betaFeedback

WorksPhysics and Astronomy

Wall-attached convection under strong inclined magnetic fields

Abstract

We employ a linear stability analysis and direct numerical simulations to study the characteristics of wall-modes in thermal convection in a rectangular box under strong and inclined magnetic fields. The walls of the convection cell are electrically insulated. The stability analysis assumes periodicity in the spanwise direction perpendicular to the plane of the homogeneous magnetic field. Our study shows that for a fixed vertical magnetic field, the imposition of horizontal magnetic fields results in an increase of the critical Rayleigh number along with a decrease in the wavelength of the wall modes. The wall modes become tilted along the direction of the resulting magnetic fields and therefore extend further into the bulk as the horizontal magnetic field is increased. Once the modes localized on the opposite walls interact, the critical Rayleigh number decreases again and eventually drops below the value for onset with a purely vertical field. We find that for sufficiently strong horizontal magnetic fields, the steady wall modes occupy the entire bulk and therefore convection is no longer restricted to the sidewalls. The above results are confirmed by direct numerical simulations of the nonlinear evolution of magnetoconvection.

Introduction

Buoyancy-driven flows of electrically conducting fluids under the influence of magnetic fields are a common occurrence in geophysical, astrophysical, as well as in several technological applications. Such flows are called magnetoconvection and their driving mechanism is the temperature dependence of the fluid density, which results in spatial density variations leading to buoyancy forces acting on the fluid. When such a fluid moves under the influence of magnetic fields, electric currents are induced in the fluid due to Faraday’s law, which, in turn, induce magnetic fields by the virtue of Ampere’s law. These electric currents interact with the applied and induced magnetic fields to generate a Lorentz force distribution that acts on the fluid. Therefore, magnetoconvective flows are governed by equations of conservation of mass, momentum, and thermal energy, along with Maxwell’s equations for electromagnetism and Ohm’s law [41]. Magnetoconvection is encountered in the Sun, stars, and planetary dynamos. In industries and technological applications, magnetoconvection is typically encountered in liquid-metal batteries [18, 33, 19], cooling liquid-metal blankets in fusion reactors [26, 27], and magnetic stirring and braking of liquid metal melts [12, 13, 24].

A simplified paradigm for magnetoconvection consists of a fluid layer that is heated from below and cooled from above (Rayleigh-Bénard convection or RBC) with imposed magnetic fields in different configurations. Typically, the Boussinesq approximation is employed for modelling the above flows; this approximation assumes that the flow is incompressible and the density variations are negligible except in the buoyancy term in the momentum equation [9, 23, 10, 37]. Magnetoconvection is governed by the following nondimensional parameters: i) Rayleigh number RaRa – the ratio of buoyancy to dissipative forces, ii) Prandtl number PrPr – the ratio of kinematic viscosity to thermal diffusivity, iii) Hartmann number HaHa – the ratio of Lorentz to viscous forces, and iv) the magnetic Prandtl number PmPm – the ratio of kinematic viscosity to magnetic diffusivity. The important nondimensional output parameters of magnetoconvection are i) the Nusselt number NuNu – the ratio of the total heat transport to the diffusive heat transport, ii) the Reynolds number ReRe – the ratio of inertial to viscous forces, and iii) the magnetic Reynolds number RmRm – the ratio of induction to diffusion of the magnetic field. In liquid-metal convection typically encountered in laboratory experiments and most industrial applications, the magnetic Reynolds number is sufficiently small such that the induced magnetic field is negligible compared to the applied magnetic field and is thus neglected in the expressions of the Lorentz force and Ohm’s law [32, 13, 38]. Such cases are referred to as quasi-static magnetoconvection where the induced magnetic field adjusts instantaneously to the changes in velocity. In the quasi-static approximation, there exists a one-way influence of the magnetic field on the flow only.

Magnetoconvection has been studied theoretically in the past (for example, Chandrasekhar [9]; Houchens et al. [16]; Busse [7]) as well as with the help of experiments (for example, Nakagawa [28]; Fauve et al. [14]; Cioni et al. [11]; Aurnou & Olson [3]; Burr & Müller [5]; King & Aurnou [20]; Vogt et al. [39, 40]; Zürner et al. [44]; Grannan et al. [15]) and numerical simulations (for example, Liu et al. [22]; Yan et al. [42]; Akhmedagaev et al. [1]; Nicoski et al. [29]; Bhattacharya et al. [4]). An application of horizontal magnetic fields causes the large-scale rolls to become quasi two-dimensional and align in the direction of the field [14, 8, 6, 43, 34, 39, 40]. These self-organized flow structures reach an optimal state wherein the heat transport and convective velocities increase significantly compared to convection without magnetic fields [40]. In contrast, strong vertical magnetic fields suppress convection [9, 11, 44, 1 b]. It must be noted that in a Rayleigh-Bénard system, convection commences only above a certain critical Rayleigh number, which is Rac≃1708Ra_c \simeq1708 for the case with infinite no-slip horizontal walls [9]. For Ra<RacRa < Ra_c, the heat transfer occurs purely by diffusion. The critical Rayleigh number increases when a vertical magnetic field is imposed and scales as Rac∼Ha2Ra_c \sim Ha^2 in the asymptotic limit of large Hartmann numbers.

The dynamics of convection under strong vertical magnetic fields become more intricate in the presence of sidewalls. Houchens et al. [16] and Busse [7] analytically showed that magnetoconvection near the sidewalls ceases at Rayleigh numbers below the ones required to completely suppress convection in the bulk. Several numerical and experimental studies on magnetoconvection with sidewalls have also revealed the presence of residual wall-attached convection at Ra<RacRa < Ra_c (Houchens et al. 2002; Liu et al. 2018; Akhmedagaev et al. 2020a,b; Zürner et al. 2020; Teimurazov et al. 2023; McCormack et al. 2023). These so-called wall modes were shown to exhibit a two-layered structure and become more closely attached to the sidewalls with the increase of Hartmann number [22].

There are a few studies only on convection with inclined magnetic fields which motivates the present work [17, 29]. The results of Hurlburt et al. [17] indicate that the mean flows tend to travel in the direction of the tilt. Nicoski et al. [29] observed qualitative similarities between convection with inclined magnetic field and that with vertical magnetic field in terms of the behavior of convection patterns, heat transport, and flow speed. However, to the best of our knowledge, there are no studies for the case with inclined magnetic fields where the Rayleigh number is close to but less than the critical Rayleigh number. Therefore, in the present work, we study thermal magnetoconvection in the wall-attached convection regime and explore the effects of additional horizontal magnetic fields on the wall modes. We use a combination of linear stability analysis and direct numerical simulations to study the dependence of the horizontal magnetic field strength, relative to the vertical magnetic field, on the wall-mode structures and their impact on large-scale heat and momentum transport.

The outline of the paper is as follows. In § 2, we discuss the problem setup, linear stability model, and the schemes for direct numerical simulations. The linear stability analysis and the results of direct numerical simulations are described in § 3. We conclude in § 4.

Numerical model

In this section, we discuss the mathematical model of our problem and the numerics employed for the stability analysis and direct numerical simulations. The study will be conducted under the quasi-static approximation, in which the induced magnetic field is neglected as it is very small compared to the applied magnetic field. This approximation is fairly accurate for magnetoconvection in liquid metals [13]. Further, we employ Boussinesq approximation, in which the variations in the density of the fluid are ignored except in the buoyancy term in the momentum equation. Hence, the flow is essentially treated as incompressible. The governing equations of magnetoconvection under the above approximations are as follows:

∇⋅u=0(1)\nabla\cdot\boldsymbol{u} = 0 \tag*{(1)}
∂u∂t+u⋅∇u=−∇pρ+αgTz^+ν∇2u+1ρ(j×B),(2)\frac{\partial\boldsymbol{u}}{\partial t} + \boldsymbol{u} \cdot\nabla\boldsymbol{u} = -\frac{\nabla p}{\rho} + \alpha g T \hat{\boldsymbol{z}} + \nu\nabla^2 \boldsymbol{u} + \frac{1}{\rho}(\boldsymbol{j} \times\boldsymbol{B}), \tag*{(2)}
∂T∂t+u⋅∇T=κ∇2T,(3)\frac{\partial T}{\partial t} + \boldsymbol{u} \cdot\nabla T = \kappa\nabla^2 T, \tag*{(3)}
j=σ(−∇ϕ+u×B),(4)\boldsymbol{j} = \sigma(-\nabla\phi+ \boldsymbol{u} \times\boldsymbol{B}), \tag*{(4)}
∇2ϕ=∇⋅(u×B),(5)\nabla^2 \phi= \nabla\cdot(\boldsymbol{u} \times\boldsymbol{B}), \tag*{(5)}

where u\boldsymbol{u}, j\boldsymbol{j}, pp, TT, and ϕ\phi are the fields of velocity, current density, pressure, temperature, and electrical potential respectively, and B=(Bx,By,Bz)\boldsymbol{B} = (B_x, B_y, B_z) is the applied magnetic field. In our work, the magnetic fields are inclined along yy-direction only, hence Bx=0B_x = 0. Further, ν\nu is the kinematic viscosity, κ\kappa is the thermal diffusivity, ρ\rho is the density, and σ\sigma is the electrical conductivity of the fluid. The last term in the momentum equation (2.2) is the Lorentz force density. Equation (2.4) is Ohm’s law. The Poisson equation (2.5) for the electric potential is a consequence of the charge conservation condition ∇⋅j=0\nabla\cdot\boldsymbol{j} = 0.

In the following, we will discuss how the above equations have been employed for our linear stability analysis and direct numerical simulations.

