Lowfrequency internal waves in magnetized rotating stellar radiation zones
I. Wave structure modification by a toroidal field
^{1}
Laboratoire AIM, CEA/DSM – CNRS – Université Paris Diderot,
IRFU/SAp Centre de Saclay,
91191
GifsurYvette,
France
email: stephane.mathis@cea.fr, nicolas.debrye@obspm.fr
^{2}
LESIA, Observatoire de Paris – CNRS – Université Paris Diderot,
Place Jules Janssen,
92195
Meudon,
France
Received:
11
August
2010
Accepted:
23
September
2010
Context. The study of helioseismology, asteroseismology, and powerful groundbased instrumentation dedicated to stellar physics is developing strongly (cf. CoRoT, KEPLER, and ESPaDOnS). This generates tight constraints on the stellar internal structure and dynamical processes. In this context, it is thus necessary to go beyond the nonrotating and the nonmagnetic picture of stellar interiors, particularly for largescale transport mechanisms and waves.
Aims. We focus on lowfrequency internal waves in magnetic, rotating, stably stratified stellar radiation zones. For frequencies, which can be close to the Alfvén and the inertial frequencies, we go beyond the nonmagnetic and nonrotating description of wave dynamics with taking the Coriolis acceleration and the Lorentz force into account. Then, we have to couple wave dynamics with fossil magnetic fields, which must have mixed configurations (both poloidal and toroidal) to survive in stellar radiation zones.
Methods. We chose to study such coupling step by step, first with purely toroidal fields and then with purely poloidal fields, to unravel their modification by each corresponding component of a realistic mixedfield. Thus, we analytically built a complete formalism, which describes both effects of the Coriolis acceleration and of the Lorentz force in a nonperturbative way in the case of an axisymmetric toroidal field. We consider here the case where both Alfvén frequency and angular velocity are chosen to be uniform, to isolate wave properties.
Results. The different approximations possible for lowfrequency internal waves in this model are examined and discussed. In this way, the traditional approximation used to describe the dynamics of lowfrequency regular elliptic gravitoinertial waves in the purely hydrodynamical case is generalized to the magnetic one. The complete structure of internal waves, which become magnetogravitoinertial waves, is then derived and compared to the nonmagnetic case. The asymptotic behaviour of such waves is obtained.
Conclusions. A global study of magnetogravitoinertial waves in stellar radiation zones is achieved in the case of an axisymmetric toroidal magnetic field. In the near future, consequences for angular momentum transport and the case of general differential rotation and azimuthal magnetic field have to be studied. Moreover, the same methodology must be applied to the case of poloidal fields, and the hyperbolic regime has to be carefully studied.
Key words: magnetohydrodynamics (MHD) / waves / methods: analytical / stars: rotation / stars: magnetic field / stars: oscillations
© ESO, 2010
1. Introduction
Stellar radiation zones are stable strongly stratified rotating magnetic regions. This means that fluid dynamics in such zones are driven by the buoyancy force, the Coriolis and the centrifugal accelerations, and the Lorentz force. Furthermore, those regions are the seat of transport and mixing processes that rule the secular evolution of stars along with nuclear reactions (see for example Zahn 1992; Meynet & Maeder 2000). Then, angular momentum, heat, magnetic field, and chemicals are transported, leading to a modification of the stellar structure.
Observations from space and ground (with helioseismology, asteroseismology, and highresolution spectropolarimetry, for example) now give us ever more refined constraints on physical processes in star interiors. In this way, we have to built models that become more and more realistic, now taking internal dynamical processes into account. A complete review of those mechanisms in stellar radiation zones is given by Talon (2008).
In this work, we focus on the waves that propagate in such regions, namely the internal (or gravity) waves. These waves transport angular momentum along with fossil magnetic field (Mestel et al. 1988; Gough & McIntyre 1998; Brun & Zahn 2006; Garaud & Guervilly 2009) and are serious candidates for explaining stellar radiation zone angular velocity profiles, as in the Sun and solartype stars (cf. Talon et al. 2002; Talon & Charbonnel 2005; Charbonnel & Talon 2005; Denissenkov et al. 2008). Since stellar radiation regions are differentially rotating and magnetic (for example, in the solar tachocline where waves are excited; see Browning et al. 2006), internal waves dynamics are modified by the Coriolis acceleration (the centrifugal one can be neglected to the first order in rotation) and by the Lorentz force, thus becoming magnetogravitoinertial waves (Schatzman 1993a; Barnes et al. 1998; Kumar et al. 1999; Schecter et al. 2001; Rogers & MacGregor 2010). They are equivalent to the MAC waves (for magneticArchimedeanCoriolis) studied in geophysics for the liquid Earth’s core (see Braginsky 1967; Braginsky & Roberts 1975; Friedlander 1987b, 1989), but the relative importance of each force is however different in the stellar case with a strong domination of the stratification restoring force. For the fossil magnetic field configuration in stellar radiation zones, the simplest geometrical configurations, such as purely poloidal and purely toroidal fields, are known to be unstable (Tayler 1973; Achesson 1978; Markey & Tayler 1973, 1974), with the best candidate for stable geometries the mixed poloidaltoroidal fields (Wright 1973; Markey & Tayler 1974; Tayler 1980; Braithwaite & Spruit 2004; Braithwaite 2009; Duez & Mathis 2010). However, it is a hard and very new task to consider such a complex helical field geometry in the study of internal wave dynamics. This is the reason we chose to treat the modification of wave propagation, and the transport they induce, first by a purely toroidal field, and second by a purely poloidal field (see for example Lee 200520072010 in the dipolar case), to unravel their modification by each corresponding component of a realistic mixed field.
The wave dynamics equations are given here for an axisymmetric toroidal (and thus azimuthal) magnetic field in Sect. 2. Examining a model of rotating and magnetic stellar radiation zone where the angular velocity and the Alfvén frequency are assumed to be uniform (cf. Fig. 1), which constitutes the first step in understanding the complex behaviour of such waves, we introduce in Sect. 3 the socalled traditional approximation, which we generalize to the MHD case. In this framework, we derive the complete spatial structure of magnetogravitoinertial waves and their asymptotic properties. Then, the validity of the MHD traditional approximation is discussed in Sect. 4 and wave families are identified. Finally, conclusions are given and perspectives for modifying waveinduced transport of angular momentum are discussed in Sect. 5.
2. Waves in rotating stars with a toroidal magnetic field
To study the dynamics of magnetogravitoinertial waves in stellar radiation zones, the classical perfect MHD inviscid dynamical equations system has to be solved. It is formed by the induction equation in the nondissipative case (here, the ohmic dissipation is not taken into account because it is negligible compared to the thermal diffusion) (1)the momentum equation (2)the continuity equation (3)the equation for the energy transport, which is given here in the adiabatic limit (4)and the Poisson’s equation for the gravitational potential (5)Here, B is the macroscopic magnetic field. It is formed by the sum of the largescale toroidal field , associated to the chosen uniform Alfvén frequency ω_{A}^{1}, and of the waveinduced magnetic field b: (6)where t is the time and r = (r,θ,ϕ) are the usual spherical coordinates with their associated unit vector basis , ρ is the density, and μ is the magnetic permeability of the medium. Next, V is the macroscopic velocity field. It is formed by the sum of the largescale azimuthal velocity V_{0}, associated to the chosen uniform rotation Ω_{s}, and of the wave velocity u: (7)where D_{t} = ∂_{t} + (V·∇) is the Lagrangian derivative. The Φ and P are the gravitational potential and the pressure, while Γ_{1} = (∂lnP/∂lnρ)_{S} is the adiabatic exponent, with S the macroscopic entropy, and G the universal gravitational constant.
Fig. 1 The setup to study internal wave dynamics in rotating and magnetic stellar radiation zones. Angular velocity (Ω_{s}) and Alfvén frequency are assumed uniform in a first step to unravel magnetogravitoinertial wave characteristics. 

