes (where radially-aligned tilts would be greater) andare greater than sim25 for high latitudes (where radially-aligned tilts would be smaller) To avoid confusion wewill henceforth not use the term ldquoradial tiltrdquo and insteaddescribe the rotation-contour tilt (in the Sun and in oursimulations) simply as ldquopositiverdquo or ldquonegativerdquo using thesign convention illustrated in Figure 5

In Figure 6 we show the values of the rotation-contourtilt angle at mid-depth for a subset of our models andfor the Sun The positive tilt for the FF cases is obviouswith the maximum tilt angle being about +15 for thehighest value of the diffusion (ν = 4 times 1012 cm2 sminus1)This is still substantially lower than the solar value forcontour tilt which attains a maximum value of sim25 inthe middle of the solar CZ The contours in the FE casesall have positive tilt at low latitudes At high latitudeshowever they have negative tilt and are tilted the most(with a tilt angle of about minus10) for the lowest value ofthe diffusion (ν = 2times 1012 cm2 sminus1)

4 THERMAL WIND BALANCE

We find that to leading order the longitudinally andtemporally averaged force balance in the meridional di-rections r and θ (or λ and z) is dominated by the Coriolispressure and buoyancy forces for each simulation in thiswork

minusnabla(〈P 〉ρ

)+〈S〉cp

g(r)er + 2〈vφ〉eφ timesΩ0 asymp 0 (16)

Here the angular brackets denote a combined temporaland longitudinal average

In the Earthrsquos atmosphere a ldquothermal windrdquo describesa situation in which geostrophic balance (pressure bal-ancing the Coriolis force) holds in the horizontal direc-tions and hydrostatic balance (pressure balancing grav-ity) holds in the vertical direction Equation (16) thusrepresents the generalization of a thermal wind to the so-lar geometry in which the vertical (radial) and horizontalextents of the flow structures are comparable (unlike inthe Earthrsquos atmosphere where the vertical extent is verysmall) Furthermore the flows in the solar geometrygenerally have a vertical component unlike in a classicalthermal wind for which the flows are purely horizontal

The colatitudinal component of Equation (16) may be

6 Matilsky et al

Figure 4 Temporally and longitudinally averaged rotation rate in the meridional plane for some of the simulations in the suite with thetwo hemispheres averaged assuming even symmetry about the equator In the titles at the top ν = κ is given of units of 1012 cm2 sminus1and Ω0 in units of Ω

Figure 5 Schematic for our definition of contour tilts showing(a) positive contour tilt (all contours tilted at a constant +25)(b) zero contour tilt and (c) negative contour tilt (all contourstilted at a constant minus25)

rearranged to yield

Ω(r θ) asymp Ω0 +1

Ω0ρr2 sin 2θ

langpartP

partθ

rang (17)

which is a purely geostrophic equation since the buoy-ancy force is radial Figure 7 shows a representative ex-ample of geostrophic balance for the FE2-3FF2-3 pairClearly Equation (17) is very well satisfied for both caseswith deviations from geostrophy being no more than 1part in 103 in the bulk of the meridional plane and 1part in 102 at isolated regions by the equator and poleThe same is true for all the cases considered in this workindicating that the differential rotation profile in our sim-ulations is almost completely determined by the pressureprofile and vice versa The fact that the differential rota-tion magnitudes are sim40 greater in the FF cases com-pared to the FE cases is thus a consequence of greaterlatitudinal pressure gradients Figure 7 also indicatesthat viscosity plays a relatively insignificant role in theforce balance at large scales

Figure 6 Tilt angle of Equation (15) shown as a function oflatitude for three of the cases rotating at 3Ω and for the SunThe profiles are taken at the middle of the shell for our modelsand the middle of the CZ for the Sun The northern and southernhemispheres have been averaged assuming odd symmetry for tiltangle Dashed lines correspond to the FE cases and solid lines tothe FF cases For the solar tilt angle we use the inversion fromGONG data 1995ndash2004 as reported in Howe et al (2005) andshown in Figure 1

To assess why there are opposite signs of tilt for therotation contours in the FF and FE simulations we dif-ferentiate Equation (17) with respect to the axial coordi-nate z and use the radial component of Equation (16) toeliminate terms (or equivalently take the φ-componentof the curl of Equation (16)) yielding

partΩ

partzasymp g

2Ω0r2 sin θcp

langpartS

partθ

rang (18)

The tilt of the rotation contours is thus determinedby the entropy distribution in the final thermodynamicstate

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

4 Matilsky et al

box that solved the equations of radiative transfer in thephotosphere They found that the near-surface motionshad a weak influence on the deep interior Regardlessof its relevance to interior flow structures correctly cap-turing the small-scale near-surface flows in global modelsis currently prohibitively expensive computationally Inorder to explore a wider range of parameter space wethus only consider the FE and FF boundary conditionshere

3 SIMULATION RESULTS

We label simulations with a prefix that signifies theouter boundary condition (ldquoFErdquo for Equation (7) andldquoFFrdquo for Equation (8)) followed by the value of the dif-fusion constant ν = κ (in units of 1012 cm2 sminus1) followedby the value of the rotation rate (in units of the siderealCarrington value for the Sun Ω equiv 287times 10minus6 rad sminus1or Ω2π equiv 456 nHz) For example ldquocase FE4-3rdquo refersto a simulation with an FE outer boundary for whichν = κ = 4 times 1012 cm2 sminus1 throughout the domain andΩ0 = 3Ω

Table 2 (in the Appendix) contains the values of thenon-dimensional parameters as well as the grid resolu-tion for each of the 18 simulations considered in thiswork Table 2 has four groupings according to FE or FFat two different rotation rates Following the notationof Featherstone amp Hindman (2016b) we parameterizethe strength of the imposed driving in each simulationthrough a bulk ldquoflux Rayleigh numberrdquo (imposed a pri-ori)

Ra equiv gFH4

cpρT ν3(9)

and the level of turbulence through bulk Reynolds orPeclet numbers (calculated a posteriori)

Re =Pe

Pr=vprimeH

ν (10)

Since the Prandtl number for all models is unity thePeclet number Pe = vprimeHκminus1 equals the Reynolds num-ber

Similarly we parameterize the influence of rotationthrough an Ekman number (imposed a priori)

Ek equiv ν

2Ω0H2(11)

and a bulk Rossby number (calculated a posteriori)

Ro equiv vprime

2Ω0H (12)

In the preceding equations the length scale H is takento be the shell depth ro minus ri the tildes refer to volumeaverages over the full spherical shell and F refers to theenergy flux associated with conduction and convectionin equilibrium (see Featherstone amp Hindman 2016b)The typical convective velocity amplitude vprime refers to therms of the velocity with the longitudinally averaged partsubtracted the mean being taken over time and over thefull volume of the shell Throughout this work temporalaverages are taken during the latter portion of run timefor which there is statistical equilibriummdashgenerally sim34of the total run time listed in Table 2