Linear stability model

We first discuss the derivation of the perturbation equations for our linear stability analysis. The equations (2.1) to (2.5) are non-dimensionalized using the cell height HH as the length scale, κ/H\kappa/H as the velocity scale, the temperature difference Δ\Delta between the two horizontal plates as the temperature scale, and BzB_z, the vertical component of the applied magnetic field. For this part of the analysis, we take the units that are typically chosen for a linear stability analysis to end with a Prandtl number-independent set of equations at the marginal stability threshold. The characteristic units in the subsequent simulation part will differ. The non-dimensionalized governing equations are as follows.

1Pr(∂u∂t+(u⋅∇)u)=−∇p+∇2u+Ra θez+Haz2 j×(ez+Rey),(6)\frac{1}{Pr}\left(\frac{\partial\mathbf{u}}{\partial t}+(\mathbf{u}\cdot\nabla)\mathbf{u}\right)=-\nabla p+\nabla^2\mathbf{u}+Ra\,\theta\mathbf{e}_z+Ha_z^2\,\mathbf{j}\times(\mathbf{e}_z+R\mathbf{e}_y), \tag*{(6)}
∂θ∂t+(u⋅∇)θ=∇2θ+uz,(7)\frac{\partial\theta}{\partial t}+(\mathbf{u}\cdot\nabla)\theta=\nabla^2\theta+u_z, \tag*{(7)}
∇⋅u=0,(8)\nabla\cdot\mathbf{u}=0, \tag*{(8)}
j=−∇ϕ+u×(ez+Rey),(9)\mathbf{j}=-\nabla\phi+\mathbf{u}\times(\mathbf{e}_z+R\mathbf{e}_y), \tag*{(9)}
∇⋅j=0.(10)\nabla\cdot\mathbf{j}=0. \tag*{(10)}

In the above system of equations, RaRa is the Rayleigh number, PrPr is the Prandtl number, HazHa_z is the Hartmann number based on the vertical component of the magnetic field, and RR is the ratio of the horizontal to the vertical magnetic field strength. These quantities are the governing parameters for our setup and are given by

Ra=αgΔH3νκ,Pr=νκ,Haz=BzHσρν,R=ByBz.(11)Ra=\frac{\alpha g\Delta H^3}{\nu\kappa},\qquad Pr=\frac{\nu}{\kappa},\qquad Ha_z=B_zH\sqrt{\frac{\sigma}{\rho\nu}},\qquad R=\frac{B_y}{B_z}. \tag*{(11)}

The quantity θ\theta is the difference between the temperature and the linear conduction profile, i.e.,

T(x,t)=θ(x,t)−z.(12)T(x,t)=\theta(x,t)-z. \tag*{(12)}

Apart from RaRa, PrPr, HazHa_z, and RR, the dynamics are also governed by the aspect ratio Γ\Gamma, which is the ratio of the length to the height of the convection cell.

We examine a total of 12 cases of magnetoconvection in a bounded, horizontally-extended domain of dimension Γ×1\Gamma\times1 in the yy-zz plane. Two aspect ratios are considered: Γ=2\Gamma=2 and Γ=4\Gamma=4. For Γ=2\Gamma=2, we consider the cases of Haz=50Ha_z=50 and Haz=100Ha_z=100, whereas for Γ=4\Gamma=4, we consider only the case of Haz=50Ha_z=50. For each HazHa_z analysed in our study, we vary RR from 0 (corresponding to a purely vertical magnetic field) to 3 in steps of 1. The convection cell is periodic in xx-direction and consists of no-slip horizontal walls at z=±1/2z=\pm1/2 and two no-slip sidewalls at y=±Γ/2y=\pm\Gamma/2. All the walls are electrically insulated. Each horizontal wall is at a constant temperature, with the bottom wall at T=0.5T=0.5 and the top wall at T=−0.5T=-0.5. The sidewalls are thermally insulated with ∂T/∂η=0\partial T/\partial\eta=0, where η\eta is the direction normal to the wall. A sketch of our setup is shown in Figure 1.

Sketch of the Rayleigh-Bénard convection setup with inclined magnetic field

Figure 1. A sketch of the Rayleigh-Bénard convection setup with inclined magnetic field employed for our linear stability analysis.

In the present work we assume that the instability is of stationary type. The nonlinear terms as well as the time derivatives in equations (2.6) and (2.7) are therefore neglected. The momentum and continuity equations reduce to the Stokes problem with additional buoyancy and Lorentz force. To avoid complications stemming from the coupling between pressure and velocity we choose a representation for the velocity that satisfies the continuity equation automatically and eliminate the pressure term. The velocity field is written as the curl of a vector streamfunction ψ\boldsymbol{\psi}

u=∇×ψ,(13)\mathbf{u}=\nabla\times\boldsymbol{\psi}, \tag*{(13)}

and the gauge condition ∇⋅ψ=0\nabla\cdot\boldsymbol{\psi} = 0 is imposed to determine ψ\boldsymbol{\psi} uniquely as in Priede et al. (2010). The dependence on xx is represented by the normal mode ansatz with wavenumber β\beta for all fields, e.g. θ(y,z)exp⁡(iβx)\theta(y,z)\exp(i\beta x) for the temperature perturbation. The gauge condition allows one to express the xx-component of ψ\boldsymbol{\psi} by

iβψx(y,z)=−∂yψy(y,z)−∂zψz(y,z).(14)i\beta\psi_x(y,z)=-\partial_y\psi_y(y,z)-\partial_z\psi_z(y,z). \tag*{(14)}

The velocity components then read

ux=∂yψz−∂zψy,(15)u_x=\partial_y\psi_z-\partial_z\psi_y, \tag*{(15)}
uy=−iβψz+iβ(∂y∂zψy+∂z2ψz),(16)u_y=-i\beta\psi_z+\frac{i}{\beta}\left(\partial_y\partial_z\psi_y+\partial_z^2\psi_z\right), \tag*{(16)}
uz=iβψy−iβ(∂y2ψy+∂y∂zψz).(17)u_z=i\beta\psi_y-\frac{i}{\beta}\left(\partial_y^2\psi_y+\partial_y\partial_z\psi_z\right). \tag*{(17)}

Equations for ψy\psi_y and ψz\psi_z are obtained by taking the curl of the definition (13) and the momentum equation (6). They are

0=∇2ψy+ωy,(18)0=\nabla^2\psi_y+\omega_y, \tag*{(18)}
0=∇2ψz+ωz,(19)0=\nabla^2\psi_z+\omega_z, \tag*{(19)}
0=∇2ωy−iβRa θ+Haz2(−∂y∂zϕ−∂zux+R(−∂y2ϕ+ωz−iβuy)),(20)0=\nabla^2\omega_y-i\beta Ra\,\theta+Ha_z^2\left(-\partial_y\partial_z\phi-\partial_z u_x+R\left(-\partial_y^2\phi+\omega_z-i\beta u_y\right)\right), \tag*{(20)}
0=∇2ωz+Haz2(−∂z2ϕ+R(−∂y∂zϕ+ωy+iβuz+R∂yux)).(21)0=\nabla^2\omega_z+Ha_z^2\left(-\partial_z^2\phi+R\left(-\partial_y\partial_z\phi+\omega_y+i\beta u_z+R\partial_y u_x\right)\right). \tag*{(21)}

The quantities ωy\omega_y and ωz\omega_z are the yy- and zz-components of the vorticity field ∇×u\nabla\times\boldsymbol{u}. Equations for the remaining quantities are

0=∇2θ+uz,(22)0=\nabla^2\theta+u_z, \tag*{(22)}
0=∇2ϕ−ωz−Rωy.(23)0=\nabla^2\phi-\omega_z-R\omega_y. \tag*{(23)}

The last equation (23) is obtained by substitution of Ohm’s law (9) into (10). Combined with boundary conditions on the top wall, bottom wall and side walls specified below, equations (18)–(23) represent a linear eigenvalue problem for the Rayleigh number RaRa that must be solved numerically. A suitable discretization of this problem is obtained by a spectral collocation method with Chebyshev polynomials Tn(z)=T_n(z)= cos⁡{narccos⁡(z)}\cos\{n\arccos(z)\}. The scalar fields such as θ\theta are expanded as

θ(y,z)=∑i∑kθikTi(2y/Γ)Tk(2z),(24)\theta(y,z)=\sum_i\sum_k \theta_{ik}T_i(2y/\Gamma)T_k(2z), \tag*{(24)}

where −Γ/2≤y≤Γ/2-\Gamma/2 \le y \le\Gamma/2 and −1/2≤z≤1/2-1/2 \le z \le1/2. The Poisson equations (2.20–2.23) and boundary conditions are imposed pointwise at the Gauss-Lobatto collocation points

yj=Γcos⁡(jπ/Ny)/2(0≤j≤Ny),zk=cos⁡(kπ/Nz)/2(0≤k≤Nz),(25)y_j=\Gamma\cos(j\pi/N_y)/2 \quad(0\le j\le N_y),\qquad z_k=\cos(k\pi/N_z)/2 \quad(0\le k\le N_z), \tag*{(25)}

where Ny+1N_y+1 and Nz+1N_z+1 are the number of expansion terms with respect to yy and zz.

The boundary conditions for the vector stream function and vorticity components are determined with the help of equations (2.15)–(2.17). Zero normal velocity on the horizontal walls requires ψy=0\psi_y=0 and ∂zψz=0\partial_z\psi_z=0. On the y=±Γ/2y=\pm\Gamma/2 sidewalls, the corresponding conditions are ψz=0\psi_z=0 and ∂yψy=0\partial_y\psi_y=0. The tangential velocity vanishes on the sidewalls if ωy=0\omega_y=0 and ∂yψz=∂zψy\partial_y\psi_z=\partial_z\psi_y. On the top and bottom walls these conditions are ωz=0\omega_z=0 and ∂yψz=∂zψy\partial_y\psi_z=\partial_z\psi_y. The remaining boundary conditions for (22) are θ=0\theta=0 on the top and bottom walls and ∂yθ=0\partial_y\theta=0 on the y=±Γ/2y=\pm\Gamma/2 sidewalls. The boundary condition for the electric potential supplementing equation (23) is the homogeneous Neumann condition.