Open with DEXTER 
To achieve our aim, Eqs. (1)–(5) are linearized around the rotating magnetic steady state. Each scalar field X (the density, the gravitational potential, and the pressure) is then expanded as the sum of its hydrostatic value and of the wave’s associated fluctuation : (8)We neglect the nonspherical character of the hydrostatic background due to the deformation associated to the centrifugal acceleration, , and to the Lorentz force associated to , .
Then, following Braginsky (1967), Braginsky & Roberts (1975), and Unno et al. (1989), the wave’s Lagrangian displacement ξ is introduced and defined such that (9)Using Eq. (7), it becomes (10)Next, inserting Eq. (9) in the induction equation (Eq. (1)), we get (11)Using the definition of given in Eq. (6), this leads to (12)Then, the momentum equation (Eq. (2)) becomes (13)where is the unit vector along the rotation axis. Next, the wave total pressure fluctuation is given by the sum of the gas pressure fluctuation and of the wave magnetic pressure : (14)Afterwards, the wave magnetic tension force is given by (15)Finally, the continuity (Eq. (3)), the energy transport (Eq. (4)), and the Poisson’s (Eq. (5)) equations are respectively given by (16)(17)(18)Next, and each vectorial field x(r,t) = {u,ξ,b} are expanded in Fourier’s series in ϕ and t: (19)(20)with σ the wave angular velocities in an inertial frame excited in the star.
On one hand, we get for the velocity field (21)In the considered rotating stellar radiation zone, waves are thus Dopplershifted by rotation (Ω_{s}) and the local wave angular velocity that corresponds to the operator ∂_{t} + Ω_{s}∂_{ϕ} appears.
On the other hand, Eq. (12) becomes (22)Then, the momentum equation is written as (23)where (24)Expanding each component explicitely gives along , , and (25)(26)(27)Here can be seen as a modified local wave angular velocity that corresponds to the modification of σ_{s} by the presence of the magnetic field. Thus, waves are only propagating if ; in the other case, they are trapped in the vertical direction (cf. Schatzman 1993a; and Barnes et al. 1998) because the azimuthal magnetic field acts as a filter in this direction (see in Figs. 6 and 7). Moreover, we define (28)which is the sum of the total wave dynamical pressure fluctuation^{2} and of the gravific potential fluctuation.
Next, the continuity equation is given by (29)Afterwards, the adiabatic energy transport equation becomes (30)where the gravity and the BruntVäisälä frequency are given by and , respectively.
Finally, the Poisson’s equation is written as (31)From now on, we adopt the Cowling’s approximation (Cowling 1941) where the wave’s gravific potential fluctuation is neglected. Therefore, Eq. (28) reduces to .
2.1. Local analysis
Following Kumar et al. (1999) who studied the dynamics of lowfrequency magnetogravitoinertial waves (cf. Appendix A), we now consider a more simplified setup than ours to obtain their dispersion relation. Indeed, we consider a stellar radiation zone where both the angular velocity and the background magnetic field are uniform (i.e. Ω = Ω_{s} and ). Then, assuming that all the perturbed quantities have a dependence proportional to exp[i(k·r + σt)], the momentum equation (Eq. (2)) becomes (32)After vectorial manipulations given in Appendix A, this yields to the dispersion relation for waves in this setup (Eq. (A.6)): (33)where , , , and . This can be solved, leading to (34)Let us now discuss this dispersion relation as done by Heng & Spitkovsky (2009). First, if we examine the nonrotating limit (Ω_{s} = 0), we get (35)Two branches are then isolated: the lowfrequency one corresponds to Alfvén waves while magnetogravity waves, for which restoring forces are both buoyancy and Lorentz forces, are linked to the higherfrequency branch (see Schecter et al. 2001).
Then, if we examine the rotating system, both wave families couple through the Coriolis acceleration. On one hand, for lowfrequency Alfvén waves, Coriolis acceleration tries to balance the Lorentz force and for this reason, those waves are then called “magnetostrophic waves”. Their approximate dispersion relation is given by (36)On the other hand, magnetogravity waves become magnetogravitoinertial waves, which are also called magnetoPoincaré waves and magnetoRossby waves for those waves that are due to the star curvature and the associated variation of the local rotation on a tangent sphere. In this case, we obtain the following approximative dispersion relation: (37)Let us introduce the expansion of k both in the spherical and cylindrical coordinates: (38)where and , with (s,ϕ,z) the cylindrical coordinates and the associated unitvector basis. Then, the dispersion relation is written as (39)In the case of highly stratified stellar radiation zones where ω_{A} ≪ N and 2Ω_{s} ≪ N, this reduces for lowfrequency waves (σ ≪ N) to (40)which leads to the following hierarchy (41)because of the anelastic approximation where acoustic waves are filtered, which becomes in the local analysis case k·ξ ≈ 0.
From now on, we focus on this type of wave (in the case where ω_{A} ≪ N and 2Ω_{s} ≪ N) for which σ can be of the same order of magnitude as ω_{A} and 2Ω_{s}, and the ratio ω_{A}/2Ω_{s} is taken as general as possible (see Fig. 2 below).
Fig. 2 Wave types in a rotating and magnetic stellar radiation zone and associated frequencies (where f_{L} is the Lamb’s frequency). Values for the Sun (⊙ ) at the bottom of the Tachocline are given in brackets. 