Before discussing our results in detail we note thatall our models have fairly high levels of thermal and vis-cous diffusion Furthermore all our models rotate at ei-ther two or three times the solar Carrington rate Thesechoices which stand in contrast to the physics of thesolar interior ensure that our models have low enoughRossby numbers to avoid antisolar differential rotation(fast poles slow equator) All global spherical-shell con-vection codes produce high velocities at large scales inthe solar context when sufficiently turbulent The in-fluence of rotation on the large scales is therefore weakwhich causes less coherence in the outward angular mo-mentum transport by the convection and ultimately lessangular momentum in the outer layers than the inner lay-ers (ie an antisolar differential rotation) The overallproblemmdashthat increasing the turbulence in simulationsleads to antisolar statesmdashis now called the ldquoconvectiveconundrumrdquo (OrsquoMara et al 2016) The antisolar statescan be avoided by raising the rotation rate raising thediffusions or lowering the luminosity We choose a com-bination of the former two which requires that our mod-els are only moderately turbulent Nonetheless the vis-cous force and heat flux are small to leading order in theprimary dynamical balances

Returning to our simulation results we quantify themagnitude of the overall differential rotation contrast asthe difference in the outer-surface rotation rate betweenthe equator and 60-latitude normalized by the framerotation rate

∆Ω

Ω0equiv Ω(ro π2)minus Ω(ro π6)

Ω0 (13)

where

Ω(r θ) equiv Ω0 +〈vφ〉r sin θ

(14)

is the rotation rate of the fluid as a function of r andθ and the angular brackets denote a combined temporaland longitudinal average From Table 2 the FF caseshave differential rotation contrasts ∆ΩΩ0 that are sig-nificantly greatermdashon average by sim40mdashthan those ofthe FE cases For comparison the solar value of the ro-tation contrast is substantially higher than in any of ourmodels ∆ΩΩ = 020 (see Howe et al 2000 Fig-ure 1) For the solar estimate we have taken Ω to bethe sidereal Carrington rate and ro to lie just below thenear-surface shear layer

Figure 2 shows how the differential rotation fractionscales with the reduced Rayleigh number which accountsfor the increase to the critical Rayleigh number for con-vective onset caused by rotation (Chandrasekhar 1961)The reduced Rayleigh number thus serves as a better pa-rameterization of the supercriticality of the system thansimply the Rayleigh number From Figure 2 each type ofboundary condition yields a similar scaling with the re-duced Rayleigh number R For the Ω0 = 3Ω cases (cir-cles in Figure 2) the rotation contrast increases mono-tonically (but with decreasing slope) with increasing Rso that the curves connected by circles in Figure 2 areconcave-down For the Ω0 = 2Ω cases (triangles in Fig-ure 2) the curves ldquooverturnrdquo so that a peak value of therotation contrast (at around R sim 27) is achieved Thisbehavior (concisely described in Gastine et al 2013) is asymptom of the convective conundrum as models grow

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 5

Figure 2 Rotation contrast achieved by the suite of simula-tions plotted as a function of the reduced Rayleigh number R equivRa Ek43 Circles mark the Ω0 = 3Ω cases and triangles theΩ = 2Ω cases

Figure 3 Temporally and longitudinally averaged rotation rate(normalized by the frame rate Ω0) at the outer surface for three ofthe cases rotating at Ω0 = 3Ω plotted versus latitude Dashedlines correspond to the FE cases and solid lines to the FF casesIn the legend ν = κ is given in units of 1012 cm2 sminus1

more turbulent the rotation contrast increases at firstbut then decreases and eventually becomes negative (ieantisolar)

To illustrate exactly where the ldquoextrardquo rotation con-trast in the FF cases is located we plot the rotationrate at the outer surface for three of the cases rotatingat 3Ω in Figure 3 Most of the additional contrast oc-curs at high latitudes where the polar regions in the FFcases rotate significantly more slowly than in their FEcounterparts Additionally the equator in the FF casesrotates slightly faster than in the FE cases For all sim-ulation pairs the difference in contrast between the FEcase and the FF case is greater the smaller the value ofthe diffusion (or the higher the level of turbulence)

Figure 4 shows contour plots of rotation rate in theupper meridional plane for some of the simulation suiteClearly there is a striking difference between the tilts ofthe rotation contours in the FE and FF simulations

In this paper we define all rotation-contour tilt angles(or simply tilts) with respect to the rotation axis withzero tilt corresponding to alignment of the contour withthe rotation axis We use the sign convention for tiltangle illustrated in Figure 5 Under this convention thesolar rotation contours have positive tilts at all latitudesin the bulk of the CZ (above the tachocline and belowthe NSSL as shown in Figure 1) We thus define the

tilt angle of a rotation contour at a given point in themeridional plane as

tilt equiv minus tanminus1[partΩpartz

partΩpartλ

] (15)

which is consistent with the sign convention shown inFigure 5 for solar-like differential rotation in which thecontours further from the rotation axis correspond to ahigher rotation rate

Describing the solar rotation contours as ldquotilted alongradial linesrdquo as is often done is technically misleadingRadial tilt implies a specific dependence of the contourtilt angle with latitude namely contours that fan radi-ally outward from the center of the Sun In Figure 1 bycontrast the bulk-CZ tilts are roughly constant at sim25

for mid-latitudes are smaller than sim25 for low lati-tudes (where radially-aligned tilts would be greater) andare greater than sim25 for high latitudes (where radially-aligned tilts would be smaller) To avoid confusion wewill henceforth not use the term ldquoradial tiltrdquo and insteaddescribe the rotation-contour tilt (in the Sun and in oursimulations) simply as ldquopositiverdquo or ldquonegativerdquo using thesign convention illustrated in Figure 5

In Figure 6 we show the values of the rotation-contourtilt angle at mid-depth for a subset of our models andfor the Sun The positive tilt for the FF cases is obviouswith the maximum tilt angle being about +15 for thehighest value of the diffusion (ν = 4 times 1012 cm2 sminus1)This is still substantially lower than the solar value forcontour tilt which attains a maximum value of sim25 inthe middle of the solar CZ The contours in the FE casesall have positive tilt at low latitudes At high latitudeshowever they have negative tilt and are tilted the most(with a tilt angle of about minus10) for the lowest value ofthe diffusion (ν = 2times 1012 cm2 sminus1)

4 THERMAL WIND BALANCE

We find that to leading order the longitudinally andtemporally averaged force balance in the meridional di-rections r and θ (or λ and z) is dominated by the Coriolispressure and buoyancy forces for each simulation in thiswork

minusnabla(〈P 〉ρ

)+〈S〉cp

g(r)er + 2〈vφ〉eφ timesΩ0 asymp 0 (16)

Here the angular brackets denote a combined temporaland longitudinal average

In the Earthrsquos atmosphere a ldquothermal windrdquo describesa situation in which geostrophic balance (pressure bal-ancing the Coriolis force) holds in the horizontal direc-tions and hydrostatic balance (pressure balancing grav-ity) holds in the vertical direction Equation (16) thusrepresents the generalization of a thermal wind to the so-lar geometry in which the vertical (radial) and horizontalextents of the flow structures are comparable (unlike inthe Earthrsquos atmosphere where the vertical extent is verysmall) Furthermore the flows in the solar geometrygenerally have a vertical component unlike in a classicalthermal wind for which the flows are purely horizontal