Since the boundary conditions (zero normal velocity) for (18) and (19) only involve ψy\psi_y and ψz\psi_z, respectively, one can represent the expansion coefficients of ψy\psi_y and ψz\psi_z by linear invertible maps through those of ωy\omega_y and ωz\omega_z (assuming the latter are augmented by the zero boundary values to be imposed on either ψy\psi_y, ψz\psi_z or its normal derivatives). The expansions for ψy\psi_y and ψz\psi_z therefore contain Ny+3N_y+3 and Nz+3N_z+3 terms, respectively. The values of ψy\psi_y, ψz\psi_z or its derivatives in equations (2.20)–(2.23) (and associated boundary conditions) at the collocation points are represented through expansion coefficients of ωy\omega_y and ωz\omega_z via these linear invertible maps. The same can be done for the electric potential, which is the sum of a contribution from ωz\omega_z and ωy\omega_y. As a result of the collocation approximation one obtains a vector YY of unknowns containing the expansion coefficients of ωy\omega_y, ωz\omega_z, θ\theta with a size of 3(Ny+1)(Nz+1)3(N_y+1)(N_z+1) and a generalized linear eigenvalue problem

AY=Ra BY.(26)AY=Ra\,BY. \tag*{(26)}

The method was implemented in Matlab [36] using the default double precision. Notice that ωy\omega_y, ωz\omega_z and ϕ\phi are real variables. According to equations (2.15–2.17) and (2.20–2.22), uyu_y, uzu_z and θ\theta would be purely imaginary quantities. They are considered as real variables in the code and below. Problem (26) was solved with Matlab’s eig routine to find all eigenvalues and eigenvectors. The routine also works with a matrix B\mathcal{B} whose rank is smaller than the rank of A\mathcal{A} (as it is the case for (26)). It associates the spurious solutions that stem from equations not containing the eigenvalue RaRa with infinite eigenvalues.

For the cases studied in this paper, the numerical resolution was typically Ny=70N_y=70, Nz=60N_z=60 for Haz=50Ha_z=50 and Ny=90N_y=90, Nz=70N_z=70 for Haz=100Ha_z=100 with aspect ratio Γ=2\Gamma=2. A decrease of NyN_y by 10 typically only resulted in a relative change of the first eigenvalue below 10−510^{-5}. The generation of the matrices and the solution of problem (26) for a given wavenumber took about 20 hours for Ny=90N_y=90, Nz=70N_z=70 and about 6 hours for Ny=70N_y=70, Nz=60N_z=60 on an Intel Xeon E5 CPU.

Direct numerical simulations

We conduct direct numerical simulations (DNS) of our magnetoconvection setup using a second-order finite difference code developed by Krasnov et al. [21]. The governing equations are made dimensionless by using the cell height HH, the imposed temperature difference Δ\Delta, and the free-fall velocity U=αgΔHU=\sqrt{\alpha g\Delta H} (where gg and α\alpha are respectively the gravitational acceleration and the volumetric coefficient of thermal expansion of the fluid). The following non-dimensional equations are employed for our DNS:

∇⋅u=0(27)\nabla\cdot\mathbf{u} = 0 \tag*{(27)}
∂u∂t+u⋅∇u=−∇p+Tz^+PrRa∇2u+Haz2PrRa(j×B),(28)\frac{\partial\mathbf{u}}{\partial t} + \mathbf{u} \cdot\nabla\mathbf{u} = -\nabla p + T\hat{\mathbf{z}} + \sqrt{\frac{Pr}{Ra}}\nabla^2\mathbf{u} + Ha_z^2\sqrt{\frac{Pr}{Ra}}(\mathbf{j} \times\mathbf{B}), \tag*{(28)}
∂T∂t+u⋅∇T=1RaPr∇2T,(29)\frac{\partial T}{\partial t} + \mathbf{u} \cdot\nabla T = \frac{1}{\sqrt{RaPr}}\nabla^2 T, \tag*{(29)}
j=−∇ϕ+(u×B),(30)\mathbf{j} = -\nabla\phi+ (\mathbf{u} \times\mathbf{B}), \tag*{(30)}
∇2ϕ=∇⋅(u×B),(31)\nabla^2 \phi= \nabla\cdot(\mathbf{u} \times\mathbf{B}), \tag*{(31)}

We numerically solve equations (27) to (31). The Prandtl number PrPr is chosen to be 0.025, which is the same as that of mercury. We choose two Hartmann numbers based on vertical magnetic field: Haz=100Ha_z = 100 and Haz=200Ha_z = 200. For these Hartmann numbers, the corresponding critical Rayleigh numbers (RacRa_c) for the case without horizontal walls obtained using the linear stability analysis of Chandrasekhar [9] is:

Rac,∞={1.245×105,Haz=100,4.48×105,Haz=200.(32)Ra_{c,\infty} = \begin{cases} 1.245 \times10^5, & Ha_z = 100,\\ 4.48 \times10^5, & Ha_z = 200. \end{cases} \tag*{(32)}

Since we are interested in the wall-attached convection regime, we choose the Rayleigh number to be slightly below the critical Rayleigh numbers given by (32). Thus, the chosen Rayleigh numbers are Ra=105Ra = 10^5 for the case of Haz=100Ha_z = 100 and Ra=4×105Ra = 4 \times10^5 for the case of Haz=400Ha_z = 400. For each HazHa_z, we vary RR from 0 to 3.

We choose the domain-size to be Γ×Γ×H=4×4×1\Gamma\times\Gamma\times H = 4 \times4 \times1 and employ a grid-resolution ranging from 1200 ×\times 1200 ×\times 300 points to 1800 ×\times 1800 ×\times 400. The horizontal walls are at z=±1/2z = \pm1/2 and the sidewalls are at x=±Γ/2x = \pm\Gamma/2 and y=±Γ/2y = \pm\Gamma/2. The mesh is non-uniform with stronger clustering of the grid points near the top and bottom boundaries. The elliptic equations for pressure, electric potential, and the temperature are solved based on applying cosine transforms in xx- and yy-directions and using a tridiagonal solver in the zz-direction. The diffusive term in the temperature transport equation is treated implicitly. The time discretization of the momentum equation uses the fully explicit Adams–Bashforth/Backward-Differentiation method of second order (Peyret [30]). A constant time step size ranging from 5×10−55 \times10^{-5} to 1×10−41 \times10^{-4} free fall time unit was chosen for our simulations, which satisfied the Courant–Friedrichs–Lewy (CFL) condition for our runs.

All the walls are rigid and electrically insulated such that the electric current density j\mathbf{j} forms closed field lines inside the cell. The top and bottom walls are held fixed at T=−0.5T = -0.5 and T=0.5T = 0.5 respectively, and the sidewalls are adiabatic with ∂T/∂η=0\partial T/\partial\eta= 0 (where η\eta is the component normal to sidewall). All the simulations are initialized with the linear conduction profile for temperature (which is a function of the zz-coordinate only) and a random noise of amplitude A=0.001A = 0.001 along the zz-direction for velocity. We run the simulations initially on a coarse grid of 120 ×\times 120 ×\times 30 points for 100 free-fall time units in which they converge to a statistically steady state. Following this, we successively refine the mesh to the required resolutions and allow the simulations to converge after each refinement. Once the simulations reach the statistically steady state at the highest resolution, they are run for another 20 to 21 free-fall time units and a snapshot of the flow field is saved after every free-fall time unit.

Since all the walls are no-slip, thin velocity boundary layers are formed adjacent to the walls. For our simulations to be well-resolved, an adequate number of gridpoints need to

RunsHazHa_zRaRaRRGrid sizenzn_znyn_ynxn_x
11001×1051 \times10^501200×1200×3001200 \times1200 \times300228989
21001×1051 \times10^50.31200×1200×3001200 \times1200 \times300223988
31001×1051 \times10^511200×1200×3001200 \times1200 \times300221579
41001×1051 \times10^521200×1200×3001200 \times1200 \times3002215120
51001×1051 \times10^531200×1200×3001200 \times1200 \times3002210106
62004×1054 \times10^501200×1200×3001200 \times1200 \times300122121
72004×1054 \times10^50.31200×1200×3001200 \times1200 \times300122121
82004×1054 \times10^511200×1200×3001200 \times1200 \times300121318
92004×1054 \times10^521800×1800×4001800 \times1800 \times400161122
102004×1054 \times10^531800×1800×4001800 \times1800 \times400161218

caption: Parameters of the simulations: the Hartmann number (HazHa_z) based on the vertical magnetic field, the Rayleigh number (RaRa), the ratio RR of the magnetic field strength along the horizontal to the vertical directions, the grid-size, the number of points in the velocity boundary layers along the horizontal walls (nzn_z), the y=±Γ/2y=\pm\Gamma/2 sidewalls (nyn_y), and the x=±Γ/2x=\pm\Gamma/2 sidewalls (nxn_x). The Prandtl number is constant in all cases at a value of Pr=0.025Pr=0.025.

:::

be present in these boundary layers. It must be noted that the boundary layer profiles are strongly influenced by the magnetic fields. For a purely vertical magnetic field, these boundary layers are categorized into Hartmann layers adjacent to the top and bottom walls and Shercliff layers adjacent to the sidewalls. The thickness of Hartmann layers is given by δH=1/Haz\delta_H=1/Ha_z and that for the Shercliff layers is given by δS=1/Haz\delta_S=1/\sqrt{Ha_z}. However, in our case, the magnetic field is inclined with respect to the vertical direction; therefore, both the horizontal walls and y=±Γ/2y=\pm\Gamma/2 sidewalls will have a mix of Hartmann and Shercliff layers. On the other hand, the x=±Γ/2x=\pm\Gamma/2 sidewalls will have purely Shercliff layers. Thus, we use δ\delta only for representing the boundary-layer thickness. For a conservative analysis, we estimate the thicknesses of these boundary layers as follows.