Open with DEXTER 
3. Lowfrequency waves in rotating magnetic stellar radiation zones
3.1. The traditional approximation
We first discuss the nonmagnetic case, namely the gravitoinertial waves. In the general case, the Poincaré operator, which governs the spatial structure of waves, is of mixed type (elliptic and hyperbolic) and not separable (see the detailed discussions in Friedlander & Siegman 1982; Dintrans et al. 1999; Dintrans & Rieutord 2000). This leads to the appearance of detached shear layers associated with the underlying singularities of the adiabatic problem that could be crucial for transport and mixing processes in stellar radiation zones, since they are the seat of strong dissipation (cf. Stewartson & Richard 1969; Stewartson & Walton 1976; Dintrans et al. 1999; and Dintrans & Rieutord 2000).
Let us now focus on the case of a uniform rotation (Ω = Ω_{s}). In the largest part of stellar radiation zones, we are in a regime where 2Ω_{s} ≪ N. Since we are interested here in lowfrequency waves (σ ≪ N), the traditional approximation (hereafter TA), which consists in neglecting the latitudinal component of the rotation vector (), , in the Coriolis acceleration, can be adopted in the superinertial regime where 2Ω_{s} < σ ≪ N (see e.g. Eckart 1960; for a modern description in a stellar context see Bildsten et al. 1996; Lee & Saio 1997; and Mathis et al. 2008). Then, variable separation in radial and horizontal eigenfunctions remains possible (cf. Friedlander 1987a), which corresponds to the ergodic (regular) elliptic gravitoinertial mode family (the E_{1} modes in Dintrans et al. 1999; and Dintrans & Rieutord 2000). This approximation has to be carefully used however, because it changes the nature of the Poincaré operator, and removes the singularities and associated shear layers that appear. Then, in the subinertial regime, where σ ≤ 2Ω_{s}, which corresponds to the equatorially trapped hyperbolic modes (the H_{2} modes in Dintrans et al. 1999; and Dintrans & Rieutord 2000), the TA fails to reproduce the waves behaviour and the complete momentum equation has to be solved (see detailed examples in Gerkema & Shrira 2005; and Gerkema et al. 2008). Note also the work by Fruman (2009), who examines the validity of the TA on the equatorial βplane in function of the 2Ω_{s}/N value.
As a first step, we thus studied the regular elliptic gravitoinertial waves for which the TA applies (Mathis et al. 2008; and Mathis 2009). In the MHD case, our purpose is to generalize the TA to , thus neglecting the latitudinal component of since 2Ω_{s} ≪ N, ω_{A} ≪ N, and σ ≪ N. Therefore, we restrict ourselves here to the regular lowfrequency waves for which the TA can be used as we show it in the next section. Its application domain is discussed in Sect. 4.
3.2. Dynamical equations
Assuming the TA for lowfrequency magnetogravitoinertial waves, we respectively get for the linearized momentum equation components: (42)(43)(44)where the linearized energy equation (Eq. (30)) has been used to express the buoyancy force .
We need to describe the different assumptions for using the TA. On one hand, since ω_{A} ≪ N and 2Ω_{s} ≪ N, we neglect the term in the radial component of the momentum equation (Eq. (25)). On the other hand, the first term also has to be neglected, but we first conserve it to make the historical link with works in the nonrotating and nonmagnetic cases (Press 1981; Schatzman 1993b; Zahn et al. 1997) and those about gravitoinertial waves (Mathis et al. 2008; Mathis 2009). Finally, since (cf. Eq. (41)) for lowfrequency internal waves, the term is neglected in the azimuthal component (Eq. (27)).
3.3. Spatial structure of the wavedisplacement, velocity, and magnetic field
3.3.1. Spatial structure of the horizontal components of the displacement
After successively eliminating and between Eqs. (43) and (44), each of them is expressed in function of as (45)(46)where x = cosθ has been introduced. Then ν_{M} is defined as (47)where (48)Here, ν_{s} is the spin parameter (Fig. 3), which equals the inverse of the Rossby number R_{o} = σ_{s}/2Ω_{s} (cf. Lee & Saio 1997), and F_{M} = ν_{M}/ν_{s} gives its modification by the magnetic field.
Fig. 3 Inverse of the Rossby number (ν_{s}) as a function of the frequency (σ/2π) for axisymmetric waves . We assume that Ω_{s}/2π = 430 nHz. In the purely hydrodynamical case (namely those of gravitoinertial waves), the TA is allowed in the superinertial regime where 2Ω_{s} < σ (i.e. ν_{s} < 1). 

Open with DEXTER 
We introduce the wave’s Elsasser number (49)which gives the ratio of the Lorentz force with the Coriolis acceleration (cf. Eqs. (13) and (15) and Fig. 4)^{3}.
Fig. 4 Elsasser number (Λ_{E}) as a function of the wave frequency in the corotating frame and of the azimuthal magnetic field amplitude at the equator just below the solar tachocline . There and we assume that Ω_{s}/2π = 430 nHz. The upper limit for the amplitude of the magnetic field has been chosen following constraints obtained by Antia et al. (2003). 