The colatitudinal component of Equation (16) may be

6 Matilsky et al

Figure 4 Temporally and longitudinally averaged rotation rate in the meridional plane for some of the simulations in the suite with thetwo hemispheres averaged assuming even symmetry about the equator In the titles at the top ν = κ is given of units of 1012 cm2 sminus1and Ω0 in units of Ω

Figure 5 Schematic for our definition of contour tilts showing(a) positive contour tilt (all contours tilted at a constant +25)(b) zero contour tilt and (c) negative contour tilt (all contourstilted at a constant minus25)

rearranged to yield

Ω(r θ) asymp Ω0 +1

Ω0ρr2 sin 2θ

langpartP

partθ

rang (17)

which is a purely geostrophic equation since the buoy-ancy force is radial Figure 7 shows a representative ex-ample of geostrophic balance for the FE2-3FF2-3 pairClearly Equation (17) is very well satisfied for both caseswith deviations from geostrophy being no more than 1part in 103 in the bulk of the meridional plane and 1part in 102 at isolated regions by the equator and poleThe same is true for all the cases considered in this workindicating that the differential rotation profile in our sim-ulations is almost completely determined by the pressureprofile and vice versa The fact that the differential rota-tion magnitudes are sim40 greater in the FF cases com-pared to the FE cases is thus a consequence of greaterlatitudinal pressure gradients Figure 7 also indicatesthat viscosity plays a relatively insignificant role in theforce balance at large scales

Figure 6 Tilt angle of Equation (15) shown as a function oflatitude for three of the cases rotating at 3Ω and for the SunThe profiles are taken at the middle of the shell for our modelsand the middle of the CZ for the Sun The northern and southernhemispheres have been averaged assuming odd symmetry for tiltangle Dashed lines correspond to the FE cases and solid lines tothe FF cases For the solar tilt angle we use the inversion fromGONG data 1995ndash2004 as reported in Howe et al (2005) andshown in Figure 1

To assess why there are opposite signs of tilt for therotation contours in the FF and FE simulations we dif-ferentiate Equation (17) with respect to the axial coordi-nate z and use the radial component of Equation (16) toeliminate terms (or equivalently take the φ-componentof the curl of Equation (16)) yielding

partΩ

partzasymp g

2Ω0r2 sin θcp

langpartS

partθ

rang (18)

The tilt of the rotation contours is thus determinedby the entropy distribution in the final thermodynamicstate

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 5

Figure 2 Rotation contrast achieved by the suite of simula-tions plotted as a function of the reduced Rayleigh number R equivRa Ek43 Circles mark the Ω0 = 3Ω cases and triangles theΩ = 2Ω cases

Figure 3 Temporally and longitudinally averaged rotation rate(normalized by the frame rate Ω0) at the outer surface for three ofthe cases rotating at Ω0 = 3Ω plotted versus latitude Dashedlines correspond to the FE cases and solid lines to the FF casesIn the legend ν = κ is given in units of 1012 cm2 sminus1

more turbulent the rotation contrast increases at firstbut then decreases and eventually becomes negative (ieantisolar)

To illustrate exactly where the ldquoextrardquo rotation con-trast in the FF cases is located we plot the rotationrate at the outer surface for three of the cases rotatingat 3Ω in Figure 3 Most of the additional contrast oc-curs at high latitudes where the polar regions in the FFcases rotate significantly more slowly than in their FEcounterparts Additionally the equator in the FF casesrotates slightly faster than in the FE cases For all sim-ulation pairs the difference in contrast between the FEcase and the FF case is greater the smaller the value ofthe diffusion (or the higher the level of turbulence)

Figure 4 shows contour plots of rotation rate in theupper meridional plane for some of the simulation suiteClearly there is a striking difference between the tilts ofthe rotation contours in the FE and FF simulations

In this paper we define all rotation-contour tilt angles(or simply tilts) with respect to the rotation axis withzero tilt corresponding to alignment of the contour withthe rotation axis We use the sign convention for tiltangle illustrated in Figure 5 Under this convention thesolar rotation contours have positive tilts at all latitudesin the bulk of the CZ (above the tachocline and belowthe NSSL as shown in Figure 1) We thus define the

tilt angle of a rotation contour at a given point in themeridional plane as

tilt equiv minus tanminus1[partΩpartz

partΩpartλ

] (15)

which is consistent with the sign convention shown inFigure 5 for solar-like differential rotation in which thecontours further from the rotation axis correspond to ahigher rotation rate

Describing the solar rotation contours as ldquotilted alongradial linesrdquo as is often done is technically misleadingRadial tilt implies a specific dependence of the contourtilt angle with latitude namely contours that fan radi-ally outward from the center of the Sun In Figure 1 bycontrast the bulk-CZ tilts are roughly constant at sim25

for mid-latitudes are smaller than sim25 for low lati-tudes (where radially-aligned tilts would be greater) andare greater than sim25 for high latitudes (where radially-aligned tilts would be smaller) To avoid confusion wewill henceforth not use the term ldquoradial tiltrdquo and insteaddescribe the rotation-contour tilt (in the Sun and in oursimulations) simply as ldquopositiverdquo or ldquonegativerdquo using thesign convention illustrated in Figure 5

In Figure 6 we show the values of the rotation-contourtilt angle at mid-depth for a subset of our models andfor the Sun The positive tilt for the FF cases is obviouswith the maximum tilt angle being about +15 for thehighest value of the diffusion (ν = 4 times 1012 cm2 sminus1)This is still substantially lower than the solar value forcontour tilt which attains a maximum value of sim25 inthe middle of the solar CZ The contours in the FE casesall have positive tilt at low latitudes At high latitudeshowever they have negative tilt and are tilted the most(with a tilt angle of about minus10) for the lowest value ofthe diffusion (ν = 2times 1012 cm2 sminus1)

4 THERMAL WIND BALANCE

We find that to leading order the longitudinally andtemporally averaged force balance in the meridional di-rections r and θ (or λ and z) is dominated by the Coriolispressure and buoyancy forces for each simulation in thiswork

minusnabla(〈P 〉ρ

)+〈S〉cp

g(r)er + 2〈vφ〉eφ timesΩ0 asymp 0 (16)

Here the angular brackets denote a combined temporaland longitudinal average

In the Earthrsquos atmosphere a ldquothermal windrdquo describesa situation in which geostrophic balance (pressure bal-ancing the Coriolis force) holds in the horizontal direc-tions and hydrostatic balance (pressure balancing grav-ity) holds in the vertical direction Equation (16) thusrepresents the generalization of a thermal wind to the so-lar geometry in which the vertical (radial) and horizontalextents of the flow structures are comparable (unlike inthe Earthrsquos atmosphere where the vertical extent is verysmall) Furthermore the flows in the solar geometrygenerally have a vertical component unlike in a classicalthermal wind for which the flows are purely horizontal