δ={min⁡(1Haz,1RHaz),Horizontal walls,min⁡(1Haz,1RHaz),y=±Γ/2 sidewalls,1(Haz2(1+R2))1/4,x=±Γ/2 sidewalls.(33)\delta= \begin{cases} \min\left(\frac{1}{Ha_z},\frac{1}{\sqrt{RHa_z}}\right), & \text{Horizontal walls},\\ \min\left(\frac{1}{\sqrt{Ha_z}},\frac{1}{RHa_z}\right), & y=\pm\Gamma/2\text{ sidewalls},\\ \frac{1}{\left(Ha_z^2(1+R^2)\right)^{1/4}}, & x=\pm\Gamma/2\text{ sidewalls}. \end{cases} \tag*{(33)}

Table 1 lists the important parameters of our simulation runs. In this table, we also report the number of points in the different velocity boundary layers. We ensure that a minimum of 10 points is present in the boundary layers so as to adequately resolve our simulation runs.

In the next section, we will discuss the results obtained from our stability analysis and numerical simulations.

Results

In this section, we make a detailed analysis of the structure of the wall-modes and their impact on the heat and momentum transport. We will first present the linear stability analysis which is followed by a discussion of the results obtained from our direct numerical simulations.

Neutral stability curves for magnetoconvection with different values of $R$

Figure 2. Neutral stability curves for magnetoconvection with different values of RR for (a) Γ=2\Gamma= 2, Haz=50Ha_z = 50; (b) Γ=4\Gamma= 4, Haz=50Ha_z = 50; and (c) Γ=2\Gamma= 2, Haz=100Ha_z = 100.

Linear stability analysis

In this subsection, we discuss the results from our linear stability model. We assume that the set of finite positive eigenvalues of problem (2.26) is sorted in ascending order, i.e. the smallest one will be referred to as the first eigenvalue etc.

In figures 2(a-c), we plot the first eigenvalue, denoted as Rayleigh number RaRa, at which the instability sets in due to disturbances at wavenumber β\beta for different values of RR. The minimum value of RaRa is the critical Rayleigh number RacRa_c, and the wavenumber corresponding to RacRa_c is the most unstable wavenumber. We also plot RaRa for the infinite plane layer for the corresponding HazHa_z, where the flow is assumed to be uniform in the yy-direction. The figures show that RacRa_c is smaller than that corresponding to the plane infinite layer (Rac,∞Ra_{c,\infty}) for all RR, HazHa_z, and aspect ratios considered in this study. This clearly implies that sidewalls destabilize the magnetoconvection system. Figure 3(a) exhibits the dependence of RacRa_c on RR for different aspect ratios and strengths of the vertical magnetic field. It is evident from the above figure that for R<1R < 1, the critical Rayleigh number increases for all aspect ratios. For the larger aspect ratio box (Γ=4\Gamma= 4), RacRa_c continues to increase with RR beyond R=1R = 1 and saturates at R=2R = 2. On the other hand, for the smaller aspect ratio box, the critical Rayleigh number starts to decrease with RR for R>1R > 1, a trend that does not seem to depend on HazHa_z.

Plots of the critical Rayleigh number and wavelength

Figure 3. For Haz=50Ha_z = 50, Γ=2\Gamma= 2 (red squares); Haz=50Ha_z = 50, Γ=4\Gamma= 4 (blue circles); and Haz=100Ha_z = 100, Γ=2\Gamma= 2 (black triangles): (a) critical Rayleigh number RacRa_c, normalized with the same for infinite horizontal layer, and (b) the wavelength of the most unstable mode at RacRa_c.

In Figure 3(b), we exhibit the plots of the most unstable wavelength λ\lambda versus RR. The most unstable wavelength is calculated as

λ=2πβc,\lambda= \frac{2\pi}{\beta_c},

where βc\beta_c is the most unstable wavenumber. The figure shows that λ\lambda, like RacRa_c, exhibits a non-monotonic variation with RR and lies between 1.1 and 1.6 for the entire range of parameters considered in our study. For small values of RR, λ\lambda decreases with RR and for the larger aspect-ratio box, it saturates at R=2R = 2. For the smaller aspect-ratio box, λ\lambda increases with increasing RR beyond R=2R = 2.

We now examine the behavior of the first three eigenvalues (RaRa) for our cases. In Figure 4(a)–(Figure 4l), we plot the variations of these eigenvalues with β\beta. The neutral stability curve for the infinite plane convection layer for the corresponding HazHa_z is also shown in dashed curves. For R=0R = 0, it can be seen in Figure 4(a,Figure 4e,Figure 4i) that the minima of the third eigenvalue (represented by green triangles) overlaps with that of the plane convection layer. This indicates that the third eigenvalue corresponds to the onset of bulk convection for different wavenumbers. The figures also indicate that minima of the third eigenvalue, which corresponds to the critical Rayleigh number for convection in the bulk, increases with RR. This indicates that the bulk convection is further suppressed as the horizontal magnetic field increases.

Dependence of the first three eigenvalues on the wavenumber $\beta$ for different aspect ratios, magnetic-field ratios, and vertical Hartmann numbers

Figure 4. Dependence of the first three eigenvalues on the wavenumber β\beta for Γ=2\Gamma= 2 and Haz=50Ha_z = 50 with (a) R=0R = 0, (b) R=1R = 1, (c) R=2R = 2, and (d) R=3R = 3; Γ=4\Gamma= 4 and Haz=50Ha_z = 50 with (e) R=0R = 0, (f) R=1R = 1, (g) R=2R = 2, and (h) R=3R = 3; and Γ=2\Gamma= 2 and Haz=100Ha_z = 100 with (i) R=0R = 0, (j) R=1R = 1, (k) R=2R = 2, and (l) R=3R = 3. Also shown is the neutral stability curve (dashed blue lines) for the corresponding infinite horizontal convection layer (Γ=∞\Gamma= \infty) with a purely vertical magnetic field (R=0R = 0). For the infinite plane layer, the flow is assumed to be uniform in the yy-direction.

The first and second eigenvalues (denoted by black squares and red circles respectively) correspond to wall-attached convection, and their minima are less than that for the third eigenvalue (corresponding to bulk convection). These eigenvalues nearly overlap for small horizontal magnetic fields but begin to diverge at large horizontal magnetic fields. These variations are shown more explicitly in Figure 5(a,Figure 5b) in which we exhibit the variations of the first two eigenvalues with RR for β=4\beta= 4 (near the most unstable wavenumber). It is evident from these figures that the eigenvalues diverge visibly for R>Rc≈1.5R > R_c \approx1.5 for Γ=2\Gamma= 2 and for R>Rc≈2.5R > R_c \approx2.5 for Γ=4\Gamma= 4. It is interesting to note that the value of RR for which the eigenvalues begin to diverge does not seem to depend on HazHa_z.

Influence of aspect ratio and vertical Hartmann number on the stability curves

Figure 5. Influence of (a) aspect ratio and (b) vertical Hartmann number on the stability curves for 1st and 2nd eigenvalues as function of the ratio of the horizontal to vertical magnetic fields at fixed wavenumber β=4\beta= 4.

We will now examine the spatial structure of the wall-modes, i.e. the eigenfunctions that correspond to the first and second eigenvalues. The eigenfunctions are normalized such that the maximum of the vertical velocity becomes equal to unity.

In Figure 6(a)–(Figure 6p), we exhibit the contour plots of the temperature perturbation θ\theta on the vertical yy-zz plane corresponding to the first and second eigenvalues for β=4\beta= 4, Γ=2\Gamma= 2, and different HazHa_z and RR. The first and third rows of the figure correspond to the first eigensolutions for Haz=50Ha_z = 50 and Haz=100Ha_z = 100, respectively, whereas the second and fourth rows correspond to the second eigensolutions. The figures show that as the horizontal magnetic field strength is increased, the wall modes get elongated and tilt along the direction of the resultant magnetic field. In fact, for R≥2R \geq2, the wall modes are no longer confined adjacent to the sidewalls; instead, they occupy almost the entire bulk and interact with each other as explained later in this section.

Eigenvector field from the linear stability analysis

Figure 6. Eigenvector field from the linear stability analysis. Contours of the temperature deviation θ\theta are shown. For β=4\beta= 4, Γ=2\Gamma= 2, and Haz=50Ha_z = 50: isocontours of θ\theta corresponding to the first solution for (a) R=0R = 0, (b) R=1R = 1, (c) R=2R = 2, and (d) R=3R = 3. For β=4\beta= 4, Γ=2\Gamma= 2, and Haz=50Ha_z = 50: isocontours of θ\theta corresponding to the second solution for (e) R=0R = 0, (f) R=1R = 1, (g) R=2R = 2, and (h) R=3R = 3. For β=4\beta= 4, Γ=2\Gamma= 2, and Haz=100Ha_z = 100: isocontours of θ\theta corresponding to the first solution for (i) R=0R = 0, (j) R=1R = 1, (k) R=2R = 2, and (l) R=3R = 3. For β=4\beta= 4, Γ=2\Gamma= 2, and Haz=100Ha_z = 100: isocontours of θ\theta corresponding to the second solution for (m) R=0R = 0, (n) R=1R = 1, (o) R=2R = 2, and (p) R=3R = 3. Also shown are the vector plots of the corresponding magnetic fields.