Open with DEXTER 
3.3.2. Spatial structure of the radial component of the displacement and of the pressure
As in the nonmagnetic case and because of the equation structure (cf. Lee & Saio 1997; Mathis et al. 2008), we choose to expand scalar quantities and the displacement’s vertical component as (50)(51)The Θ_{k,m} functions are the orthogonal Hough functions for which where δ is the Kronecker symbol (Laplace 1799; Hough 1898; LonguetHiggins 1968; Lindzen & Chapman 1969). They are the eigenfunctions (with the associated eigenvalues Λ_{k,m}) of the socalled “Laplace tidal equation” (hereafter LTE): (52)where the Laplace tidal operator is given by (53)For a detailed discussion of the boundary conditions of the LTE, we refer the reader to the Lee & Saio work (1997). Examples of Hough functions are given in Fig. 5.
As emphasized in Mathis et al. (2008), pro and retrograde eigenfunctions (and eigenvalues) are different. This crucial point for induced angular momentum transport will be discussed in detail in Paper II.
Fig. 5 Left: hough functions Θ_{k,m}(cosθ;ν) (Top) and associated latitudinal and azimuthal functions (Middle and Bottom) m = 2, ν = 0 (continuous lines); m = 2, ν = 0.86 (dotted lines); m = −2, ν = 0.86 (dashed lines). Right: (Top) Eigenvalue (Λ) of Laplace’s tidal equation (in the presence of rotation and of magnetic field) in the range relevant for a solar model and the MHD TA . (Bottom) Equivalent horizontal eigenvalue () (see Eq. (85)). (Left) Class I waves with 1 ≤ l ≤ 6, and m = −l + 2 (black continuous line), m = 0 (long blue dashed line), and m = l (dashed red line). (Middle) Class III waves (s = 0, s being defined in Eq. (89)) have negative values of m (m = 0,··· , −5). (Right) Class IV waves (s = −1) have indices m = −1,··· , −6. Waves classes are discussed in Sect. 4.2. (see also Mathis et al. 2008). 

Open with DEXTER 
This allows separating the radial and latitudinal variables for and . Introducing the amplitude of the horizontal displacement leads to (54)and the latitudinal component of ξ′ is written (55)where (56)with (57)In the same way, its azimuthal component is given by (58)where (59)with (60)Then, the radial component of the momentum equation assuming the TA (Eq. (42)), the continuity equation (Eq. (29)), and the energy transport equation (Eq. (30)) become (61)(62)(63)Eliminating between the previous equations and assuming the Cowling’s approximation, we get the following system for r^{2}ξ_{r;k,m} and : (64)(65)where we explicitly introduce the wave magnetic pressure expansion (66)Terms where it explicitly appears are introduced through in the energy equation (Eq. (63)) because of Eq. (14).
This is a generalization of the system obtained by Press (1981) in the nonrotating and nonmagnetic case where σ_{M} and respectively replace σ and l(l + 1), where l is the orbital number of the classical spherical harmonics.
Adopting the anelastic approximation, where magnetoacoustic waves are filtered out (i.e. and the terms proportional to are neglected, where is the sound speed), and following the procedure given in Zahn (1970, 1975) and in Press (1981), we respectively get for the vertical displacement (67)where , and for the pressure fluctuation (68)where .
3.3.3. The final wavedisplacement, velocity, and magnetic field
Once one solves Eq. (67) or Eq. (68), the associated final expression for ξ, u and b are derived. First, using Eqs. ((51)−(45)−(46)−(54)), the wavedisplacement is given by (69)where (70)(71)(72)Then, the wave velocity field and magnetic field are straightforwardly derived using u = i σ_{s} ξ and (cf. Eqs. (21) and (22)).
Finally, the dynamical pressure is given by (73)where .
Therefore, one of the main important conclusions of the study of this model, which assumes uniform angular velocity (Ω_{s}) and Alfvén frequency , is that the waveinduced perturbations in this case can be completely expressed as a function of the Hough functions used in the nonmagnetic uniformly rotating case. In that framework, the spin parameter (ν_{s}) is now replaced by ν_{M}, which accounts for magnetic field. The simplest case of magnetogravity waves is treated in Appendix B.
However, we have to emphasize that such a neat variable separation is not possible in the case of a general differential rotation and azimuthal magnetic field (Ω(r,θ), ), even if the TA applies (cf. Mathis 2009). In that case, which is beyond the scope of the present paper, a more subtle treatment of the dynamical equations has to be adopted.
3.3.4. Asymptotic analysis
Let us now consider the lowfrequency waves for which σ ≪ N. We introduce the vertical wave number (74)then, Eq. (67) is written (75)For these waves the J.W.K.B. approximation can be adopted (cf. Landau & Lifshitz 1966; Press 1981; Schatzman 1993b; Zahn et al. 1997). Applying it to Eq. (75), we get^{4}(76)where r_{c} is the radius of the convectionradiation border while the amplitude function is given by (77)Here, A_{k,m} is the wavedisplacement amplitude that has to be obtained using excitation models corresponding to the stochastic excitation by turbulent convective movements (Kiraga et al. 2003, 2005; Dintrans et al. 2005; Rogers & Glatzmaier 2005, 2006; Belkacem et al. 2009) or to the κmechanism (Lee & Saio 1987; Unno et al. 1989; Lee & Baraffe 1995).
Fig. 6 Top: ν_{M} as a function of ν_{s} and Λ_{E} for nonaxisymmetric retrograde (m = 1) and prograde (m = −1) waves. The surfaces such that ν_{M} = 1 are given by the thick black lines and the isoν_{M} lines for ν_{M} < 1 and ν_{M} > 1 are given by the red and the blue lines. The big external white region in the topright corner in each case corresponds to the domain where , the wave being thus vertically trapped there. The MHD TA applies when ν_{M} < 1 and (red lines). In the other case (blue and black lines, i.e.  ν_{M}  ≥ 1 and ), it does not apply due to the singularity at the critical latitude (θ_{c}) given in Eq. (88) and plotted in Fig. (7). Bottom: same for m = 2 and m = −2. 

Open with DEXTER 
Fig. 7 Top: θ_{c} as a function of ν_{s} and Λ_{E} for nonaxisymmetric retrograde (m = 1) and prograde (m = −1) waves. As in Fig. (6), the large white topright corner region corresponds to the domain where and where the waves are trapped in the vertical direction. In the white region at the left, the MHD TA applies (i.e. ν_{M} < 1 and ) and there is no critical latitude. Finally, the blue θ_{c}iso lines correspond to the regime where the TA does not apply (i.e. ν_{M} ≥ 1 and ). Middle: same for m = 2 and m = −2. Bottom: cartoon for the wave propagation region: traditional elliptic waves are regular for every latitude and live in the whole sphere (E), while hyperbolic ones are trapped in the equatorial belt where θ ∈ [θ_{c},π − θ_{c}] (H). 

Open with DEXTER 
We then derive ξ, u, b and Π. First, we get (78)where (79)(80)(81)Next, u and b are obtained once again using Eqs. (21) and (22). Finally, we get using Eqs. (61) and (63) in the anelastic approximation regime (82)with (83)On the other hand, following Pantillon et al. (2007), we can define a horizontal wave number given by (84)where (cf. Fig. 5) (85) and is the horizontal spherical Laplacian.
In the absence of rotation and of magnetic field (ν_{M} = 0), we recover . The associated monochromatic vertical wave group velocity is given by (86)where is the monochromatic wave phase velocity in solartype stars. Signs are due to the direction of the excitation kinetic energy flux directed from the convective envelope to the star’s center; in massive stars, signs are inverted. Finally, to determine eigenfrequencies one has to follow strictly the same methodology as described in Friedlander (1987a,b) and in Lee & Saio (1989).
4. The traditional approximation in the MHD case
4.1. Validity domain
The MHD TA can be applied in spherical shells where (87)thus as long ν_{M} < 1 and ∀r and ∀θ ∈ [0,π] (cf. Fig. 6). In this regime, waves are elliptic and regular at all latitudes. In the other case, where both and (ν_{M} ≥ 1 and ), waves become hyperbolic and trapped in an equatorial belt (see Dintrans & Rieutord 2000, for example) where θ ∈ [θ_{c},π − θ_{c}], with θ_{c} the critical colatitude (88)where (cf. Fig. 7). Then, the adiabatic velocity field is singular and the MHD TA cannot be applied (see Sect. 3.1, and references therein), because the wave behaviour is possibly ruled by the dissipation in wave attractors (see for example in the hydrodynamical case Dintrans et al. 1999; Dintrans & Rieutord 2000; and Maas & Harlander 2007). This is a strict generalization of the criteria derived in the hydrodynamical case for gravitoinertial waves (Mathis et al. 2008; Mathis 2009).
4.2. Wave classification
Under the MHD TA (as long as ν_{M} < 1), four types of magnetogravitoinertial waves can be identified (see Townsend 2003; and Mathis et al. 2008, and references therein in the hydrodynamical case).

Class I waves. They are internal gravity waves, which exist in the nonrotating and in the nonmagnetic cases, which are modified both by Coriolis acceleration and the Lorentz force. Their eigenvalue (Λ_{k,m}), hence their radial wave number, , are increased. Such waves are also called magnetoPoincaré waves (see Zaqarashvili et al. 2009; Heng & Spitovsky 2009 and Sect. 2).

Class II waves. They are purely retrograde waves (m > 0), which only exist in the case of ν_{M} high values. Their dynamics are driven by the conservation of the specific vorticity combined with the effects of curvature and by the Lorentz force. However, since ν_{M} ≥ 1 in this case, they cannot be treated using the MHD TA. In the hydrodynamical case, they are called “quasiinertial” waves, which corresponds to the geophysical Rossby waves (see Provost et al. 1981). Such magnetoRossby waves have recently been studied by Zaqarashvili et al. (2007), Zaqarashvili et al. (2009), and Heng & Spitovsky (2009) in the limit of a thin stratified layer.

Class III waves. These are mixed class I and class II waves, and m ≤ 0 waves exist in the absence of rotation and of magnetic field. Waves with m > 0 appear when ν_{M} = m + 1 with low eigenvalues, while their horizontal eigenfunctions are . When they appear and have low eigenvalues, they behave mostly like class II waves; m ≤ 0 and m > 0 waves with high eigenvalues behave rather like class I waves. Their eigenvalues are much lower than those of class I waves. Thus, they have a smaller vertical wave number. They may be identified with the geophysical Yanai waves (Yanai & Maruyama 1966).

Class IV waves. They are purely prograde waves (m < 0) whose characteristics change little with ν_{M}, since their displacement in the θ direction is very small. Their dynamics are driven by the conservation of the specific vorticity combined with the stratification effects and by the Lorentz force; their eigenvalues are lower than those of both class I and class III waves. Hence, their vertical wave number is smaller. In the hydrodynamical case, they may be identified with the geophysical Kelvin waves (Pedlosky 1998).
If we define a new index s by (89)then Class IV waves have s = −1, class III waves s = 0, and class I waves have s = 1,2,3,...
5. Conclusion and perspectives
In this work, we examine of lowfrequency internal wave behaviour in stellar radiation zones, which are stable highly stratified rotating magnetic regions. Then, waves dynamics are driven by the buoyancy force, the Coriolis acceleration, and the Lorentz force. Internal waves thus become magnetogravitoinertial waves that are equivalent to the MAC waves (magnetic – Archimedean – Coriolis) which have been studied for the Earth’s core. In this work in the stellar context, we studied the case of radiation zones with an axisymmetric purely toroidal magnetic field, where both the angular velocity (Ω_{s}) and the Alfvén pulsation (ω_{A}) are assumed to be uniform. This allowed to extract the main properties of the system. First, we derived dynamical equations, following Braginsky’s work and using stellar notations. Then, the MHD TA, which can be used only in the strongly stratified case when ν_{M} < 1 and , has been introduced and discussed. This allows to simplify the equations system and to obtain in this case the wavevelocity field, the wavemagnetic field, and the total pressure fluctuation as a function of the usual Hough functions and associated special latitudinal and azimuthal functions used in the nonmagnetic case for gravitoinertial waves in the superinertial strongly stratified regime (ν_{s} < 1). Then, four classes of magnetogravitoinertial waves are indentified, while the asymptotic behaviour of such lowfrequency waves is discussed.
Once this study is achieved, two main works have to be engaged in the near future. The first one, is to examine the angular momentum transport by such waves in the “weak differential rotation case” (cf. Mathis et al. 2008), assuming a uniform Alfvén frequency to isolate the impact of an axisymmetric toroidal magnetic field as a function of its intensity (Paper II). Then, the more general case of general differential rotation and azimuthal magnetic field has to be studied, because it has been done in Mathis (2009) in the hydrodynamical case (Paper III). Moreover, the same methodology must be applied to axisymmetric poloidal magnetic fields. This will lead to a coherent picture of angular momentum waveinduced transport taking the impact of an axisymmetric fossil magnetic field into account.
The Alfvén speed associated to an axisymmetric toroidal magnetic field () is given by . Since ω_{A} = V_{A}/s, where s = rsinθ is the distance from the symmetry axis, we get . Here, we choose a uniform ω_{A} as a first step. The case of a general realistic axisymmetric toroidal field (ω_{A}(r,θ)) will be treated in Paper III.
Here, we consider the case of solartype stars for which the group velocity is negative (cf. Eq. (86)). In the case of massive stars, where it is positive, the phase function is .
Acknowledgments
Authors thank Drs. A.S. Brun, S. Talon, S. TurckChièze, and J.P. Zahn for fruitful discussions during this work, which was supported in part by the Programme National de Physique Stellaire (CNRS/INSU), the CNES/GOLF grant at the Service d’Astrophysique (CEA – Saclay), and the CNRS Physique théorique et ses interfaces programme.
References
 Acheson, D. J. 1978, R. Soc. London Phil. Trans. Ser. A, 289, 459 [NASA ADS] [CrossRef] [MathSciNet] [Google Scholar]
 Antia, H. M., Chitre, S. M., & Thompson, M. J. 2003, A&A, 399, 329 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Barnes, G., MacGregor, K. B., & Charbonneau, P. 1998, ApJ, 498, L169 [NASA ADS] [CrossRef] [Google Scholar]
 Belkacem, K., Samadi, R., Goupil, M. J., et al. 2009, A&A, 494, 191 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Bildsten, L., Ushomirsky, G., & Cutler, C. 1996, ApJ, 460, 827 [NASA ADS] [CrossRef] [Google Scholar]
 Braginsky, S. I. 1967, Geomagnetism and Aeronomy, 7, 851 [NASA ADS] [Google Scholar]
 Braginsky, S. I., & Roberts, P. H. 1975, Proc. Royal Soc. London A, 347, 125 [NASA ADS] [CrossRef] [Google Scholar]
 Braithwaite, J. 2009, MNRAS, 397, 763 [NASA ADS] [CrossRef] [Google Scholar]
 Braithwaite, J., & Spruit, H. C. 2004, Nature, 431, 819 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Browning, M. K., Miesch, M. S., Brun, A. S., & Toomre, J. 2006, ApJ, 648, L157 [NASA ADS] [CrossRef] [Google Scholar]
 Brun, A.S., & Zahn, J.P. 2006, A&A, 457, 665 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Charbonnel, C., & Talon, S. 2005, Science, 309, 2189 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
 Cowling, T. G. 1941, MNRAS, 101, 367 [NASA ADS] [Google Scholar]
 Denissenkov, P. A., Pinsonneault, M., & MacGregor, K. B. 2008, ApJ, 684, 757 [NASA ADS] [CrossRef] [Google Scholar]
 Dintrans, B., & Rieutord, M. 2000, A&A, 354, 86 [NASA ADS] [Google Scholar]
 Dintrans, B., Rieutord, M., & Valdettaro, L. 1999, J. Fluid Mech., 398, 271 [NASA ADS] [CrossRef] [Google Scholar]
 Dintrans, B., Brandenburg, A., Nordlund, Å, & Stein, R. F. 2005, A&A, 438, 365 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Duez, V., & Mathis, S. 2010, A&A , 517, A58 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Eckart, C. 1960, Hydrodynamics of Oceans and Atmospheres (Oxford: Pergamon Press) [Google Scholar]
 Friedlander, S. 1987a, Geophys. J. Royal Astron. Soc., 89, 637 [Google Scholar]
 Friedlander, S. 1987b, Geophys. Astrophys. Fluid Dyn., 39, 315 [NASA ADS] [CrossRef] [Google Scholar]
 Friedlander, S. 1989, Geophys. Astrophys. Fluid Dyn., 48, 53 [NASA ADS] [CrossRef] [Google Scholar]
 Friedlander, S., & Siegmann, W. L. 1982, Geophys. Astrophys. Fluid Dyn., 19, 267 [NASA ADS] [CrossRef] [Google Scholar]
 Fruman, M. D. 2009, J. Atmos. Sci., 66, 2937 [NASA ADS] [CrossRef] [Google Scholar]
 Garaud, P., & Guervilly, C. 2009, ApJ, 695, 799 [NASA ADS] [CrossRef] [Google Scholar]
 Gerkema, T., & Shrira, V. I. 2005, J. Fluid Mech., 529, 195 [NASA ADS] [CrossRef] [Google Scholar]
 Gerkema, T., Zimmerman, J. T. F., Mass, L. R. M., & Van Haren, H. 2008, Rev. Geophys., 46, CiteID RG2004 [Google Scholar]
 Gough, D. O., & McIntyre, M. E. 1998, Nature, 394, 755 [NASA ADS] [CrossRef] [Google Scholar]
 Heng, K., & Spitkovsky, A. 2009, ApJ, 703, 1819 [NASA ADS] [CrossRef] [Google Scholar]
 Hough, S. S. 1898, Philos. Trans. Royal Soc. London A, 191, 139 [NASA ADS] [CrossRef] [Google Scholar]
 Kiraga, M., Jahn, K., Stepien, K., & Zahn, J.P. 2003, Acta Astron., 53, 321 [NASA ADS] [Google Scholar]
 Kiraga, M., Stepien, K., & Jahn, K. 2005, Acta Astron., 55, 205 [NASA ADS] [Google Scholar]
 Kumar, P., Talon, S., & Zahn, J.P. 1999, ApJ, 520, 859 [NASA ADS] [CrossRef] [Google Scholar]
 Landau, L. D., & Lifchitz, E. M. 1966, Theoretical Physics: Quantum Mechanics (Moscow: Mir) [Google Scholar]
 Laplace, P.S. 1799, Mécanique céleste, Bureau des Longitudes, Paris [Google Scholar]
 Lee, U. 2005, MNRAS, 357, 97 [NASA ADS] [CrossRef] [Google Scholar]
 Lee, U. 2007, MNRAS, 374, 1015 [NASA ADS] [CrossRef] [Google Scholar]
 Lee, U. 2010, MNRAS, 405, 1444 [NASA ADS] [Google Scholar]
 Lee, U., & Saio, H. 1987, MNRAS, 225, 643 [NASA ADS] [Google Scholar]
 Lee, U., & Saio, H. 1989, MNRAS, 237, 875 [NASA ADS] [Google Scholar]
 Lee, U., & Baraffe, I. 1995, A&A, 301, 419 [NASA ADS] [Google Scholar]
 Lee, U., & Saio, H. 1997, ApJ, 491, 839 [NASA ADS] [CrossRef] [Google Scholar]
 Lindzen, R. S., & Chapman, S. 1969, Space Sci. Rev., 10, 3 [NASA ADS] [CrossRef] [Google Scholar]
 LonguetHiggins, M. S. 1968, Philos. Trans. Royal Soc. of London A, 262, 511 [NASA ADS] [CrossRef] [Google Scholar]
 Maas, L., & Harlander, U. 2007, J. Fluid Mech., 570, 47 [NASA ADS] [CrossRef] [Google Scholar]
 Markey, P., & Tayler, R. J. 1973, MNRAS, 163, 77 [NASA ADS] [CrossRef] [Google Scholar]
 Markey, P., & Tayler, R. J. 1974, MNRAS, 168, 505 [NASA ADS] [CrossRef] [Google Scholar]
 Mathis, S. 2009, A&A, 506, 811 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Mathis, S., Talon, S., Pantillon, F.P., & Zahn, J.P. 2008, Sol. Phys., 251, 101 [NASA ADS] [CrossRef] [Google Scholar]
 Meynet, G., & Maeder, A. 2000, A&A, 361, 101 [NASA ADS] [Google Scholar]
 Mestel, L., Tayler, R. J., & Moss, D. L. 1988, MNRAS, 231, 873 [NASA ADS] [CrossRef] [Google Scholar]
 Pantillon, F. P., Talon, S., & Charbonnel, C. 2007, A&A, 474, 155 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Pedlosky, J. 1998, Geophysical fluid dynamics, 2nd edn. (Springer) [Google Scholar]
 Press, W. 1981, ApJ, 245, 286 [NASA ADS] [CrossRef] [Google Scholar]
 Provost, J., Berthomieu, G., & Rocca, G. 1981, A&A, 94, 126 [NASA ADS] [Google Scholar]
 Rogers, T. M., & Glatzmaier, G. A. 2005, MNRAS, 364, 1135 [NASA ADS] [CrossRef] [Google Scholar]
 Rogers, T. M., & Glatzmaier, G. A. 2006, ApJ, 653, 756 [NASA ADS] [CrossRef] [Google Scholar]
 Rogers, T. M., & MacGregor, K. B. 2010, MNRAS, 401, 191 [NASA ADS] [CrossRef] [Google Scholar]
 Schatzman, E. 1993a, A&A, 271, L29 [NASA ADS] [Google Scholar]
 Schatzman, E. 1993b, A&A, 279, 431 [NASA ADS] [Google Scholar]
 Schecter, D. A., Boyd, J. F., & Gilman, P. A. 2001, ApJ, 551, L185 [NASA ADS] [CrossRef] [Google Scholar]
 Stewartson, K., & Richard, J. 1969, J. Fluid Mech., 35, 759 [NASA ADS] [CrossRef] [Google Scholar]
 Stewartson, K., & Walton, I. C. 1976, Proc. Royal Soc. London A, 349, 141 [NASA ADS] [CrossRef] [Google Scholar]
 Talon, S. 2008, EAS PS, 32, 81 [CrossRef] [EDP Sciences] [Google Scholar]
 Talon, S., & Charbonnel, C. 2005, A&A, 440, 981 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Talon, S., Kumar, P., & Zahn, J.P. 2002, ApJ, 574, L175 [NASA ADS] [CrossRef] [Google Scholar]
 Tayler, R. J. 1973, MNRAS, 161, 365 [NASA ADS] [CrossRef] [Google Scholar]
 Tayler, R. J. 1980, MNRAS, 191, 151 [NASA ADS] [Google Scholar]
 Townsend, R. H. D. 2003, MNRAS, 340, 1020 [NASA ADS] [CrossRef] [Google Scholar]
 Unno, W., Osaki, Y., Ando, H., Saio, H., & Shibahashi, H. 1989, Nonradial oscillations of stars, second edition (Tokyo: University of Tokyo Press) [Google Scholar]
 Wright, G. A. E. 1973, MNRAS, 162, 339 [NASA ADS] [CrossRef] [Google Scholar]
 Yanai, M., & Maruyama, T. 1966, J. Meteorol. Soc. Japan, 44, 291 [Google Scholar]
 Zahn, J.P. 1970, A&A, 4, 452 [NASA ADS] [Google Scholar]
 Zahn, J.P. 1975, A&A, 41, 329 [NASA ADS] [Google Scholar]
 Zahn, J.P. 1992, A&A, 265, 115 [NASA ADS] [Google Scholar]
 Zahn, J.P., Talon, S., & Matias, J. 1997, A&A, 322, 320 [NASA ADS] [Google Scholar]
 Zaqarashvili, T. V., Olivier, R., Ballester, J. L., & Shergelashvili, B. M. 2007, A&A, 470, 815 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
 Zaqarashvili, T. V., Olivier, R., & Ballester, J. L. 2009, ApJ, 691, L41 [NASA ADS] [CrossRef] [Google Scholar]
Appendix A: Magnetogravitoinertial waves dispersion relation
To establish the dispersion relation of magnetogravitoinertial waves, we follow the method proposed in Kumar et al. (1999). As in our simplified model, we assume that the angular velocity is uniform (Ω = Ω_{s}), while the unperturbed magnetic field is here (and only here) assumed to also be uniform (i.e. ). We consider the shortwavelength limit, where spatial and temporal dependence of all perturbed quantities are taken as proportional to exp[i(k·r + σt)]. Then, the linearized momentum equation (Eq. (2)) is given by (A.1)where . Using the energy transport equation in the adiabatic limit (Eq. (30)), this becomes (A.2)Substituting for b using the linearized induction equation (Eq. (1)) in the case of a perfectly conducting fluid and using the anelastic approximation that becomes in the shortwavelength k·ξ = 0, we obtain (A.3)Taking the vector cross product of the previous equation with k, we get (A.4)where is the Alfvén velocity and . Taking the cross product of the previous equation with k allows k × ξ as a function of ξ. Substituting back in Eq. (A.4), we obtain (A.5)Taking the rcomponent of this equation gives the searched dispersion relation: (A.6)where we have defined and , which can be inverted: (A.7)Two branches can be isolated: the lowerfrequency branch corresponds to the Alfvén waves modified by the Coriolis acceleration, and the higherfrequency one is linked to the magnetogravitoinertial waves.
Appendix B: The magnetogravity waves case
In this short appendix, we simplify the general case treated here to unravel the magnetic field impact on gravity waves. We thus do not take the Coriolis acceleration into account (i.e. ). Then, ν_{M} becomes (B.1)and the waves’ horizontal structure is described by associated Hough functions. Moreover, the MHD TA can be applied as long as ν_{M} < 1, which corresponds to (B.2)for propagative waves for which .
All Figures
Fig. 1 The setup to study internal wave dynamics in rotating and magnetic stellar radiation zones. Angular velocity (Ω_{s}) and Alfvén frequency are assumed uniform in a first step to unravel magnetogravitoinertial wave characteristics. 

Open with DEXTER  
In the text 
Fig. 2 Wave types in a rotating and magnetic stellar radiation zone and associated frequencies (where f_{L} is the Lamb’s frequency). Values for the Sun (⊙ ) at the bottom of the Tachocline are given in brackets. 

Open with DEXTER  
In the text 
Fig. 3 Inverse of the Rossby number (ν_{s}) as a function of the frequency (σ/2π) for axisymmetric waves . We assume that Ω_{s}/2π = 430 nHz. In the purely hydrodynamical case (namely those of gravitoinertial waves), the TA is allowed in the superinertial regime where 2Ω_{s} < σ (i.e. ν_{s} < 1). 

Open with DEXTER  
In the text 
Fig. 4 Elsasser number (Λ_{E}) as a function of the wave frequency in the corotating frame and of the azimuthal magnetic field amplitude at the equator just below the solar tachocline . There and we assume that Ω_{s}/2π = 430 nHz. The upper limit for the amplitude of the magnetic field has been chosen following constraints obtained by Antia et al. (2003). 

Open with DEXTER  
In the text 
Fig. 5 Left: hough functions Θ_{k,m}(cosθ;ν) (Top) and associated latitudinal and azimuthal functions (Middle and Bottom) m = 2, ν = 0 (continuous lines); m = 2, ν = 0.86 (dotted lines); m = −2, ν = 0.86 (dashed lines). Right: (Top) Eigenvalue (Λ) of Laplace’s tidal equation (in the presence of rotation and of magnetic field) in the range relevant for a solar model and the MHD TA . (Bottom) Equivalent horizontal eigenvalue () (see Eq. (85)). (Left) Class I waves with 1 ≤ l ≤ 6, and m = −l + 2 (black continuous line), m = 0 (long blue dashed line), and m = l (dashed red line). (Middle) Class III waves (s = 0, s being defined in Eq. (89)) have negative values of m (m = 0,··· , −5). (Right) Class IV waves (s = −1) have indices m = −1,··· , −6. Waves classes are discussed in Sect. 4.2. (see also Mathis et al. 2008). 

Open with DEXTER  
In the text 
Fig. 6 Top: ν_{M} as a function of ν_{s} and Λ_{E} for nonaxisymmetric retrograde (m = 1) and prograde (m = −1) waves. The surfaces such that ν_{M} = 1 are given by the thick black lines and the isoν_{M} lines for ν_{M} < 1 and ν_{M} > 1 are given by the red and the blue lines. The big external white region in the topright corner in each case corresponds to the domain where , the wave being thus vertically trapped there. The MHD TA applies when ν_{M} < 1 and (red lines). In the other case (blue and black lines, i.e.  ν_{M}  ≥ 1 and ), it does not apply due to the singularity at the critical latitude (θ_{c}) given in Eq. (88) and plotted in Fig. (7). Bottom: same for m = 2 and m = −2. 

Open with DEXTER  
In the text 
Fig. 7 Top: θ_{c} as a function of ν_{s} and Λ_{E} for nonaxisymmetric retrograde (m = 1) and prograde (m = −1) waves. As in Fig. (6), the large white topright corner region corresponds to the domain where and where the waves are trapped in the vertical direction. In the white region at the left, the MHD TA applies (i.e. ν_{M} < 1 and ) and there is no critical latitude. Finally, the blue θ_{c}iso lines correspond to the regime where the TA does not apply (i.e. ν_{M} ≥ 1 and ). Middle: same for m = 2 and m = −2. Bottom: cartoon for the wave propagation region: traditional elliptic waves are regular for every latitude and live in the whole sphere (E), while hyperbolic ones are trapped in the equatorial belt where θ ∈ [θ_{c},π − θ_{c}] (H). 

Open with DEXTER  
In the text 