The colatitudinal component of Equation (16) may be

6 Matilsky et al

Figure 4 Temporally and longitudinally averaged rotation rate in the meridional plane for some of the simulations in the suite with thetwo hemispheres averaged assuming even symmetry about the equator In the titles at the top ν = κ is given of units of 1012 cm2 sminus1and Ω0 in units of Ω

Figure 5 Schematic for our definition of contour tilts showing(a) positive contour tilt (all contours tilted at a constant +25)(b) zero contour tilt and (c) negative contour tilt (all contourstilted at a constant minus25)

rearranged to yield

Ω(r θ) asymp Ω0 +1

Ω0ρr2 sin 2θ

langpartP

partθ

rang (17)

which is a purely geostrophic equation since the buoy-ancy force is radial Figure 7 shows a representative ex-ample of geostrophic balance for the FE2-3FF2-3 pairClearly Equation (17) is very well satisfied for both caseswith deviations from geostrophy being no more than 1part in 103 in the bulk of the meridional plane and 1part in 102 at isolated regions by the equator and poleThe same is true for all the cases considered in this workindicating that the differential rotation profile in our sim-ulations is almost completely determined by the pressureprofile and vice versa The fact that the differential rota-tion magnitudes are sim40 greater in the FF cases com-pared to the FE cases is thus a consequence of greaterlatitudinal pressure gradients Figure 7 also indicatesthat viscosity plays a relatively insignificant role in theforce balance at large scales

Figure 6 Tilt angle of Equation (15) shown as a function oflatitude for three of the cases rotating at 3Ω and for the SunThe profiles are taken at the middle of the shell for our modelsand the middle of the CZ for the Sun The northern and southernhemispheres have been averaged assuming odd symmetry for tiltangle Dashed lines correspond to the FE cases and solid lines tothe FF cases For the solar tilt angle we use the inversion fromGONG data 1995ndash2004 as reported in Howe et al (2005) andshown in Figure 1

To assess why there are opposite signs of tilt for therotation contours in the FF and FE simulations we dif-ferentiate Equation (17) with respect to the axial coordi-nate z and use the radial component of Equation (16) toeliminate terms (or equivalently take the φ-componentof the curl of Equation (16)) yielding

partΩ

partzasymp g

2Ω0r2 sin θcp

langpartS

partθ

rang (18)

The tilt of the rotation contours is thus determinedby the entropy distribution in the final thermodynamicstate

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

6 Matilsky et al

Figure 4 Temporally and longitudinally averaged rotation rate in the meridional plane for some of the simulations in the suite with thetwo hemispheres averaged assuming even symmetry about the equator In the titles at the top ν = κ is given of units of 1012 cm2 sminus1and Ω0 in units of Ω

Figure 5 Schematic for our definition of contour tilts showing(a) positive contour tilt (all contours tilted at a constant +25)(b) zero contour tilt and (c) negative contour tilt (all contourstilted at a constant minus25)

rearranged to yield

Ω(r θ) asymp Ω0 +1

Ω0ρr2 sin 2θ

langpartP

partθ

rang (17)

which is a purely geostrophic equation since the buoy-ancy force is radial Figure 7 shows a representative ex-ample of geostrophic balance for the FE2-3FF2-3 pairClearly Equation (17) is very well satisfied for both caseswith deviations from geostrophy being no more than 1part in 103 in the bulk of the meridional plane and 1part in 102 at isolated regions by the equator and poleThe same is true for all the cases considered in this workindicating that the differential rotation profile in our sim-ulations is almost completely determined by the pressureprofile and vice versa The fact that the differential rota-tion magnitudes are sim40 greater in the FF cases com-pared to the FE cases is thus a consequence of greaterlatitudinal pressure gradients Figure 7 also indicatesthat viscosity plays a relatively insignificant role in theforce balance at large scales

Figure 6 Tilt angle of Equation (15) shown as a function oflatitude for three of the cases rotating at 3Ω and for the SunThe profiles are taken at the middle of the shell for our modelsand the middle of the CZ for the Sun The northern and southernhemispheres have been averaged assuming odd symmetry for tiltangle Dashed lines correspond to the FE cases and solid lines tothe FF cases For the solar tilt angle we use the inversion fromGONG data 1995ndash2004 as reported in Howe et al (2005) andshown in Figure 1

To assess why there are opposite signs of tilt for therotation contours in the FF and FE simulations we dif-ferentiate Equation (17) with respect to the axial coordi-nate z and use the radial component of Equation (16) toeliminate terms (or equivalently take the φ-componentof the curl of Equation (16)) yielding

partΩ

partzasymp g

2Ω0r2 sin θcp

langpartS

partθ

rang (18)

The tilt of the rotation contours is thus determinedby the entropy distribution in the final thermodynamicstate

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 7

Figure 7 Point-by-point colatitudinal force balance in the merid-ional plane for representative pair (a) case FE2-3 and (b) caseFF2-3 We show the difference between the temporally and longi-tudinally averaged rotation rate Ω and the RHS of Equation (17)which we denote by ΩP (rotation rate from the pressure) normal-ized by the frame rotation rate Ω0 The two hemispheres have beenaveraged assuming even symmetry about the equator

Figure 8 Temporally and longitudinally averaged entropy pres-sure and temperature deviations from the spherically symmetricmean in the meridional half-plane (averaged assuming even sym-metry about the equator) for cases FE2-3 and FF2-3 normalizedby the reference state profiles The spherical mean 〈middot middot middot 〉sph hasbeen removed to show the variation from equator to pole

In Figure 8 we show the average profiles for entropypressure and temperature in the meridional half-planefor the FE2-3FF2-3 pair Case FF2-3 (which is rep-resentative of all the FF cases in the simulation suite)displays a monotonically increasing entropy from equa-tor to pole Case FE2-3 on the other hand has a non-monotonic entropy profile except on the outer boundarythe entropy from equator to pole increases up to sim20

latitude then decreases In both cases the pressure andtemperature deviations (normalized by the backgroundreference state) are substantially greater (by a factor ofsim30 in the case of the pressure) than the entropy devi-ation The profiles of pressure and temperature in themeridional plane thus tend to mirror one another withhigh temperature regions corresponding to high pressureregions and vice versa (compare the last two columns ofFigure 8)

The balance described by Equation (18) is shown forthe representative simulation pair FE2-3FF2-3 in Fig-ure 9 There is good balance in the deep layers althoughsignificant departures near the outer surface which has