Figure 6(a)–(Figure 6p) also show that two there are two types of spatial structures corresponding to the first two eigensolutions. The first type consists of a hot plume (θ>0\theta> 0) adjacent to the y=−Γ/2y = -\Gamma/2 sidewall and a cold plume (θ<0\theta< 0) adjacent to y=Γ/2y = \Gamma/2 sidewall; the corresponding eigensolution will be referred to as antisymmetric solution, see panel (a). The second type consists of cold (or hot) plumes adjacent to both the sidewalls; the corresponding eigensolution will be referred to as symmetric solution, see panel (e). For R<1.5R < 1.5, there is only a marginal difference between the symmetric (RasRa_s) and antisymmetric (RaaRa_a) eigenvalues. As exhibited in Figure 7, this difference is less than 0.5%0.5\% for R<0.8R < 0.8 and is approximately 10%10\% at R=1.5R = 1.5 for the case of Haz=50Ha_z = 50 and Γ=2\Gamma= 2. Thus, for R<1.5R < 1.5, there is nearly an equal preference for symmetric and antisymmetric structures to develop at the onset of convection. However, these eigenvalues deviate significantly once RR exceeds the threshold Rc≈1.5R_c \approx1.5, above which the eigenvalue corresponding to the symmetric solution becomes significantly smaller. This implies that there is a stronger preference for the symmetric structures to develop at the onset of convection for R>RcR > R_c. In this regime of RR, the symmetric eigensolution comprises of a large plume developed by the merging of two cold (or hot) plumes adjacent to the opposite walls as visible in Figure 6(c), (Figure 6d), (Figure 6k), and (Figure 6l). On the other hand, as seen in Figure 6(g), (Figure 6h), (Figure 6o), and (Figure 6p), the antisymmetric eigensolutions for R>RcR > R_c comprise of a cold plume extending on the top of a hot plume (or vice-versa) extended from the opposite wall, resulting in two convection rolls on top of each other. Our observations

Ratio of the difference between the critical Rayleigh numbers for antisymmetric and symmetric solutions

Figure 7. Ratio of the difference between the critical Rayleigh numbers for antisymmetric solution (RaaRa_a) and the symmetric solution (RasRa_s) to the critical Rayleigh number of the symmetric solution. The inset exhibits a magnified version of the above plot with RR ranging from 0 to 0.8. Parameters are Haz=50Ha_z = 50, Γ=2\Gamma= 2 and β=4\beta= 4.

imply that a clear separation of the symmetric and antisymmetric solutions occurs when the wall modes adjacent to the opposite walls for symmetric solutions start interacting with each other. The merging of the plumes and the resultant formation of a merged roll results in an increased heat and momentum transport and hence in the decrease of the critical Rayleigh number.

Figure 8(a–Figure 8h) exhibit the contour plots of θ\theta on vertical yy-zz midplane corresponding to the first and second eigensolutions for Γ=4\Gamma= 4, β=4\beta= 4, and Haz=50Ha_z = 50. These figures again show that the wall modes tend to elongate along the direction of the resultant magnetic field. Similar to the Γ=2\Gamma= 2 cases, the two eigenvalues correspond to symmetric and antisymmetric solutions respectively. The wall modes are symmetric for the first eigensolution and antisymmetric for the second. It can be recalled from Figure 5(a) that the first and second eigenvalues for Γ=4\Gamma= 4 box begin to diverge at a higher value of RR compared to the Γ=2\Gamma= 2 box. A clear separation occurs near Rc≈2.5R_c \approx2.5; this is because owing to the larger aspect ratio of the box, the plumes adjacent to opposite walls are able to interact and merge only at higher tilts, and hence at larger RR. The merged plume is exhibited in Figure 8(g) which displays the contours of θ\theta corresponding to the symmetric solution for R=3R = 3.

Eigenvector field from the linear stability analysis

Figure 8. Eigenvector field from the linear stability analysis. Contours of the temperature deviation θ\theta are shown. For Haz=50Ha_z = 50, β=4\beta= 4, and Γ=4\Gamma= 4: isocontours of θ\theta corresponding to the first eigensolutions for (a) R=0R = 0, (c) R=1R = 1, (e) R=2R = 2, and (g) R=3R = 3, and corresponding to second eigensolutions for (b) R=0R = 0, (d) R=1R = 1, (f) R=2R = 2, and (h) R=3R = 3. Also shown are the vector plots of the corresponding magnetic fields.

Our analysis suggests that the nonmonotonic behaviour of RacRa_c and wavelength λ\lambda with respect to RR is due to the increasing interaction of the plumes on opposite sidewalls. As long as the opposite plumes do not merge, an increase in the horizontal magnetic field component further stabilizes the magnetoconvection system with RacRa_c increasing and the wavelength λ\lambda decreasing with RR. The system gets destabilized due to the merging of the opposite plumes and the variations of RacRa_c and λ\lambda with RR get reversed. In the next subsection, we will discuss the results of direct numerical simulations of the nonlinear evolution of magnetoconvection.

Results of direct numerical simulations

In this subsection, we analyze our steady-state DNS results and examine the structures of the wall modes, their role in the heat and momentum transport, and the effects of initial conditions on the formation of the wall modes.

Structure of the wall modes

We use our numerical data to study the spatial convection structures for Haz=100Ha_z=100 and Haz=200Ha_z=200 with RR ranging from 0 to 3.

In Figure 9(a)–(Figure 9j), we exhibit the isosurfaces of uz=±0.01u_z=\pm0.01 for all our runs. In Figure 10(a)–(Figure 10h), we display the contours of uz≥0.01u_z\ge0.01 (red) and uz≤0.01u_z\le0.01 (blue); these contours give a visualization of the upwelling and downwelling plumes respectively. These figures show the presence of wall-attached convection with suppressed fluid flow in the bulk. Consistent with the results of the stability analysis in the previous subsection, these wall modes become elongated and align themselves along the direction of the resultant magnetic field as the horizontal magnetic field is increased. Figures 9 and 10 also show that with the exception of the case with Haz=100Ha_z=100 and R=3R=3, the wall modes are antisymmetric, that is, the plumes are upwelling (or downwelling) adjacent to y=−Γ/2y=-\Gamma/2 wall and downwelling (or upwelling) adjacent to y=Γ/2y=\Gamma/2 wall. The wall modes occupy the entire bulk for R=3R=3, consistent with the stability analysis of the Γ=4\Gamma=4 box. For the above field ratio, the wall modes for the Haz=100Ha_z=100 case are symmetric and the plumes adjacent to the opposite walls merge to form a large plume as displayed in Figure 9(e) and 10(d).

Figure 9. Results of direct numerical simulations for the vertical velocity component. Isosurfaces of uz=0.01u_z=0.01 (red) and uz=−0.01u_z=-0.01 (blue) for Haz=100Ha_z=100 with (a) R=0R=0, (b) R=0.3R=0.3, (c) R=1R=1, (d) R=2R=2, and (e) R=3R=3, and for Haz=200Ha_z=200 with (f) R=0R=0, (g) R=0.3R=0.3, (h) R=1R=1, (i) R=2R=2, and (j) R=3R=3.

Figure 10. Results of direct numerical simulations for the wall modes. Contours of uz≥0.01u_z \ge0.01 (red) and uz≤−0.01u_z \le-0.01 (blue) along with the vector plots of B\mathbf{B} on x=0x=0 midplane for Haz=100Ha_z=100 with (a) R=0R=0, (b) R=1R=1, (c) R=2R=2, (d) R=3R=3, and for Haz=200Ha_z=200 with (e) R=0R=0, (f) R=1R=1, (g) R=2R=2, (h) R=3R=3. The wall modes align themselves along the direction of B\mathbf{B}.

It can also be observed in Figure 9(a–Figure 9j) that there are slight irregularities in the wall-mode structures. These irregularities seem to be associated with small temporal fluctuations (∼1%\sim1\%) in integral quantities that remain after our simulations reach an apparently stationary state. These fluctuations indicate that the wall modes are still evolving, albeit slowly. However, it is important to note that we could not observe any noticeable change in the structures of the wall modes over the time-span of our simulations. Any change in the spatial arrangement of the modes is likely to be visible only after the simulations are run for several times the diffusion time scale, which is around 150 to 200 free-fall time units. Hence our solutions can be considered to be effectively steady.

We estimate the wavelength of the wall modes by visual inspection of the vertical velocity isosurfaces in Figure 9(a)–(Figure 9j). We consider only those wall modes that are adjacent to y=±Γ/2y=\pm\Gamma/2 sidewalls; these walls are not parallel to the resultant magnetic field. The estimation of the wavelength is done as follows. We count the number of upwelling (or downwelling) plumes adjacent to both y=−Γ/2y=-\Gamma/2 and y=Γ/2y=\Gamma/2 sidewalls and calculate their average as NpN_p. The wavelength λwm\lambda_{wm} of the wall modes is given by

λwm=ΓNp.(34)\lambda_{wm}=\frac{\Gamma}{N_p}. \tag*{(34)}

We plot the estimated λwm\lambda_{wm} versus RR in figure 11. The figure shows that λwm\lambda_{wm} tends to decrease as the horizontal magnetic field increases and saturates at large RR. Further, the wavelength corresponding to Haz=200Ha_z=200 is smaller than that for Haz=100Ha_z=100 except for R=3R=3. This is in line with the fact that the threshold wavelength decreases with increasing vertical magnetic field strength [9]. It can be recalled that a similar trend was observed in the variation of the most unstable wavelength with RR in our stability analysis in § 3.1. However, it must be kept in mind that since the Rayleigh numbers considered in our DNS are higher than the threshold Rayleigh number above which the wall modes appear, there is always a chance for secondary modes with smaller wavenumbers to develop.

Heat and momentum transport

We will now explore the influence of the wall modes on the heat and momentum transport. We compute the Nusselt (NuNu) and the Reynolds numbers (ReRe) using our numerical data as follows:

Figure 11. Variations of the wavelength λwm\lambda_{wm} of the wall modes with RR. The wavelength tends to decrease with an increase in the horizontal magnetic field.

Nu=1+RaPr⟨uzT⟩V,(35)Nu = 1 + \sqrt{RaPr}\langle u_zT\rangle_V, \tag*{(35)}
Re=RaPrUrms,(36)Re = \sqrt{\frac{Ra}{Pr}}U_{rms}, \tag*{(36)}

where Urms=⟨ux2+uy2+uz2⟩VU_{rms} = \sqrt{\langle u_x^2+u_y^2+u_z^2\rangle_V} is the root mean square velocity and ⟨⋅⟩V\langle\cdot\rangle_V denotes volume averaging. The second term of the right hand side of (3.2) is the normalized convective heat flux. We plot the normalized heat flux and the Reynolds number versus RR for all our runs in Figure 12(a) and (Figure 12b) respectively. The figures show that for R<1R < 1, both NuNu and ReRe decrease with RR. This is consistent with our conclusion from the stability analysis in § 3.1 that for small values of RR, an increase in the horizontal magnetic field results in the stabilization of the magnetoconvection system. There is, however, no clear trend in the variations of NuNu and ReRe for R>1R > 1.

Plots of the nondimensional convective heat flux and Reynolds number

Figure 12. Plots of (a) the nondimensional convective heat flux Nu−1Nu-1 and (b) the Reynolds number ReRe versus RR. For R≤1R \le1, the heat and momentum transport decreases as RR is increased.

Having studied the trends of the global heat and momentum transport with RR, we will now explore the spatial variation of heat and momentum transport for different regimes of RR. In Figure 13(a)–(Figure 13e), we plot twice the kinetic energy E=0.5(ux2+uy2+uz2)E = 0.5(u_x^2+u_y^2+u_z^2), averaged over the xx-zz plane, versus yy, which is the direction parallel to the horizontal magnetic field. For R=0R = 0, there are two sharp peaks close to y=−2y = -2 and y=2y = 2 for both Haz=100Ha_z = 100 and Haz=200Ha_z = 200; these peaks correspond to the wall modes. The bulk region lies in between these two peaks. The bulk consists of several smaller peaks and the kinetic energy in this region is much less compared to the near-wall regions. As RR is increased, the height of the near-wall peaks decreases and of those in the bulk increases. Thus, the distribution of kinetic energy along yy becomes more uniform as RR is increased. It is a known fact that in magnetohydrodynamic flows, the Lorentz force tends to modify the flow-field so as to minimize the gradients of velocity along the direction of the magnetic field [13]. Thus, in our case, as RR is increased, the Lorentz force generated due to the yy-component of the magnetic field becomes strong and suppresses the gradients of velocity along yy.

Spatial distribution of kinetic energy and convective heat flux

Figure 13. Spatial distribution of kinetic energy EE and the convective heat flux uzTu_zT. Plots of 2⟨E⟩x,z=⟨ux2+uy2+uz2⟩x,z2\langle E\rangle_{x,z}=\langle u_x^2+u_y^2+u_z^2\rangle_{x,z} for (a) R=0R=0, (b) R=0.3R=0.3, (c) R=1R=1, (d) R=2R=2, and (e) R=3R=3. Plots of ⟨uzT⟩x,z\langle u_zT\rangle_{x,z} for (f) R=0R=0, (g) R=0.3R=0.3, (h) R=1R=1, (i) R=2R=2, and (j) R=3R=3. In the above, ⟨⋅⟩x,z\langle\cdot\rangle_{x,z} represents averaging over xx-zz plane.

In Figure 13(f)–(Figure 13j), we plot the local convective heat flux uzTu_zT, averaged over the xx-zz plane, along yy. Although the volume-averaged heat flux is positive in thermal convection, the local heat flux for R=0R=0 fluctuates between positive and negative values as one proceeds along yy. These fluctuations even out as RR is increased, and for R=3R=3, the local heat flux remains positive throughout with very small gradients along yy. Again, this is due to the strong Lorentz forces generated by the horizontal component of the magnetic field which suppresses the gradients of velocity along the magnetic field’s direction.

In Figure 14(a)–(Figure 14e), we plot twice the kinetic energy, averaged over the yy-zz plane, versus xx, which is the direction perpendicular to the horizontal magnetic field. In the absence of horizontal magnetic field (R=0R=0), the variation of 2⟨E⟩y,z2\langle E\rangle_{y,z} versus xx is similar to that of 2⟨E⟩x,z2\langle E\rangle_{x,z} versus yy due to the symmetry of the problem. Again, there are two sharp peaks corresponding to the wall modes close to x=−2x=-2 and x=2x=2 for both Haz=100Ha_z=100 and Haz=200Ha_z=200. However, unlike ⟨E⟩x,z\langle E\rangle_{x,z}, the values of ⟨E⟩y,z\langle E\rangle_{y,z} at the peaks recede only marginally as RR is increased from 0 to 2. As the wall modes extend fully into the bulk at R=3R=3, the above peaks recede sharply with the value of 2⟨E⟩y,z2\langle E\rangle_{y,z} at these peaks being close to its values at the peaks in the bulk. It must be noted that ⟨E⟩y,z\langle E\rangle_{y,z} continues to fluctuate as it is varied with xx at R=3R=3 and does not smooth out unlike ⟨E⟩x,z\langle E\rangle_{x,z}.

Spatial distribution of kinetic energy and convective heat flux for $R=0$, $R=0.3$, $R=1$, $R=2$, and $R=3$

Figure 14. Spatial distribution of kinetic energy EE and the convective heat flux uzTu_zT. Plots of 2⟨E⟩y,z=⟨ux2+uy2+uz2⟩y,z2\langle E\rangle_{y,z}=\langle u_x^2+u_y^2+u_z^2\rangle_{y,z} for (a) R=0R=0, (b) R=0.3R=0.3, (c) R=1R=1, (d) R=2R=2, and (e) R=3R=3. Plots of ⟨uzT⟩y,z\langle u_zT\rangle_{y,z} for (f) R=0R=0, (g) R=0.3R=0.3, (h) R=1R=1, (i) R=2R=2, and (j) R=3R=3. In the above, ⟨⋅⟩y,z\langle\cdot\rangle_{y,z} represents averaging over yy-zz plane.

Figure 14(f)–(Figure 14j) exhibit the variations of the local convective heat flux uzTu_zT, averaged over the yy-zz plane, along xx. The figures show that ⟨uzT⟩y,z\langle u_zT\rangle_{y,z} fluctuates between positive and negative values. It is clear from the figures that for the Haz=100Ha_z=100 case, the spatial fluctuations increase with increasing RR. This is because as discussed before, the gradients along the yy-direction are reduced as RR is increased. Therefore, for large RR, if the gradients along zz are also small, any quantity averaged over yy-zz plane will fluctuate between the quantity’s two extremities. The case of R=3R=3 for Haz=100Ha_z=100 consists of plumes from the opposite y=±Γy=\pm\Gamma walls merged with each other; therefore, the gradients of the velocity, temperature, and heat flux along the zz-direction are small. Thus, ⟨uzT⟩y,z\langle u_zT\rangle_{y,z} exhibits strong fluctuations along the xx-direction at R=3R=3.

For Haz=200Ha_z=200 and R≤2R\leq2, ⟨uzT⟩y,z\langle u_zT\rangle_{y,z} follows a similar trend in that its fluctuations increase with increasing RR. However, for R=3R=3, its fluctuations get suppressed and ⟨uzT⟩y,z\langle u_zT\rangle_{y,z} is mostly positive. Now, let us recall that unlike in the case of Haz=100Ha_z=100, the structures for the case of Haz=200Ha_z=200 are antisymmetric. Thus, although the gradients along the yy-direction are small, there are fluctuations along the zz-direction for Haz=200Ha_z=200 due to the presence of upwelling and downwelling plumes on top of each other. An averaging over the yy-zz plane cancels out the opposing effects of the upwelling and downwelling plumes, thus resulting in the suppression of fluctuations of ⟨uzT⟩y,z\langle u_zT\rangle_{y,z}.

Finally, we analyze the contributions by the bulk and near-wall regions to the total kinetic energy and heat flux of the system. Towards this objective, we define the near-wall region as follows. For R=0R=0 and for both the Hartmann numbers, we determine NuSNu_S, which is the Nusselt number averaged over successively smaller concentric volumes S=Γ×[ry,Γ−ry]×1S=\Gamma\times[r_y,\Gamma-r_y]\times1. In this definition, ryr_y is the normal distance from the y=±Γ/2y=\pm\Gamma/2 sidewalls. We plot NuSNu_S versus ryr_y in figure 15. The figure shows that NuSNu_S initially decreases with ryr_y upto a point of local minima and then begins to increase with ryr_y. The distance between the point of the local minima and the sidewall is taken as the width δw\delta_w of the near-wall region. These widths are computed to be δw=0.37\delta_w=0.37 for Haz=100Ha_z=100 and δw=0.33\delta_w=0.33 for Haz=200Ha_z=200. Although the above values of δw\delta_w were computed for R=0R=0, these values will be assumed to hold for all RR.