Figure 9 Temporally and longitudinally averaged azimuthal vor-ticity balance in the meridional plane for representative pair FE2-3and FF2-3 The two hemispheres have been averaged assuming oddsymmetry about the equator The left-hand column (a c) showsthe axial derivative of the rotation rate partΩpartz The right-handcolumn (b d) shows the difference between partΩpartz and the RHS ofEquation (18) ((partΩpartz)S or the axial derivative of rotation ratefrom the entropy) normalized by 15times 10minus17 sminus1 cmminus1

been noted frequently in past work (eg Brun et al2011 Augustson et al 2012 Hotta et al 2015) Quan-titatively the error in Equation (18) (shown in the right-hand column of Figure 9) is sim10 in the lower 80 of thelayer and sim50 in the upper 20 of the layer For solar-like differential rotation (fast equator and slow poles)positively-tilted rotation contours (the FF cases) corre-spond to partΩpartz lt 0 which arises from 〈partSpartθ〉 lt 0 atall latitudes as in Figure 8(d) Similarly the FE cases(which have contours tilted negatively at high-latitudesand positively at low latitudes) all have 〈partSpartθ〉 gt 0 athigh latitudes and 〈partSpartθ〉 lt 0 at low latitudes as inFigure 8(a)

5 POLEWARD ENERGY TRANSPORT FROM BUSSECOLUMNS

In light of Equations (17) and (18) a thermal windin spherical geometry fundamentally consists of pres-sure and entropy differences in latitude Poles that arehigh-pressure and high-entropy relative to lower latitudes(which we have shown lead to strong differential rota-tion with positively-tilted contours) are expected to beestablished by the preferentially poleward transport ofenergy In our simulations this transport arises fromthe action of the convective Busse-column rolls Theserolls manifest at convective onset as an overstable low-frequency prograde wave (eg Unno et al 1989) oras it is called in the geophysics literature a thermalRossby wave This wave consists of a series of convec-tive rolls or Busse columns that gird the equator Each

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

8 Matilsky et al

roll is rotationally-aligned and the sign of the vorticityalternates from roll to roll Furthermore each roll is ingeostrophic balance hence the alternating sign of thevorticity corresponds to every other roll being a zone ofhigh pressure with low-pressure rolls in between Sincethe ends of the columns (at mid-latitudes) have neutralpressure the pressure anomalies at the equator causepoleward axial flow in the high-pressure rolls and equa-torward flow in the low-pressure rolls (eg Figure 1 inGilman 1983) The resulting strong correlation betweenpressure and the direction of axial flow leads to a netpoleward enthalpy flux through pressure work

The effect just described is easiest to illustrate for mod-els that are barely supercritical Here the profiles forthe velocity and thermodynamic variables are dominatedby the wavenumber associated with the most unstablemode For the range of Ekman numbers spanned by oursimulation suite the resulting Busse columns are mostlylocalized in the outer half of the shell by radius and atlow latitudes (see Jones et al 2009 for a linear sta-bility analysis of the problem) Figures 10(a b) showthe instantaneous convective radial velocity and convec-tive colatitudinal energy transport in the highly diffusivecase FE10-3 which lies in the barely supercritical regimeEach upflow and downflow (pairs of which trace oneBusse column roll) has an associated colatitudinal en-ergy transport that is on average negative in the north-ern hemisphere and positive in the southern hemisphereimplying preferentially poleward energy transport Notethat under the spherical-coordinate convention with θ asthe colatitude the positive-θ direction is always orientednorth-to-south Thus equatorward (poleward) transportof energy corresponds to positive (negative) colatitudi-nal transport in the northern hemisphere and negative(positive) transport in the southern hemisphere

Figures 10(c d) show the radial velocity and convec-tive energy transport in the comparatively more turbu-lent case FF2-3 The flow structures are more intricateand fine-scale than in the barely supercritical regime butthe imprint of the most unstable mode remains ManyBusse column rollsmdashwhich can be seen at low latitudesas columnar red and blue features in Figure 10(c)mdashcorrespond to sites of negative colatitudinal energy trans-port (blue in Figure 10(d)) in the northern hemisphereand positive transport (red in Figure 10(d)) in the south-ern hemisphere Overall there are more sites of polewardenergy transport (from the Busse columns) than sites ofequatorward transport in each hemisphere Under an az-imuthal average the Busse columns in the more turbu-lent case FF2-3 thus yield preferentially poleward energytransport just as in the barely supercritical case FE10-3

The geostrophic nature of the Busse columns is illus-trated in Figure 11 as is the resulting axial componentof the flow In the top row (case FE10-3) panel a showsthat the Busse-column rolls alternate between high andlow pressure Panel b shows that the high-pressure rollsare each anticyclonic (have negative vorticity) while thelow-pressure rolls are cyclonic Finally panel c showsthat each high-pressure anomaly corresponds to pole-ward flow (vz gt 0 in the northern hemisphere) whileeach low-pressure anomaly corresponds to equatorwardflow (vz lt 0) In the bottom row (the more supercriticalcase FF4-3) the Busse columns are less regularly spacedbut still largely alternate between anticyclonic regions of

high pressure and cyclonic regions of low pressure (panelsd e) The axial flow associated with the Busse columnsin case FF4-3 (panel f ) then leads to poleward energytransport through pressure work just as in case FE10-3

It has long been known that Busse columns transportangular momentum outward We have just shown thatBusse columns also transport heat poleward The Bussecolumns thus define a purely hydrodynamic mechanismcoupling entropy and angular velocity Balbus et al(2009b) posited the presence of such a convective mech-anism in the Sun and further argued that the motionsresponsible should fall along surfaces of both constantentropy and constant angular velocity In that picturethe isorotational and isentropic contours should thus co-incide The Busse columns in our simulation suite donot completely behave in this way as evidenced by noneof our simulations having good alignment of the isorota-tional and isentropic contours Independent of whetherthe constant-entropy and constant-angular-velocity sur-faces coincide in the Sun a key point from our work isthat the Busse columns provide an explicit convectivemechanism to couple entropy and angular velocity

6 EFFECT OF OUTER THERMAL BOUNDARYCONDITION

Given that Busse columns direct energy polewardequilibrium can be achieved by forming conductive gradi-ents that balance the poleward convective enthalpy fluxIn general such conductive transport can be achievedin two distinct ways As the pole heats up and theequator-to-pole contrast increases a latitudinal gradientwill form that transports heat equatorward Addition-ally the increased temperature of the pole can lead toenhancement of the radial gradients in the outer ther-mal boundary layer thus causing the poles to lose heatmore efficiently (ie become superluminous) In the FFcases the outer thermal boundary condition precludesthe second of these options because the radial gradientsare fixed Hence the FF models must rely solely on thedevelopment of a pole-to-equator conductive gradientIn the FE models both types of gradients are possibleTherefore the amount that the pole must be heated be-fore equilibrium can be achieved is less for the FE modelsthan it is for the FF models The outer thermal boundarycondition thus has a direct influence on the latitudinalcontrast in the temperature entropy and pressure withthe FF boundary condition being conducive to strongcontrast in all the thermodynamic variables In the pres-ence of thermal-wind balance the FF boundary condi-tion thus leads to enhanced contrast in the differentialrotation and positively-tilted isocontours in the rotationrate

Mathematically we illustrate the combined effects ofthe outer thermal boundary condition and latitudinal en-ergy transport using the steady-state total energy equa-tion for the fluid Using Equations (1)ndash(3) this equationtakes the form of a balance of fluxes

nabla middotF = 0 (19)

where

F equiv Fconv + Fcond + F rad + Fvisc + Fcirc (20)

is the temporally and longitudinally averaged total en-ergy flux in the meridional plane and we have defined the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 9

Figure 10 Spherical snapshots (at rro = 0910) of the radial velocity vprimer (lefthand panels) and colatitudinal convective energy flux

vprimeθ[TSprime + P prime + (12)ρ(vprime)2] (righthand panels) shown in global Mollweide projections The three terms in the convective energy fluxcorrespond to advection of heat pressure work and advection of kinetic energy respectively The top row of panels is taken from thehighly diffusive barely supercritical case FE10-3 and the bottom row of panels is taken from case FF2-3 In both cases the Busse-columnsites in the lefthand panels are associated with sites of poleward convective energy flux in the righthand panelsmdashie panels (b) and (d)are on average blue in the north and red in the south

averaged convective conductive radiative viscous andmeridional-circulation fluxes through

Fconv equiv ρ(cp〈T primevprime〉+

1

2〈v2v〉

) (21a)

Fcond equiv minusκρT 〈nablaS〉 (21b)

F rad equiv

(1

r2

int ro

r

Q(x)x2dx

)er (21c)

Fvisc equiv minus〈D middot v〉 (21d)

and Fcirc equiv ρcp〈T 〉〈v〉 (21e)

respectively Note that ρTS + P = cpT so the termswith 〈T primevprime〉 and 〈T 〉〈v〉 in the convective and meridional-circulation fluxes represent the combined effects of heatadvection and pressure work Technically the flux dueto transport of kinetic energy (proportional to 〈v2v〉)has convective terms (eg 〈(vprime)2vprime〉) and meridional-circulation terms (eg 〈v〉2〈v〉) For simplicity we in-clude all the kinetic-energy terms in the convective fluxsince they are in general small

We are interested in the total latitudinal transport ofenergy and so we integrate the total flux in Equation

(20) over conical surfaces at constant latitude

Iθ(θ) equiv 2π sin θ

int ro

ri

Fθ(r θ)rdr (22)

For the FF cases there can be no net transport of en-ergy in latitude due to the absence of conductive losses inthe polar regions through the outer boundary In otherwords Iθ(θ) equiv 0 For the FE cases by contrast thereis a net poleward energy transport because the poles areallowed to be superluminous Thus Iθ(θ) will in gen-eral be negative in the northern hemisphere and positivein the southern hemisphere Recall that equatorward(poleward) transport of energy corresponds to positive(negative) Iθ(θ) in the northern hemisphere and nega-tive (positive) Iθ(θ) in the southern hemisphere

Figure 12(a) shows the integrated colatitudinal energyflux in case FF2-3 after the system has achieved statis-tical equilibrium The total flux is very close to zero atall latitudes indicating a well-equilibrated state Thedominant transport components are the convective fluxwhich transports energy preferentially poleward due tothe Busse columns and the conductive flux which trans-ports energy equatorward The monotonic entropy gra-dient of Figure 8(d) and by extension radially tilted con-tours in the FF cases is thus seen to be a result of the

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

10 Matilsky et al

Figure 11 Geostrophic balance in Busse columns for cases FE10-3 (top row) and FF4-3 (bottom row) Here we use the standardCartesian coordinates x y and z (a and d) Snapshots of the non-axisymmetric pressure (P prime = P minus 〈P 〉) shown as 3D volume renderingswith the view from slightly north of the equator and the scene cut by the plane zro = 02 (b and e) Closeup views of P prime (shown in color)and ρvprimeh equiv ρ(vprimexex + vprimey ey) (shown as a vector field) in a portion of the plane zro = 02 (c and f ) Closeup views of ρvprimez and ρvprimeh in the

same portion of the plane as in (b and e) In all panels red tones indicate positive values and blue tones indicate negative values

response by conduction to the convective transport ofenergy to the poles

Figure 12(b) shows the integrated colatitudinal en-ergy transport in case FE2-3 The poles are clearlysuperluminousmdashie there is a net poleward energytransport due to the convection For all the FE casesexplored here the energy loss at the poles is even greaterthan the heating by the convection the conductive fluxis thus forced to change sign at mid-latitudes (aroundplusmn20) transporting energy poleward in concert with theBusse columns This results in the non-monotonic en-tropy profile of Figure 8(a) leading to the tilts of therotation contours being negative at high latitudes andpositive at low latitudes as in the top row of Figure 4

7 DISCUSSION AND CONCLUSIONS

We have shown that the differential rotation achievedin global 3D convection simulations is well-described bya thermal wind and highly sensitive to the outer ther-mal boundary condition The FF boundary tends toyield more solar-like rotation profiles (strong contrastwith positively-tilted contours) while the FE boundaryyields weaker contrast and negatively-tilted contours Inlight of these results we now discuss the likelihood thatthe Sunrsquos strong rotation contrast and positively-tiltedcontours arise from thermal-wind balance in the deepinterior coupled with the observation that the radiativeflux from the solar photosphere does not vary apprecia-bly with latitude

The first question is whether the force balance Equa-tion (16) which should in general be true for low Rossbynumbers holds in the Sun The interior solar Rossbynumber is currently unknown but recent helioseismicestimates (Hanasoge et al 2012 Greer 2015) give

Ro 01 in the deep interior Thus it is likely thatEquation (16) (and the derivative thermal wind Equa-tions (17) and (18)) apply in the solar CZ except perhapsin the layers just below the photosphere Thermal-windbalance in the Sun is also not prohibited by observationalresults We can derive the temperature in the solar CZfrom the rotation rate of Figure 1 assuming thermal-windbalance holds integrating Equations (17) and (18) to getP and S in the meridional plane and using Equation (4)to get T In this calculation we set Ω0 to the siderealCarrington rate and use the same polytropic referencestate employed in our models for the solar profiles Theresult is shown in Figure 13 The equator-to-pole tem-perature contrast required to drive the solar-like differ-ential rotation is sim10 K (approximately uniform withradius) which is well below the detection limit of helio-seismology (eg Brun et al 2010)

The second question is whether the Sunrsquos Bussecolumns send energy preferentially poleward In gen-eral stellar convection transitions through a series ofconvective regimes as the supercriticality (measured bythe reduced Rayleigh number R) increases (Hindman etal 2020) Both the least supercritical case in our work(FE10-3 for which R sim 2) and the most supercriticalcases (the pair FF2-2 and FE2-2 for which R sim 50)have a strong preference for poleward transport by Bussecolumns suggesting that the poleward transport is a fea-ture of the most unstable mode of convection that stayspresent as the flows get ever more complex

Finally it is an open question how the Sun might trans-port energy equatorward to maintain equilibrium Inour simulations the net poleward transport of energy byBusse columns is at its maximum a few per cent of the

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 11

Figure 12 Total colatitudinal energy transport for (a) case FF2-3 and (b) case FE2-3 or integrals of the fluxes in Equation (21)over conical surfaces at constant latitude The integrated fluxesare plotted as functions of latitude (= π2 minus θ) Contributionsfrom the various fluxes are indicated in the legend ldquototrdquo denotingthe sum of all the other fluxes Positive (negative) Iθ indicatesequatorward (poleward) transport in the northern hemisphere andpoleward (equatorward) transport in the southern hemisphere

Figure 13 Temperature deviation from the spherical mean in thesolar CZ assuming thermal-wind balance holds The temperaturehas been calculated from the rotation rate inferred from helioseis-mology (Figure 1) using Equations (17) (18) and(4)

Figure 14 Net poleward transport of heat from Busse columnsfor all simulations in the suite (except the barely supercritical casesFE10-3 and FF10-3) plotted versus the reduced Rayleigh number

R equiv Ra Ek43 The poleward heat flux |Iθ conv| has been aver-aged over latitude Circles mark the Ω0 = 3Ω cases and trianglesthe Ω = 2Ω cases

solar luminosity which is counteracted almost entirelyby conduction in the FF cases (Figure 12) In the Sunthe thermal diffusivity associated with radiative heat-ing is sim107 cm2 sminus1 at mid-CZ (eg Hindman et al2020) which (if a thermal wind were operating with atemperature contrast of sim10 K) would correspond to alatitudinal energy flux of sim10minus7L Figure 14 showshow the Busse-column heat flux scales with supercrit-icality The trends are different between the FE andFF cases (and the Ω0 = 3Ω and Ω0 = 2Ω cases)but there is clearly a tendency for the net transport togo down for our more supercritical FF simulations (bluecurves) This indicates that the balance between conduc-tive and convective heat flux could hold in the Sun justwith much smaller flux magnitudes We admit that theseresults are only suggestive since the flux in all our mod-els is still orders of magnitude higher than the presumedsolar value of 10minus7L

On a more practical note it is advantageous to use anFF outer boundary condition in solar simulations for tworeasons First maintaining a strong differential rotationis particularly relevant for dynamo models since the dy-namo cannot be maintained without sufficient shear (egBrown et al 2010 Guerrero et al 2016 Matilsky ampToomre 2020 Bice amp Toomre 2020) Second usingan FE outer boundary condition leads to superluminouspoles which are directly at odds with solar observationsFigure 15 shows the conductive flux as a function of lat-itude at the top of the domain for the FE cases Forcase FE2-3 the flux in the polar regions reaches a valuein excess of the solar luminosity by about 20 This isfar greater than the observationally-constrained value oflt 1 for the Sun (Rast et al 2008)

We very much view this paper as a complement to Mi-esch et al (2006) In that work a systematic tilt of therotation contours was achieved by imposing a small lati-tudinal entropy gradient at the inner boundary therebyensuring that the entropy increased monotonically fromequator to pole And indeed for all our FF cases thereis a similar monotonic equator-to-pole entropy gradientat the inner boundary Monotonicity is not achieved inthe FE cases In other words Miesch et al (2006)showed that it is possible to tilt the rotation contoursby imposing a preferred geostrophic balance and here

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

12 Matilsky et al

Figure 15 Latitudinal profile of Fcond r at the outer boundaryfor several of the FE cases normalized by the flux needed to carryout the solar luminosity equally at all latitudes The flux has beenaveraged over time and longitude

we have shown how this preferred balance is naturallyestablished as the result of poleward energy transport byBusse columns and the FF outer boundary condition

We conclude that any successful dynamical model ofthe Sunrsquos convection zone must be consistent with threekey observations (1) the Sun possesses strong latitu-dinal differential rotation (2) the isorotation contoursare tilted positively in the sense of Figure 5 and (3)the emergent intensity is nearly spherically symmetricWe have demonstrated in this paper that the transportof angular momentum and heat by Busse columns canyield results consistent with observations (1) and (2)Busse columns transport angular momentum outwardaway from the rotation axis thus speeding up the equa-tor They simultaneously transport heat poleward viathe enthalpy flux associated with pressure work Thesetwo transports reach equilibrium once thermal-wind bal-ance is achieved

While it is satisfying to address the observations (1)and (2) with a single mechanism (ie the action of Bussecolumns) observation (3) presents a theoretical problemThe thermal-wind balance resulting from Busse columnsrequires that the solar poles be at a higher temperaturethan the equator with a contrast of sim10 K throughoutthe CZ The emergent intensity at the solar surface how-ever is nearly uniform in latitude with less than sim25K variation in effective temperature This suggests thatthe thermal-wind balance must only be maintained inthe deep CZ and the surface layers (perhaps within thenear-surface shear layer) may have different dynamicsThose dynamics must somehow screen the surface fromthe imposition of the latitudinal temperature gradientfrom below

In our models the proper proxy for the emergent in-tensity is the radial conductive flux at the outer surfaceFor the FF cases we impose spherical symmetry of thisflux as a boundary condition Further we have shownthat this boundary condition is conducive to achievingstrong differential rotation and properly tilted rotationalisocontours However a fully self-consistent model of thesolar interior would reproduce the spherical symmetryat the outer boundary as a natural consequence of the

near-surface convection and not as an enforced boundarycondition

Since the underlying dynamics giving rise tolatitudinally-independent emissivity in the Sun are notwell-understood we cannot predict a priori whether thatsame mechnism would be sufficient to enforce spheri-cal symmetry for other stars It has long been postu-lated that oblateness induced by rotation in high-massstars (in which a radiation layer overlies an interior CZ)could result in superluminous poles (eg von Zeipel1924 Collins 1963) Here we have identified Bussecolumns as a possible mechanism for producing hot polesin low-mass stars (in which the CZ overlies the radia-tion layer) that does not depend on the starrsquos distor-tion Furthermore Busse columns do not necessarilyrequire rapid rotationmdashthey simply require strong ro-tational constraint or in other words low Rossby num-ber Thus small dim stars that are expected to haveweak convection may have active Busse columns and hotpoles even with only moderate rotation Finally sincethe homogeneity (or lack thereof) of the stellar emissiveflux may significantly affect the interior differential rota-tion we should not assume that the positive-tilts of therotational isocontours inferred for the Sun are a generalfeature of other low-mass stars

APPENDIX DIAGNOSTIC PARAMETERS FOR THESIMULATION SUITE

Table 2 contains various input and diagnostic param-eters for the entire simulation suite as defined in Sec-tion 3 Recall that for each simulation name ldquoFErdquo orldquoFFrdquo refers to the type of outer thermal boundary condi-tion employed the first number gives the diffusion valueν = κ in units of 1012 cm2 sminus1 and the second numberafter the dash gives the frame rotation rate Ω0 in unitsof Ω equiv 287times 10minus6 rad sminus1 or Ω2π = 456 nHz NrNθ and Nφ refer to the number of radial colatitudinaland longitudinal grid points respectively The run timeis given in units of the thermal diffusion time defined tobe H2κ with H the shell depth

We thank Mark Miesch for elucidating how pole-ward energy transport might arise from Busse columnsand Rachel Howe for providing invaluable helioseis-mic data We thank the anonymous reviewer forproviding detailed feedback that resulted in a signif-icantly improved manuscript Loren Matilsky waspartly supported during this work by the Future In-vestigators in NASA Earth and Space Sciences Tech-nology (FINESST) award 80NSSC19K1428 and bya George Ellery Hale Graduate Fellowship Thisresearch was primarily supported by NASA Helio-physics through grants NNX13AG18G NNX17AG22G80NSSC18K1127 80NSSC17K0008 80NSSC18K1125and 80NSSC19K0267 Computational resources wereprovided by the NASA High-End Computing (HEC)Program through the NASA Advanced Supercomputing(NAS) Division at Ames Research Center Rayleigh hasbeen developed by Nicholas Featherstone with supportby the National Science Foundation through the Com-putational Infrastructure for Geodynamics (CIG) underNSF grants NSF-0949446 and NSF-1550901

REFERENCES

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665

Revisiting the Sunrsquos Strong Differential Rotation along Radial Lines 13

Table 2Fluid diagnostic parameters for the different simulations

Name (Nr Nθ Nφ) Ra Re = Pe Ek Ro R equiv Ra Ek43 ∆ΩΩ0 Run time

FE2-3 (128 384 768) 852times 105 632 462times 10minus4 338times 10minus2 304 0097 236FE23-3 (128 384 768) 560times 105 518 531times 10minus4 321times 10minus2 241 0092 777FE3-3 (96 384 768) 252times 105 352 693times 10minus4 290times 10minus2 155 0078 596FE4-3 (128 192 384) 107times 105 219 923times 10minus4 255times 10minus2 962 0062 334FE10-3 (64 96 192) 682times 103 254 231times 10minus3 745times 10minus3 208 00048 2385

FE2-2 (128 576 1152) 852times 105 756 693times 10minus4 591times 10minus2 522 0082 778FE23-2 (128 576 1152) 560times 105 622 796times 10minus4 561times 10minus2 413 0088 914FE3-2 (96 384 768) 252times 105 428 104times 10minus3 509times 10minus2 266 0092 119FE4-2 (64 288 576) 107times 105 280 139times 10minus3 451times 10minus2 166 0086 477

FF2-3 (128 576 1152) 852times 105 645 462times 10minus4 345times 10minus2 304 0140 147FF23-3 (128 384 768) 560times 105 529 531times 10minus4 328times 10minus2 241 0131 790FF3-3 (96 384 768) 252times 105 358 693times 10minus4 295times 10minus2 155 0108 144FF4-3 (128 192 384) 107times 105 224 923times 10minus4 246times 10minus2 962 0080 255FF10-3 (64 96 192) 682times 103 289 231times 10minus3 858times 10minus3 208 00024 159

FF2-2 (128 576 1152) 852times 105 812 693times 10minus4 633times 10minus2 522 0110 668FF23-2 (128 576 1152) 560times 105 653 796times 10minus4 589times 10minus2 413 0121 827FF3-2 (96 384 768) 252times 105 444 104times 10minus3 530times 10minus2 266 0128 103FF4-2 (128 384 768) 107times 105 289 139times 10minus3 466times 10minus2 166 0117 704

Note mdash The 18 simulations here are in four groupings according to FE or FF at two different rotation rates

Anders EH Vasil GM Brown BP amp Korre L 2020PhRvF submitted arXiv200300026

Augustson KC Brown BP Brun AS Miesch MS ampToomre J 2012 ApJ 756 169

Balbus SA 2009 MNRAS 395 2056Balbus SA Bonart J Latter HN amp Weiss NO 2009

MNRAS 400 176Bice CP amp Toomre J 2020 ApJ 893 107Brown BP Browning MK Brun AS Miesch MS amp

Toomre J 2010 ApJ 711 424Brun AS amp Toomre J 2002 ApJ 570 865Brun AS Antia HM Chitre SM 2010 ApJ 510 A33Brun AS Miesch MS amp Toomre J 2002 ApJ 742 79Busse FH 2002 Phys Fluids 14 4Chandrasekhar S 1961 Hydrodynamic and Hydromagnetic

Stability (Oxford Clarendon)Christensen-Dalsgaard J Dappen W Ajukov SV et al 1996

Sci 272 1286Collins GW II 1963 Astrophysical Journal ApJ 138 1134Edwards JM 1990 GAFD 55 1Elliot JR Miesch MS amp Toomre J 2000 ApJ 533 546Featherstone NA amp Hindman BW 2016a ApJ 818 (1) 38Featherstone NA amp Hindman BW 2016b ApJL 830 L15Featherstone N 2018 Rayleigh 091 doi

httpdoiorg105281zenodo1236565Gastine T Wicht J amp Aurnou JM 2013 Icarus 225 156Gilman PA amp Glatzmaier GA 1981 ApJS 45 335Gilman PA 1983 ApJ 53 243Greer BJ Hindman BW Featherstone NA amp Toomre J

2015 ApJL 803 L17Gough DO 1969 J Atmos Sci 26 448

Guerrero G Smolarkiewicz EW de Gouveia Dal Pino EMKosovichev AG amp Mansour NN 2016 ApJL 828 L3

Hanasoge SM Duvall TL Jr amp Sreenivasan KR 2012PNAS 109 11928

Hindman BW Featherstone NA amp Julien K 2020 ApJsubmitted

Hotta H Rempel M amp Yokoyama T 2015 ApJ 798 51Hotta H Iijimi H amp Kusano K 2019 Sci Adv 5 eaau2307Howe R Christensen-Dalsgaard J Hill F et al 2000 Sci 287

2456Howe R Christensen-Dalsgaard J Hill F et al 2005 ApJ

634 1405Hurle DTJ Jakeman E amp Pike ER 1966 Proc R Soc

Lond A 296 469Jones CA Kuzanyan KM amp Mitchell RH 2009 J Fl

Mech 634 291Jones CA Boronski P Brun AS et al 2011 Icarus 216 120OrsquoMara B Miesch MS Featherstone NA amp Augustson KC

2016 AdSpR 58 1475

Matilsky LI Hindman BW amp Toomre J 2019 ApJ 871 217Matilsky LI amp Toomre J 2020 ApJ 892 106Matsui H Heien E Aubert J Aurnou JM Avery M et al

2016 Geochem Geophys 17 1586Miesch MS Brun AS amp Toomre J 2006 ApJ 641 618Nelson NJ Featherstone NA Miesch MS amp Toomre J

2018 ApJ 859 117Rast MP Ortiz A amp Meisner RW 2008 ApJ 673 1209Thompson M J Christensen-Dalsgaard J Miesch M S amp

Toomre J 2003 ARAampA 41 599Unno W Osaki Y Ando H et al 1989 Nonradial

Oscillations of Stars 2nd ed (University of Tokyo Press)von Zeipel H 1924 MNRAS 84 665