The total kinetic energy E\mathcal{E} and the convective heat flux H\mathcal{H} can be expressed as the sum of their bulk and near-wall contributions. Therefore,

E=Ebulk+Enw,(37)\mathcal{E}=\mathcal{E}_{\mathrm{bulk}}+\mathcal{E}_{\mathrm{nw}}, \tag*{(37)}
H=Hbulk+Hnw.(38)\mathcal{H}=\mathcal{H}_{\mathrm{bulk}}+\mathcal{H}_{\mathrm{nw}}. \tag*{(38)}

where Ebulk\mathcal{E}_{bulk} and Enw\mathcal{E}_{nw} are respectively the bulk and near-wall contributions to the total kinetic energy, and Hbulk\mathcal{H}_{bulk} and Hnw\mathcal{H}_{nw} are respectively the bulk and near-wall contributions to the total heat flux. The total and bulk contributions to these quantities are defined as

E=12∫−Γ/2Γ/2∫−Γ/2Γ/2∫−1/21/2(ux2+uy2+uz2) dz dy dx,(39)\mathcal{E} = \frac{1}{2}\int_{-\Gamma/2}^{\Gamma/2}\int_{-\Gamma/2}^{\Gamma/2}\int_{-1/2}^{1/2}(u_x^2+u_y^2+u_z^2)\,dz\,dy\,dx, \tag*{(39)}
Ebulk=12∫−Γ/2+δwΓ/2−δw∫−Γ/2+δwΓ/2−δw∫−1/21/2(ux2+uy2+uz2) dz dy dx.(40)\mathcal{E}_{bulk} = \frac{1}{2}\int_{-\Gamma/2+\delta_w}^{\Gamma/2-\delta_w}\int_{-\Gamma/2+\delta_w}^{\Gamma/2-\delta_w}\int_{-1/2}^{1/2}(u_x^2+u_y^2+u_z^2)\,dz\,dy\,dx. \tag*{(40)}
H=∫−Γ/2Γ/2∫−Γ/2Γ/2∫−1/21/2uzT dz dy dx,(41)\mathcal{H} = \int_{-\Gamma/2}^{\Gamma/2}\int_{-\Gamma/2}^{\Gamma/2}\int_{-1/2}^{1/2}u_zT\,dz\,dy\,dx, \tag*{(41)}
Hbulk=∫−Γ/2+δwΓ/2−δw∫−Γ/2+δwΓ/2−δw∫−1/21/2uzT dz dy dx.(42)\mathcal{H}_{bulk} = \int_{-\Gamma/2+\delta_w}^{\Gamma/2-\delta_w}\int_{-\Gamma/2+\delta_w}^{\Gamma/2-\delta_w}\int_{-1/2}^{1/2}u_zT\,dz\,dy\,dx. \tag*{(42)}
Equation not extracted
(43)

We compute the relative strengths of the bulk and near-wall kinetic energies and heat fluxes using our numerical data and plot them versus RR in Figure 15(a)–(Figure 15d). For small values of RR, the bulk contribution to the total kinetic energy and heat flux is very small (less than 10%). This is expected because convection is completely suppressed in the bulk at small RR. It is clear from the figures that as RR increases, the bulk contributions to the total kinetic energy and heat flux increase. This is due to the fact that the wall modes extend further into bulk as RR increases. In fact, for R>2R > 2, the bulk and sidewall contributions become comparable to each other because the wall modes at such high values of RR extend fully into the bulk. It is also interesting to note that for Haz=100Ha_z = 100, the bulk contribution to the heat flux decreases as RR is increased from 2 to 3. The reason for this anomalous behaviour is yet to be understood.

Figure 15. Determination of the near-wall region for our analysis from the results of our DNS for R=0R = 0. Variation of the Nusselt number NuSNu_S averaged over successively smaller concentric volumes with the sidewall-normal distance ryr_y. The width of the near-wall region is given by the point of the first minima (represented as black dashed vertical line for Haz=100Ha_z = 100 and blue dashed vertical line for Haz=200Ha_z = 200) in the NuSNu_S profile

Effects of initial conditions

The direct numerical simulations of magnetoconvection for R=3R = 3 resulted in different structural arrangements of the convection plumes for Haz=100Ha_z = 100 and Haz=200Ha_z = 200. On one hand, we obtained a symmetric arrangement of the plumes adjacent to the opposite walls merging with each other for Haz=100Ha_z = 100. On the other hand, an antisymmetric arrangement of the plumes was obtained for Haz=200Ha_z = 200 with the upwelling and downwelling plumes on top of each other. In this subsection, we will explore the sensitivity of the above results to initial conditions.

Image not extracted

Figure 16. Results of direct numerical simulations: Relative strengths of the bulk and boundary layer contributions to the total kinetic energy E\mathcal{E} for (a) Haz=100Ha_z = 100 and (b) Haz=200Ha_z = 200, and to the total heat flux H\mathcal{H} for (c) Haz=100Ha_z = 100 and (d) Haz=200Ha_z = 200. The bulk contributions to kinetic energy and heat flux increase with increasing RR.

We conduct two more direct numerical simulations of magnetoconvection for R=3R = 3; one with Haz=100Ha_z = 100, Ra=105Ra = 10^{5}, and the other with Haz=200Ha_z = 200, Ra=4×105Ra = 4 \times10^{5}. However, we change the initial conditions in this set of simulations as follows. The results of the old simulation of Haz=200Ha_z = 200 are taken as the initial condition for the new simulation of Haz=100Ha_z = 100. In the same way, the results of the old simulation of Haz=100Ha_z = 100 are taken as the initial condition for the new simulation of Haz=200Ha_z = 200. Both these simulations were allowed to run for 20 free-fall time units after reaching the steady state. Henceforth, we refer to the old set of simulations as IC1 and the new set of simulations of R=3R = 3 as IC2.

We examine the structure of the wall-modes for the new set of simulations by plotting the vertical velocity isosurfaces for uz=±0.01u_z = \pm0.01 in Figure 17(a) and (Figure 17c). The figures show that the simulation of Haz=100Ha_z = 100 with the new initial conditions yields antisymmetric arrangement structures with upwelling and downwelling plumes on top of each other. This is unlike the result of the simulation with old initial conditions which resulted in a symmetric arrangement of structures with merged plumes. Similarly, the simulation of Haz=200Ha_z = 200 with the new initial conditions yield symmetric structures with merged plumes, unlike the case with old initial conditions which resulted in antisymmetric arrangement of structures. Our results indicate that that the solution of magnetoconvection with R=3R = 3 for Haz=100Ha_z = 100 and 200 is non-unique, and the arrangement of the wall modes depends on the initial conditions.

Image not extracted

Figure 17. Results of direct numerical simulations with new sets of initial conditions (IC2): Isosurfaces of uz=0.01u_z = 0.01 (red) and uz=−0.01u_z = -0.01 (blue) for R=3R = 3 and (a) Haz=100Ha_z = 100 and (c) Haz=200Ha_z = 200. For comparison, the vertical velocity isosurfaces using results of the simulations of R=3R = 3 with the old initial conditions are shown for (b) Haz=100Ha_z = 100 and (d) Haz=200Ha_z = 200

We will now explore the impact of the arrangement of the plumes on the global heat and momentum transport. In Figure 18(a,Figure 18b), we replot the computed values of Nu−1Nu - 1 and ReRe for our runs with the old initial conditions and also add the plots of these quantities computed using the results of our simulations with new initial conditions. The figures show that for Haz=200Ha_z = 200, the Reynolds and Nusselt numbers computed using the solutions of simulations with new initial conditions are significantly higher than those corresponding to the old initial conditions. On the other hand, for Haz=100Ha_z = 100, the Reynolds and Nusselt numbers computed using the solutions of simulations with new initial conditions are lower, albeit marginally, than those corresponding to the old initial conditions. The above observations clearly indicate that the solutions that are symmetric consisting of merged plumes have higher heat and momentum transport. This is because merged plumes give rise to larger convection rolls which, in turn, result in more efficient transport of heat and momentum. It is worth noting that for Haz=200Ha_z = 200, the increase of

Image not extracted

Figure 18. Results of direct numerical simulations including those using the new initial conditions (filled markers) along with the old initial conditions (unfilled markers): Plots of (a) the nondimensional convective heat flux Nu−1Nu - 1 and (b) the Reynolds number ReRe versus RR. The heat and momentum transport for the case of Haz=200Ha_z = 200, R=3R = 3 is significantly more for the new initial condition (IC2) compared to the old initial condition (IC1)

Reynolds number at R=3R = 3 due to the merged plumes is so significant that its value is larger than for R=0R = 0.

Our results indicate that the non-unique nature of the solutions for R=3R = 3 has a profound effect on the heat and momentum transport, especially for Haz=200Ha_z = 200. We conclude in the next section.

Summary and conclusions

In this paper, we systematically examined the effects of strong inclined magnetic fields on wall-attached magnetoconvection using a combination of linear stability analysis and direct numerical simulations. The linear stability analysis was conducted for lower Hartmann numbers and aspect-ratio box whereas the DNS were conducted for higher Hartmann numbers and aspect ratio. The ratio R=By/BzR = B_y/B_z of the outer imposed horizontal to vertical magnetic field was varied from 0 to 3.

The linear stability analysis revealed that the critical Rayleigh number varies non-monotonically with the relative strength of the horizontal magnetic field. The plumes at the onset of instability get elongated along the direction the resultant magnetic field and extend more into the bulk as RR is increased. At sufficiently high RR, the plumes extend fully into the bulk. The linear stability analysis further revealed the existence of non-unique solutions at the onset at low RR. One eigensolution corresponds to a symmetric arrangement of hot or cold plumes adjacent to the opposite sidewalls, whereas the other eigensolution corresponds to an antisymmetric arrangement of hot or cold plumes. For the symmetric solution, when the opposite plumes extend sufficiently into the bulk at a large RR, they merge into a single large plume. Below this critical value of R=RcR = R_c, the symmetric and antisymmetric eigensolutions overlap; however, at R>RcR > R_c, these solutions start to diverge with the symmetric eigensolution being more unstable than the antisymmetric solution. The critical Rayleigh number increases with RR for R<RcR < R_c and decreases with RR for R>RcR > R_c.

The direct numerical simulations of the fully nonlinear regime reinforced the observation from the linear stability analysis that the wall modes get elongated along the direction of the resultant magnetic field and extend further into the bulk as RR is increased. The heat and momentum transport was observed to decrease with RR for R<1R < 1 but not show any visible trend for R>1R > 1. The fluctuations of the local heat flux and kinetic energy along the direction of the horizontal magnetic field get suppressed as RR is increased. Since the wall modes extend more into the bulk as RR is increased, the relative contribution to the total heat transport and kinetic energy by the near-wall regions decrease with an increase of RR.

The analysis from direct numerical simulations further revealed that at least for R=3R = 3, the solutions are dependent on the initial conditions. The solutions for R=3R = 3 can either comprise of an antisymmetric arrangement of upwelling and downwelling plumes on top of each other, or a symmetric arrangement of merged upwelling or downwelling plumes. The solution with merged plumes corresponds to higher heat and momentum transport because the merged plumes give rise to larger convection rolls and hence more efficient heat transfer.

Our present work provides important insights into the dynamics of wall-attached convection under inclined magnetic fields which will be a typical situation in view to most applications. Our work may find applications in several industrial flows such as cooling blankets in fusion reactors. Although we worked on a small set of parameters, we expect our results to hold for higher Rayleigh numbers as well. In the future, we plan to conduct a similar analysis for fluids at different Prandtl numbers.

Acknowledgements

The authors thank M. K. Verma and M. Brynjell-Rahkola for useful discussions. The authors acknowledge the computing time provided by Leibniz Supercomputing Center, Garching, Germany, under the project ‘pn49ma’. The authors further acknowledge the Computing Center of Technische Universität Ilmenau for the resources provided to them for the linear stability computations as well as for the postprocessing and visualization of their simulation data.

Funding

The work of S.B. is sponsored by a Postdoctoral Fellowship of the Alexander von Humboldt Foundation, Germany.

Declaration of interests

The authors report no conflict of interest.

Author ORCIDs

Shashwat Bhattacharya https://orcid.org/0000-0001-7462-7680

Thomas Boeck https://orcid.org/0000-0002-0814-7432

Dmitry Krasnov https://orcid.org/0000-0002-8339-7749

Jörg Schumacher https://orcid.org/0000-0002-1359-4536

References

  1. [1]Akhmedagaev, R., Zikanov, O., Krasnov, D. & Schumacher, J. 2020a Rayleigh-Bénard convection in strong vertical magnetic field: flow structure and verification of numerical method. Magnetohydrodynamics 56, 157–166.arxiv.org/abs/2002.09404
  2. [3]Aurnou, J. M. & Olson, P. L. 2001 Experiments on Rayleigh-Bénard convection, magnetoconvection and rotating magnetoconvection in liquid gallium. J. Fluid Mech. 430, 283–307.DOI
  3. [4]Bhattacharya, S., Boeck, T., Krasnov, D. & Schumacher, J. 2023 Effects of strong fringing magnetic fields on turbulent thermal convection. J. Fluid Mech. 964, A31.DOI
  4. [5]Burr, U. & Müller, U. 2001 Rayleigh-Bénard convection in liquid metal layers under the influence of a vertical magnetic field. Phys. Fluids 13, 3247–3257.DOI
  5. [6]Burr, U. & Müller, U. 2002 Rayleigh-Bénard convection in liquid metal layers under the influence of a horizontal magnetic field. J. Fluid Mech. 453, 345–369.DOI
  6. [7]Busse, F. H. 2008 Asymptotic theory of wall-attached convection in a horizontal fluid layer with a vertical magnetic field. Phys. Fluids 20 (2), 024102.DOI
  7. [8]Busse, F. H. & Clever, R. M. 1983 Stability of convection rolls in the presence of a horizontal magnetic field. J. Theor. Appl. Mech. 2, 495–502.DOI
  8. [9]Chandrasekhar, S. 1981 Hydrodynamic and Hydromagnetic Stability. Oxford: Dover publications.
  9. [10]Chillà, F. & Schumacher, J. 2012 New perspectives in turbulent Rayleigh-Bénard convection. Eur. Phys. J. E 35 (7), 58.DOI
  10. [11]Cioni, S., Chaumat, S. & Sommeria, J. 2000 Effect of a vertical magnetic field on turbulent Rayleigh-Bénard convection. Phys. Rev. Lett. 62 (4), R4520–R4523.pubmed.ncbi.nlm.nih.gov/11089073
  11. [12]Davidson, Peter A. 1999 Magnetohydrodynamics in materials processing. Annu. Rev. Fluid Mech. 31 (1), 273–300.DOI
  12. [13]Davidson, Peter A. 2017 An Introduction to Magnetohydrodynamics, 2nd edn. Cambridge: Cambridge University Press.
  13. [14]Fauve, S., Laroche, C. & Libchaber, A. 1981 Effect of a horizontal magnetic field on convective instabilities in mercury. J. Physique Lett. 42 (21), L455.DOI
  14. [15]Grannan, A. M., Cheng, J. S., Aggarwal, A., Hawkins, E. K., Xu, Y., Horn, S., Sánchez-Álvarez, J. & Aurnou, J. M. 2022 Experimental pub crawl from Rayleigh-Bénard to magnetostrophic convection. J. Fluid Mech. 939, R1.DOI
  15. [16]Houchens, B. C., Witkowski, L. Martin & Walker, J. S. 2002 Rayleigh–Bénard instability in a vertical cylinder with a vertical magnetic field. J. Fluid Mech. 469, 189–207.DOI
  16. [17]Hurlburt, N., Matthews, P. & Proctor, M. 1996 Nonlinear compressible convection in oblique magnetic fields. Ap. J. 457, 933.DOI
  17. [18]Kelley, D. & Sadouway, D. R. 2014 Mixing in a liquid metal electrode. Phys. Fluids 26, 057102.DOI
  18. [19]Kelley, D. & Weier, T. 2018 Fluid Mechanics of Liquid Metal Batteries. Appl. Mech. Rev. 70, 020801.arxiv.org/abs/1710.03150
  19. [20]King, E. M. & Aurnou, J. M. 2015 Magnetostrophic balance as the optimal state for turbulent magnetoconvection. Proc. Natl. Acad. Sci. U.S.A. 112, 990–994.DOI
  20. [21]Krasnov, D., Zikanov, O. & Boeck, T. 2011 Comparative study of finite difference approaches in simulation of magnetohydrodynamic turbulence at low magnetic reynolds number. Comput. Fluids 50 (1), 46 – 59.DOI
  21. [22]Liu, W., Krasnov, D. & Schumacher, J. 2018 Wall modes in magnetoconvection at high Hartmann numbers. J. Fluid Mech. 849, R2.DOI
  22. [23]Lohse, Detlef & Xia, Ke-Qing 2010 Small-scale properties of turbulent Rayleigh–Bénard convection. Annu. Rev. Fluid Mech. 42 (1), 335–364.DOI
  23. [24]Lyubimov, D. V., Burnysheva, A. V., Benhadid, H., Lyubimova, T. P. & Henry, D. 2010 Rotating magnetic field effect on convection and its stability in a horizontal cylinder subjected to a longitudinal temperature gradient. J. Fluid Mech. 664, 108–137.DOI
  24. [26]Mistrangelo, C., Bühler, L. & Klüber, V. 2020 Three-dimensional magneto convective flows in geometries relevant for DCLL blankets. Fusion Eng. Des. 159, 111686.DOI
  25. [27]Mistrangelo, C., Bühler, L., Smolentsev, S., Klüber, V., Maione, I. & Aubert, J. 2021 MHD flow in liquid metal blankets: Major design issues, MHD guidelines and numerical analysis. Fusion Eng. Des. 173, 112795.DOI
  26. [28]Nakagawa, Y 1957 Experiments on the inhibition of thermal convection by a magnetic field. Proc. R. Soc. Lond. 240, 108–113.DOI
  27. [29]Nicoski, J. A., Yan, M. & Calkins, M. A. 2022 Quasistatic magnetoconvection with a tilted magnetic field. Phys. Rev. Fluids 7, 043504.DOI
  28. [30]Peyret, R. 2002 Spectral Methods for Incompressible Viscous Flows. New York: Springer.
  29. [32]Roberts, P. H. 1967 An introduction to magnetohydrodynamics. London: Longmans.
  30. [33]Shen, Y. & Zikanov, O. 2016 Thermal convection in a liquid metal battery. Theor. Comput. Fluid Dyn. 30, 275–294.DOI
  31. [34]Tasaka, Y., Igaki, K., Yanagisawa, T., Vogt, T., Zürner, T. & Eckert, S. 2016 Regular flow reversals in Rayleigh-Bénard convection in a horizontal magnetic field. Phys. Rev. E 93, 043109.DOI
  32. [36]The MathWorks Inc. 2022 Matlab version: 9.13.0 (r2022b).
  33. [37]Verma, Mahendra K. 2018 Physics of Buoyant Flows: From Instabilities to Turbulence. Singapore: World Scientific.DOI
  34. [38]Verma, M. K. 2019 Energy transfers in Fluid Flows: Multiscale and Spectral Perspectives. Cambridge: Cambridge University Press.DOI
  35. [39]Vogt, T., Ishmi, W., Yanagisawa, T., Tasaka, Y., Sakuraba, A. & Eckert, S. 2018 Transition between quasi-two-dimensional and three-dimensional Rayleigh-Bénard convection in a horizontal magnetic field. Phys. Rev. Fluids 3, 013503.
  36. [40]VOGT, T., YANG, J.-C., SCHINDLER, F. & ECKERT, S. 2021 Free-fall velocities and heat transport enhancement in liquid metal magneto-convection. J. Fluid Mech. 915, A68.DOI
  37. [41]WEISS, N. O. & PROCTOR, M. R. E. 2014 Magnetoconvection. Cambridge: Cambridge University Press.
  38. [42]YAN, M., CALKINS, M., MAFFEI, A., JULIEN, K., TOBIAS, S. & MARTI, P. 2019 Heat transfer and flow regimes in quasi-static magnetoconvection with a vertical magnetic field. J. Fluid Mech. 877, 1186–1206.
  39. [43]YANAGISAWA, T., HAMANO, Y., MIYAGOSHI, T., YAMAGISHI, Y., TASAKA, Y. & TAKEDA, Y. 2013 Convection patterns in a liquid metal under an imposed horizontal magnetic field. Phys. Rev. E 88 (6), 063020.pubmed.ncbi.nlm.nih.gov/24483570
  40. [44]ZÜ RNER, T., SCHINDLER, F., VOGT, T., ECKERT, S. & SCHUMACHER, J. 2020 Flow regimes of Rayleigh-Bénard convection in a vertical magnetic field. J. Fluid Mech. 894, A21.DOI

Paper details

vlnc:W_etga9j9x
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents