Hostname: page-component-745bb68f8f-s22k5 Total loading time: 0 Render date: 2025-01-25T08:12:30.159Z Has data issue: false hasContentIssue false

Stability analysis of partially ionized plasma in a porous medium with local thermal non-equilibrium effects

Published online by Cambridge University Press:  03 January 2025

Vishal Chandel*
Affiliation:
Department of Mathematics and Scientific Computing, National Institute of Technology Hamirpur, Hamirpur 177005, India
Sunil
Affiliation:
Department of Mathematics and Scientific Computing, National Institute of Technology Hamirpur, Hamirpur 177005, India
*
Email address for correspondence: vcvishal1950chandel@gmail.com

Abstract

This study investigates the impact of local thermal non-equilibrium on the stability analysis of partially ionized plasma within a porous medium. The plasma, heated from below, is enclosed by various combinations of bounding surfaces. Both nonlinear (via the energy method) and linear (utilizing the normal mode analysis method) analyses are performed. Eigenvalue problems for both analyses are formulated and solved using the Galerkin method. The study also explores the effects of compressibility, medium permeability and magnetic fields on system stability. The collisional frequency among plasma components and the thermal diffusivity ratio significantly influence energy decay. The results reveal that the Rayleigh–Darcy number is identical for both nonlinear and linear analyses, thus eliminating the possibility of a subcritical region and confirming global stability. The principle of exchange of stabilities is validated, indicating the absence of oscillatory convection modes. Medium permeability, heat-transfer coefficient and compressibility delay the onset of convection, demonstrating stabilizing effects. Conversely, the porosity-modified conductivity ratio hastens the convection process, indicating destabilizing effects. Rigid–rigid bounding surfaces are found to be thermally more stable for confining the partially ionized plasma. Additionally, the magnetic field exerts a stabilizing influence.

Type
Research Article
Copyright
Copyright © The Author(s), 2025. Published by Cambridge University Press

1 Introduction

Plasma, often called the fourth state of matter, stands apart from solids, liquids and gases due to its unique properties. The transition from solid to liquid to gas occurs with increasing heat, further heating of gas at sufficiently high temperatures causes its atoms to ionize, shedding the outermost electrons and resulting in a mix of positively charged ions and negatively charged electrons, known as plasma (Krishan Reference Krishan2022). Plasma can exist in various forms, categorized by the degree of ionization, with partially ionized plasma (PIP) consisting of both neutral and charged particles. Unlike fully ionized plasma, where nearly all particles are ionized, PIP allows for varied degrees of ionization depending on conditions and plasma source characteristics (Ballai Reference Ballai2019). In astrophysical contexts such as molecular clouds and the solar atmosphere, PIP plays a crucial role, with significant implications for understanding phenomena like solar flares and cometary tails (Ballester et al. Reference Ballester2018; Soler & Ballester Reference Soler and Ballester2022). These environments exhibit varying ionization levels, influencing electromagnetic processes and the plasma dynamics (Krishan Reference Krishan2022). Conversely, technological applications harness highly ionized plasmas in devices such as fusion reactors and plasma TVs. Several types of fluid instabilities, including Kelvin–Helmholtz, Rayleigh–Taylor, thermal and thermosolutal instabilities, are influenced by partial ionization effects (Soler & Ballester Reference Soler and Ballester2022). The study of PIP extends into diverse fields, including atmospheric science, plasma technology, laboratory research and astrophysics, owing to its complex behaviour and wide-ranging applications (Ballester et al. Reference Ballester, Soler, Carbonell and Terradas2021; Kumar et al. Reference Kumar, Poser, Schöttler, Kleinschmidt, Dietrich, Wicht, French and Redmer2021).

Thermal convection, a fundamental process in fluid dynamics, occurs when a fluid is heated from below, leading to a less dense lower layer than the upper layer, resulting in an unstable, top-heavy configuration. When the temperature difference or the depth of the layers overcomes the effect of gravity, the fluid ascends, revealing a cellular structure. This phenomenon, known as Bénard convection, is a significant topic in fluid dynamics and is thoroughly examined in Chandrasekhar's monograph (Chandrasekhar Reference Chandrasekhar1981). Thermal convection is crucial in various astronomical, natural and industrial processes, including atmospheric circulation, ocean flow and industrial heat transfer (Maheshwari & Bhatia Reference Maheshwari and Bhatia1976; Kaothekar Reference Kaothekar2018). Numerous studies have investigated thermal convection, contributing to our understanding of this complex phenomenon (Sharma Reference Sharma1972; Maheshwari & Bhatia Reference Maheshwari and Bhatia1976; Sharma & Sharma Reference Sharma and Sharma1978, Reference Sharma and Sharma1989; Sharma & Sunil Reference Sharma and Sunil1995, Reference Sharma and Sunil1996; Kaothekar Reference Kaothekar2018; Chandel & Sunil Reference Chandel and Sunil2024; Chandel, Sunil & Sharma Reference Chandel, Sunil and Sharma2024; Mahajan & Raj Reference Mahajan and Raj2024; Sharma, Sunil & Sharma Reference Sharma, Sunil and Sharma2024; Thakur, Kumar & Devi Reference Thakur, Kumar and Devi2024).

The phenomenon of thermal convection within porous media has numerous real-world applications, such as in oil reservoir modelling, geothermal energy utilization, building thermal insulation, food processing and nuclear water disposal (Malashetty, Swamy & Kulkarni Reference Malashetty, Swamy and Kulkarni2007; Shivakumara et al. Reference Shivakumara, Lee, Mamatha and Ravisha2011). The instability of a horizontal fluid-saturated porous layer when heated from below has been extensively researched and the growing volume of work devoted to this area is well documented by Straughan (Reference Straughan2008) and Nield & Bejan (Reference Nield and Bejan2013). A porous medium is a material containing pores (voids), where thermal convection occurs as fluid moves through these pores under a temperature gradient. In their study on the nonlinear stability of a rotating porous layer, Qin & Kaloni (Reference Qin and Kaloni1995) noted that, for highly porous materials, the Brinkman model, which accounts for the boundary layer effect, is superior to the Darcy model. This enhanced understanding of thermal convection in porous media underscores its significance in both theoretical exploration and practical applications, contributing to advancements across various scientific and industrial domains.

Studies on thermal convection in porous media heated from below often assume local thermal equilibrium (LTE), where the temperature gradient between fluid and solid phases is negligible at any location (Kuznetsov Reference Kuznetsov1998; Malashetty et al. Reference Malashetty, Swamy and Kulkarni2007). However, in many practical applications involving high-speed flows or substantial temperature differences between the fluid and solid phases, the LTE assumption proves inadequate (Kuznetsov Reference Kuznetsov1998; Malashetty et al. Reference Malashetty, Swamy and Kulkarni2007). In such scenarios, it is crucial to consider local thermal non-equilibrium (LTNE) effects by employing a two-field model for the energy equation, with separate representations for the fluid and solid phases (Kuznetsov Reference Kuznetsov1998). Local thermal non-equilibrium theory finds particular relevance in diverse applications such as food drying, freezing processes, microwave heating, rapid heat transfer in computer chips utilizing porous metal foams and heat pipe technology (Malashetty et al. Reference Malashetty, Swamy and Kulkarni2007; Shivakumara et al. Reference Shivakumara, Lee, Mamatha and Ravisha2011). These applications highlight the pivotal role LTNE theory is expected to play in future advancements.

Recent investigations have been dedicated to examining the effects of LTNE on forced and free convection in porous media. Comprehensive reviews of this research can be found in the works authored by Ingham & Pop (Reference Ingham and Pop2005), Straughan (Reference Straughan2008) and Nield & Bejan (Reference Nield and Bejan2013). Kuznetsov (Reference Kuznetsov1998) provides detailed information about thermal non-equilibrium effects on internal forced convection flows. Postelnicu & Rees (Reference Postelnicu and Rees2003) and Postelnicu (Reference Postelnicu2008) examined convection onset using a thermal non-equilibrium model, focusing on stress-free and isothermal rigid boundaries. Rees & Pop (Reference Rees and Pop2005) offer an excellent review of research on LTNE phenomena in porous medium convection, primarily free and forced convection boundary layers and free convection within cavities. Straughan (Reference Straughan2006) considered thermal convection in a fluid-saturated porous layer using a global nonlinear stability analysis with a thermal non-equilibrium model, establishing the equivalence of linear instability and nonlinear stability boundaries for thermal convection in a rotating porous layer with the Darcy law. Malashetty, Swamy & Heera (Reference Malashetty, Swamy and Heera2008) studied double-diffusive convection in a fluid-saturated porous medium when the fluid and solid phases are not in LTE, using both linear and nonlinear stability analyses. Sunil, Sharma & Mahajan (Reference Sunil, Sharma and Mahajan2010) conducted an energy stability analysis of thermo-convective magnetized ferrofluid in a porous medium under thermal non-equilibrium conditions. Shivakumara et al. (Reference Shivakumara, Lee, Mamatha and Ravisha2011) explored the effects of boundary and LTNE on the onset of convection in a sparsely packed horizontal anisotropic porous layer. Yadav & Lee (Reference Yadav and Lee2015) investigated the onset of nanofluid convection in a rotating porous layer with zero nanoparticle flux boundary conditions under LTNE effects. Bansal & Suthar (Reference Bansal and Suthar2022) and Bansal & Suthar (Reference Bansal and Suthar2024) studied temperature modulation effects on Darcy–Bénard convection using the LTNE model. Arnone, Capone & Gianfrani (Reference Arnone, Capone and Gianfrani2024) have studied the stability of penetrative convection in a Darcy–Brinkman porous medium under the hypothesis of thermal non-equilibrium.

Despite extensive research, the field remains in a much-to-be-desired state, particularly regarding the effect of LTNE on the stability of a layer of PIP saturating a porous medium. To the best knowledge of the authors, no work has yet addressed this specific problem. Investigating the impact of LTNE on the stability of PIP heated from below is crucial for improving modelling accuracy and understanding plasma stability. Such insights are pivotal for applications in astrophysics, fusion research and various industrial sectors, including spacecraft propulsion, medical technologies and environmental sciences, thereby fostering scientific and practical advancements. Given the significance and identified gaps in the literature, in this study, we undertake both nonlinear and linear analyses to explore the LTNE effect on thermal convection in compressible PIP within a porous medium enclosed by various combinations of bounding surfaces. Linear analysis is examined using the normal mode analysis method (Chandrasekhar Reference Chandrasekhar1981), while nonlinear analysis employs the energy method (Straughan Reference Straughan2004, Reference Straughan2008). For numerical analysis, the Galerkin method (Yadav, Bhargava & Agarwal Reference Yadav, Bhargava and Agarwal2013) has been employed.

The paper is structured as follows: § 2 outlines the physical problem and presents the governing equations. In § 3, we solve the governing equations for the basic state, assuming the flow to be quiescent, and introduce perturbations to the system, deriving the non-dimensional perturbation equations. Section 4 is dedicated to nonlinear analysis, including the energy decay and the formulation of the eigenvalue problem for nonlinear analysis. In § 5, we focus on linear analysis and prove the principle of exchange of stability. Section 6 details the numerical methods for solving eigenvalue problems and provides expressions for the Rayleigh–Darcy number for various bounding surface configurations. Section 7 presents the results in graphical form and discusses the outcomes in detail. Section 8 explores magnetic field effects. Finally, § 9 summarizes the major outcomes of our work.

2 Problem formulation

Consider an infinite horizontal layer of porous material saturated with compressible PIP, heated from below and confined between two surfaces at $z=0$ and $z=d$. Let $\rho _n$ denote the density of the neutral components of PIP, $\rho _i$ the density of the ionized components of PIP and $K_1$ the permeability of the porous medium. The gravitational force, $\boldsymbol {g}=( 0,0,-g )$, acts downward. A geometrical representation of the problem is illustrated in figure 1.

Figure 1. Geometrical representation of the problem.

We assume that the solid phase of the porous medium and the ionized components of the PIP are not in LTE. The temperature of the solid phase, $T_s$, and the plasma phase, $T_i$, are maintained constant at the surfaces $z=0$ and $z=d$

(2.1a,b)\begin{equation} T_s=T_i=T_0, \quad \text{at} \ z=0 \quad \text{and} \quad T_s=T_i=T_d, \quad \text{at} \ z=d, \end{equation}

where $T_0>T_d$. A two-field model is employed to separately represent the temperature fields of the solid phase of the porous medium and the ionized components of the PIP in the energy equation.

In this analysis, the PIP is treated under the continuum hypothesis, behaving as a continuous fluid. Effects of LTNE, pressure, gravity and medium permeability on the neutral components of the PIP are considered negligibly small and are thus neglected.

Assuming adherence to the Boussinesq approximation (Spiegel & Veronis Reference Spiegel and Veronis1960), the governing equations are (cf. Sharma & Sharma Reference Sharma and Sharma1978; Kuznetsov Reference Kuznetsov1998; Straughan Reference Straughan2006; Nield & Bejan Reference Nield and Bejan2013)

(2.2)\begin{gather}\frac{1}{\epsilon} \frac{\partial \boldsymbol{q}}{\partial t} ={-}\frac{1}{{\rho_{m}}}\boldsymbol{\nabla} p + \boldsymbol{g} \left[ 1-\alpha_m (T-T_{m})+K_{m}\left(p-p_{m}\right)\right] \nonumber\\ -\frac{1}{\rho_{m}}\left[ \frac{\mu }{{{K}_{1}}}-\tilde{\mu }{{\nabla}^{2}} \right]\boldsymbol{q} + \frac{{{\rho }_{d}}{{\nu }_{c}}}{{{\rho }_{m}\epsilon}}\left( \boldsymbol{q}_{d}-\boldsymbol{q} \right), \end{gather}
(2.3)\begin{gather} \epsilon\frac{\partial\rho}{\partial t}+\left(\boldsymbol{q}\boldsymbol{\cdot}\boldsymbol{\nabla}\right)\rho+\rho\boldsymbol{\nabla}\boldsymbol{\cdot} \boldsymbol{q}=0, \end{gather}
(2.4)\begin{gather} \epsilon \left(\rho c\right)_i \frac{\partial T_i}{\partial t} + \left(\rho c\right)_i \left( \boldsymbol{q}\boldsymbol{\cdot}\boldsymbol{\nabla}\right)T_i+ p \boldsymbol{\nabla}\boldsymbol{\cdot}\boldsymbol{q}= \epsilon k_i \nabla^2 T_i +h\left(T_s-T_i\right), \end{gather}
(2.5)\begin{gather}\left(1-\epsilon\right) \left(\rho c\right)_s \frac{\partial T_s}{\partial t} = \left(1-\epsilon\right) k_s \nabla^2 T_s - h\left(T_s-T_i\right), \end{gather}
(2.6)\begin{gather}\frac{\partial \boldsymbol{q}_{d}}{\partial t}+\frac{1}{\epsilon}(\boldsymbol{q}_{d} \boldsymbol{\cdot} \boldsymbol{\nabla}) \boldsymbol{q}_{d} ={-}\nu_{c}(\boldsymbol{q}_{d}-\boldsymbol{q}). \end{gather}

Here, $\epsilon$ is the porosity, $\boldsymbol {q}=(u, v, w)$ and $\boldsymbol {q}_{d}=(\ell, r, s)$ are the velocities of the ionized and neutral components of PIP, respectively, $t$ is time, $p$ is pressure, $\rho _m$, $T_m$ and $p_m$ are the constant space averages of density, temperature and pressure, respectively, $\tilde {\mu }$ is the effective viscosity of PIP, $\mu$ is the dynamic viscosity of PIP, $\nu _c$ is the collisional frequency between components of PIP, $k_i$ is the thermal conductivity of ionized components of PIP, $h$ is the inter-phase heat transfer coefficient, $c_i$ and $c_s$ are the specific heats of the ionized components of the PIP and the solid phase of the porous medium, respectively, and $k_s$ is the thermal conductivity of the solid phase of the porous medium. In (2.4) and (2.5), $(\rho c)_i = \rho _i c_i$ and $(\rho c)_s = \rho _s c_s$.

Additionally, $\alpha _m$ and $K_m$ are defined as

(2.7a,b)\begin{equation} \alpha_m={-}\left(\frac{1}{\rho}\frac{\partial\rho}{\partial t}\right)_m \left(=\alpha,say\right), \quad K_m=\left(\frac{1}{\rho}\frac{\partial\rho}{\partial p}\right)_m, \end{equation}

where $\alpha$ is the coefficient of thermal expansion.

3 Basic state and non-dimensionalized perturbed equations

The steady basic state (Sharma & Sunil Reference Sharma and Sunil1995) is defined as follows:

(3.1a-e)\begin{equation} \boldsymbol{q}={\boldsymbol{q}_{b}}=\textbf{0}, \quad p={{p}_{b}}\left( z \right), \quad T_i=T_{{i}_{b}}\left( z \right), \quad T_s=T_{{s}_{b}}\left( z \right) \quad \text{and} \quad {\boldsymbol{q}_{d}}={\boldsymbol{q}_{d}}_{_{b}}=\textbf{0}, \end{equation}

where

(3.2)\begin{equation} \left.\begin{gathered} \frac{1}{{\rho_{m}}}\boldsymbol{\nabla} {{p}_{b}}\left( z \right) = \boldsymbol{g} \left[ 1-\alpha_m (T-T_{m})+K_{m}\left(p-p_{m}\right)\right],\\ T_{{i}_{b}}\left( z \right)={-}\beta z+T_m=T_{{s}_{b}}\left( z \right). \end{gathered}\right\} \end{equation}

Here, the subscript ‘$b$’ denotes the basic state and $\beta (=(T_0-T_d)/d)$ is the temperature gradient.

Consider small disturbances in the form of

(3.3a-e)\begin{equation} \boldsymbol{{q}'}=\left(u',v',w'\right), \quad p', \quad {\boldsymbol{q}'_{d}}=\left(\ell',r',s'\right), \quad \theta, \quad \text{and} \quad \psi, \end{equation}

representing perturbations in $\boldsymbol {q}$, $p$, $\boldsymbol {q}^{}_{d}$, $T_i$ and $T_s$, respectively. The nonlinear perturbation equations are

(3.4)\begin{gather} \frac{1}{\epsilon} \frac{\partial \boldsymbol{{q}'}}{\partial t} ={-}\frac{1}{{\rho_{m}}}\boldsymbol{\nabla} p' - \boldsymbol{g} \alpha \theta -\frac{1}{\rho_{m}}\left[ \frac{\mu }{{{K}_{1}}}-\tilde{\mu }{{\nabla}^{2}} \right]\boldsymbol{q'} + \frac{{{\rho }_{d}}{{\nu }_{c}}}{{{\rho }_{m}\epsilon}}\left( {\boldsymbol{q}'_{d}}-\boldsymbol{{q}'} \right), \end{gather}
(3.5)\begin{gather}\boldsymbol{\nabla}\boldsymbol{\cdot}\boldsymbol{q'}=0, \end{gather}
(3.6)\begin{gather} \epsilon \left(\rho c\right)_i \frac{\partial \theta}{\partial t} + \left(\rho c\right)_i \left( \boldsymbol{q'}\boldsymbol{\cdot}\boldsymbol{\nabla}\right)\theta= \epsilon k_i \nabla^2 \theta +h\left(\psi-\theta\right)+\left(\rho c\right)_i \left(\beta -\frac{g}{c_i}\right) w', \end{gather}
(3.7)\begin{gather}\left(1-\epsilon\right) \left(\rho c\right)_s \frac{\partial \psi}{\partial t} = \left(1-\epsilon\right) k_s \nabla^2 \psi - h\left(\psi-\theta\right), \end{gather}
(3.8)\begin{gather} \frac{\partial {\boldsymbol{q}'_{d}}}{\partial t}+\frac{1}{\epsilon}\left({\boldsymbol{q}'_{d}} \boldsymbol{\cdot}\boldsymbol{\nabla}\right) {\boldsymbol{q}'_{d}} ={-}\nu_{c}\left({\boldsymbol{q}'_{d}}-\boldsymbol{{q}'}\right). \end{gather}

The perturbed equations (3.4)–(3.8) are non-dimensionalized using the following scales:

(3.9)\begin{equation} \left.\begin{gathered} t^*=\frac{k_i}{d^2 \left( \rho c \right)_i}t, \quad \boldsymbol{q^*}=\frac{\left(\rho c\right)_i d}{\epsilon k_i} \boldsymbol{{q'}}, \quad \boldsymbol{q}^{*}_{d}=\frac{\left(\rho c\right)_i d}{\epsilon k_i}{\boldsymbol{q}'_{d}}, \quad z^*=\frac{z}{d}, \\ p^*=\frac{K_1\left(\rho c\right)_i}{\mu \epsilon k_i}p^{*}, \quad \theta^*=\frac{\sqrt{Ra}}{\beta d}\theta, \quad \psi^*=\frac{\sqrt{Ra}}{\beta d}\psi. \end{gathered}\right\} \end{equation}

As a result, the non-dimensionalized perturbed equations (after removing the asterisk) are given as

(3.10)\begin{gather} \frac{1}{Va} \frac{\partial \boldsymbol{q}}{\partial t} ={-}\boldsymbol{\nabla} p + \sqrt{Ra}\theta \hat{\boldsymbol{k}} - \boldsymbol{q} + \widetilde{Da}\nabla^2 \boldsymbol{q} + \frac{FLPr}{Va}\left( \boldsymbol{q}_d - \boldsymbol{q}\right), \end{gather}
(3.11)\begin{gather} \boldsymbol{\nabla}\boldsymbol{\cdot}\boldsymbol{q}=0, \end{gather}
(3.12)\begin{gather}\frac{\partial \theta}{\partial t}+ \left(\boldsymbol{q}\boldsymbol{\cdot}\boldsymbol{\nabla} \right)\theta= \nabla^2\theta + \mathcal{H}\left( \psi-\theta \right) + \sqrt{Ra}\left(1-\frac{1}{G}\right)w, \end{gather}
(3.13)\begin{gather} \mathcal{A} \frac{\partial \psi}{\partial t}= \nabla^2\psi-\mathcal{H}\gamma \left( \psi-\theta \right), \end{gather}
(3.14)\begin{gather}\frac{\partial \boldsymbol{q}_d}{\partial t}+ \left( \boldsymbol{q}_d\boldsymbol{\cdot}\boldsymbol{\nabla} \right)\boldsymbol{q}_d={-}L Pr \left( \boldsymbol{q}_d-\boldsymbol{q} \right). \end{gather}

Here, $Va=(\rho c)_i d^2 \mu \epsilon / k_i \rho _m K_1$ is the Vadasz number, $Ra=g\alpha \beta d^2K_1\rho _m(\rho c)_i/ \mu \epsilon k_i$ is the Rayleigh–Darcy number, $\vphantom{{1}^{2^{2^{2}}}}\widetilde {Da}=\tilde {\mu } K_1/ \mu d^2$ is the Darcy–Brinkman number, $L=\nu _c d^2 \rho _m/ \mu$ is the collisional frequency parameter of PIP, $Pr=\mu (\rho c)_i/ k_i \rho _m$ is the Prandtl number, $F=\rho _n/\rho _m$ is the ratio of densities of neutral to ionized components of PIP, $\mathcal {H}=hd^2/\epsilon k_i$ is the scaled inter-phase heat-transfer coefficient, $G=g\rho _m/ (\rho c)_i\beta$ is the compressibility parameter, $\mathcal {A}=(\rho c)_s k_i/ (\rho c)_i K_s$ is the diffusivity ratio and $\gamma =\epsilon k_i/ (1-\epsilon ) k_s$ is the porosity-modified conductivity ratio.

To investigate the impact of the collisional frequency of two components of PIP on the system's stability, we neglect the convective term in (3.14) as it is independent of the collisional frequency. Subsequently, using the normal mode analysis technique on (3.14), we obtain

(3.15)\begin{equation} {{\boldsymbol{q}}_{d}}=\left(\frac{Pr L}{n+L}\right)\boldsymbol{q}, \end{equation}

where $n (\equiv \partial /\partial t)$ is the frequency of harmonic disturbances (Sharma & Sharma Reference Sharma and Sharma1978). Substituting (3.15) into (3.10), we get

(3.16)\begin{equation} \frac{1}{Va} \left[1+\frac{FLPr}{n+LPr}\right] \frac{\partial \boldsymbol{q}}{\partial t} ={-}\boldsymbol{\nabla} p + \sqrt{Ra}\theta \hat{\boldsymbol{k}} - \boldsymbol{q} + \widetilde{Da}\nabla^2 \boldsymbol{q}. \end{equation}

The boundary conditions (BCs) associated with (3.11)–(3.13) and (3.16) are

(3.17)\begin{equation} \boldsymbol{q}=\boldsymbol{0}, \quad \theta =0, \quad\psi=0 \quad \text{on} \ z=0, 1, \end{equation}

and plane tiling periodicity in $x$ and $y$ is satisfied by $\boldsymbol {{q}}$, $\theta$ and $\psi$ (Straughan Reference Straughan2006).

4 Nonlinear analysis

To perform a nonlinear analysis, we employ the energy method. Let $\| \cdot \|$ denote the ${{L}^{2}}( V )$ norm, $\langle \cdot \rangle$ denote integration over $V$ and $V$ denote the three-dimensional periodicity cell. To begin, multiply (3.16) by $\boldsymbol {q}$, (3.12) by $\theta$ and (3.13) by $\psi /\gamma$. Integrating the resulting equations over $V$ and using (3.11) and the BCs, we obtain

(4.1)\begin{gather} \frac{1}{2Va} \left[1+\frac{FLPr}{n+LPr}\right] \frac{{\rm d}}{{\rm d} t}{{\left\| \boldsymbol{q} \right\|}^{2}}= \sqrt{Ra} \langle w\theta \rangle-\left\| \boldsymbol{q} \right\|^2 -\widetilde{Da} \left\| \boldsymbol{\nabla}\boldsymbol{q} \right\|^2, \end{gather}
(4.2)\begin{gather}\frac{1}{2} \frac{{\rm d}}{{\rm d}t} \left\| \theta \right\|^2 = \sqrt{Ra} \left( 1-\frac{1}{G} \right) \langle w\theta \rangle - \mathcal{H} \langle \theta \left( \theta -\psi \right) \rangle -\left\| \boldsymbol{\nabla} \theta \right\|^2 , \end{gather}
(4.3)\begin{gather} \frac{\mathcal{A}}{2\gamma} \frac{{\rm d}}{{\rm d}t} \left\| \psi \right\|^2 ={-} \mathcal{H} \langle \psi \left(\psi-\theta\right) \rangle -\frac{1}{\gamma} \left\| \boldsymbol{\nabla} \psi \right\|^2. \end{gather}

By adding (4.1)–(4.3), we get

(4.4)\begin{equation} \frac{{\rm d}E}{{\rm d}t}=I_0-D_0, \end{equation}

where $E(t)$ is energy, $I_0$ is the production term and $D_0$ is the dissipation term. Here,

(4.5)\begin{gather} E(t)=\frac{1}{2} \left\| \theta \right\|^2 + \frac{\lambda}{2Va} \left[1+\frac{FLPr}{n+LPr}\right] {{\left\| \boldsymbol{q} \right\|}^{2}} + \frac{\mathcal{A}}{2\gamma} \left\| \psi \right\|^2, \end{gather}
(4.6)\begin{gather} I_0=\lambda\sqrt{Ra} \langle w\theta \rangle + \sqrt{Ra} \left( 1-\frac{1}{G} \right) \langle w\theta \rangle, \end{gather}
(4.7)\begin{gather}D_0=\lambda \left( \widetilde{Da} \left\| \nabla\boldsymbol{q} \right\|^2 + \left\| \boldsymbol{q} \right\|^2 \right)+ \left\| \boldsymbol{\nabla} \theta \right\|^2 + \mathcal{H} \left\| \theta-\psi \right\|^2 + \frac{1}{\gamma} \left\| \boldsymbol{\nabla} \psi \right\|^2, \end{gather}

where $\lambda$ is the coupling parameter.

From (4.4), we obtain

(4.8)\begin{equation} \frac{{\rm d}E}{{\rm d}t} \leqslant \left(m-1 \right)D_0, \end{equation}

where

(4.9)\begin{equation} m=\underset{\mathcal{G}}{{\max}}\,\frac{I_0}{D_0}, \end{equation}

and $\mathcal {G}$ is the admissible solution space.

Using the Poincaré inequality, (4.7) becomes

(4.10)\begin{equation} D_0\geqslant K^{*} E(t), \end{equation}

where

(4.11)\begin{equation} K^{*}=2{\rm \pi}^2\min\left\{1,\frac{1}{\mathcal{A}},\frac{1}{Va} \left(\frac{1}{{\rm \pi}^2}+\widetilde{Da}\right)\left[1+\frac{FLPr}{n+LPr}\right]^{{-}1}\right\}. \end{equation}

Using inequality (4.10) in inequality (4.8), we get

(4.12)\begin{equation} \frac{{\rm d}E}{{\rm d}t} \leqslant -a_0K^{*} E(t), \end{equation}

where $a_0=1-m>0$.

By integrating inequality (4.12), we get

(4.13)\begin{equation} E(t) \leqslant E(0)\exp({-a_oK^*t}). \end{equation}

From inequality (4.13), it follows that $E( t )\to 0$ exponentially as $t\to \infty$. This ensures the stability for all values of $E( 0 )$. The collisional frequency plays a significant role in energy decay, particularly when

(4.14)\begin{equation} \left\{1, \frac{1}{\mathcal{A}}\right\} > \frac{1}{Va}\left(\frac{1}{{\rm \pi}^2}+\widetilde{Da}\right) \left[1+\frac{FLPr}{n+LPr}\right]^{{-}1}. \end{equation}

Similarly, the thermal diffusivity ratio is important if

(4.15)\begin{equation} \left\{1,\frac{1}{Va}\left(\frac{1}{{\rm \pi}^2}+\widetilde{Da}\right) \left[1+\frac{FLPr}{n+LPr}\right]^{{-}1} \right\}> \frac{1}{\mathcal{A}}. \end{equation}

These conditions define thresholds where the collisional frequency and thermal diffusivity ratio significantly influence the stability and energy dynamics of the system under consideration.

4.1 Eigenvalue problem for nonlinear analysis

The value of $Ra$ is calculated from the Euler–Lagrange equations derived from (4.9), namely

(4.16)\begin{gather} \frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] \theta -2\boldsymbol{q} +2\widetilde{Da}\nabla^2\boldsymbol{q}=0, \end{gather}
(4.17)\begin{gather}\frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] w + 2 \nabla^2\theta - 2\mathcal{H}\left( \theta -\psi \right)=0, \end{gather}
(4.18)\begin{gather}\frac{2}{\gamma} \nabla^2\psi + 2\mathcal{H} \left( \theta -\psi \right)=0. \end{gather}

Operating $\hat {\boldsymbol {k}}\boldsymbol {\cdot }\textrm {curlcurl}$ on (4.16), we get

(4.19)\begin{equation} \frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] \nabla^{2}_{1}\theta - 2\nabla^2w +2\widetilde{Da}\nabla^4w=0, \end{equation}

where

(4.20)\begin{equation} \nabla^{2}_{1}\equiv\frac{{{\partial }^{2}}}{\partial {{x}^{2}}}+\frac{{{\partial }^{2}}}{\partial {{y}^{2}}}. \end{equation}

We assume a plane tiling of the form

(4.21)\begin{equation} \left\{ w, \theta, \psi \right\}= \left\{ W(z), \varTheta(z), \varPsi(z) \right\}f(x,y), \end{equation}

where $f$ is a planform satisfying $\nabla _{1}^{2} f +a^2 f =0$, $a$ being a wavenumber.

Using the plane form (4.21), (4.19), (4.17) and (4.18) become

(4.22)\begin{gather} \frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] a^2\varTheta - 2\widetilde{Da}\left(D^2-a^2\right)^2W+ 2\left(D^2-a^2 \right)W =0, \end{gather}
(4.23)\begin{gather}\frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] W + 2 \left( D^2-a^2 \right)\varTheta - 2\mathcal{H}\left( \varTheta -\varPsi \right)=0, \end{gather}
(4.24)\begin{gather}\frac{2}{\gamma} \left( D^2-a^2 \right)\varPsi + 2\mathcal{H} \left( \varTheta -\varPsi \right)=0, \end{gather}

where $D={\textrm {d}}/{\textrm {d}z}$. The BCs associated with (4.22)–(4.24) are

(4.25)\begin{equation} \left.\begin{gathered}W=\varPsi=\varTheta=0 \quad \text{for} \ z=0, 1,\\ \text{and} \quad DW=0 \quad \text{for rigid surface},\\ \text{and} \quad D^2W=0 \quad \text{for free surface}. \end{gathered}\right\} \end{equation}

The set of equations (4.22)–(4.24), together with the BCs (4.25), constitutes the eigenvalue problem for nonlinear analysis.

5 Linear analysis

To study the linear analysis, by considering the perturbation to be infinitesimally small, we ignore the nonlinear terms from the non-dimensional perturbed equations and obtain the linearized non-dimensional perturbed equations as

(5.1)\begin{gather} \frac{1}{Va} \left[1+\frac{FLPr}{n+LPr}\right] \frac{\partial \boldsymbol{q}}{\partial t} ={-}\boldsymbol{\nabla} p + \sqrt{Ra}\theta \hat{\boldsymbol{k}} - \boldsymbol{q} + \widetilde{Da}\nabla^2 \boldsymbol{q}, \end{gather}
(5.2)\begin{gather}\frac{\partial \theta}{\partial t}= \sqrt{Ra}\left(1-\frac{1}{G}\right)w + \mathcal{H}\left( \psi-\theta \right)+ \nabla^2\theta, \end{gather}
(5.3)\begin{gather}\mathcal{A} \frac{\partial \psi}{\partial t}={-}\mathcal{H}\gamma \left( \psi-\theta \right) + \nabla^2\psi. \end{gather}

Operating $\hat {\boldsymbol {k}}\boldsymbol {\cdot }\textrm {curlcurl}$ on (5.1), we get

(5.4)\begin{equation} \frac{1}{Va} \left[1+\frac{FLPr}{n+LPr}\right] \frac{\partial}{\partial t} \left( \nabla^2w \right) = \sqrt{Ra}\nabla^{2}_{1}\theta - \nabla^2w + \widetilde{Da}\nabla^4 w. \end{equation}

By applying the normal mode analysis method to equations (5.4), (5.2) and (5.3), where we assumed the perturbed quantities of the form

(5.5)\begin{equation} \left\{ w, \theta, \psi \right\}= \left\{ W(z), \varTheta(z), \varPsi(z) \right\} \exp\left\{ \iota \left( a_xx+a_yy \right) + nt \right\}. \end{equation}

Here, $\sqrt {a^2_x+a^2_y}=a$ is the wavenumber.

Using (5.5), equations (5.4), (5.2) and (5.3) become

(5.6)\begin{gather} \frac{1}{Va} \left[1+\frac{FLPr}{n+LPr}\right] n \left( D^2-a^2 \right)W ={-}\sqrt{Ra}a^2\varTheta - \left( D^2-a^2 \right)W + \widetilde{Da}\left( D^2-a^2 \right)^2W, \end{gather}
(5.7)\begin{gather}n\varTheta= \left( D^2-a^2 \right)\varTheta + \mathcal{H}\left( \varPsi-\varTheta \right) + \sqrt{Ra}\left(1-\frac{1}{G}\right)W, \end{gather}
(5.8)\begin{gather}\mathcal{A} n \varPsi= \left( D^2-a^2 \right)\varPsi-\mathcal{H}\gamma \left( \varPsi-\varTheta \right). \end{gather}

The BCs associated with (5.6)–(5.8) are same as (4.25).

5.1 Exchange of stabilities

Here, we prove that the principle of exchange of stabilities is valid, namely that marginally stable modes with $\sigma =0$ also have $\omega =0$, where $\sigma +\iota \omega =n$.

Multiply (5.6) by $W^*$ (complex conjugate of $W$), (5.7) by $\varTheta ^*$ (complex conjugate of $\varTheta$) and (5.8) by $\varPsi ^*$ (complex conjugate of $\varPsi$), integrating resulting equations over the range of $z$ and using BCs (4.25), we get

(5.9)\begin{gather} -\frac{n}{Va}\left[1+\frac{FLPr}{n+LPr}\right] \mathcal{I}_1 + \sqrt{Ra}a^2 \int_{0}^{1} W^* \varTheta \,{\rm d}z - \mathcal{I}_1-\widetilde{Da} \mathcal{I}_2=0, \end{gather}
(5.10)\begin{gather} n\mathcal{I}_3 + \mathcal{I}_4 - \mathcal{H} \int_{0}^{1} \varTheta^* \varPsi \,{\rm d}z + \mathcal{H} \mathcal{I}_3 - \sqrt{Ra}\left(1-\frac{1}{G}\right) \int_{0}^{1} W \varTheta^* \,{\rm d}z=0, \end{gather}
(5.11)\begin{gather}\mathcal{A}n\mathcal{I}_5 + \mathcal{I}_6 + \mathcal{H}\gamma\mathcal{I}_5-\mathcal{H}\gamma \int_{0}^{1} \varTheta \varPsi^* \,{\rm d}z, \end{gather}

where

(5.12)\begin{align} \left.\begin{gathered} \mathcal{I}_1=\int_{0}^{1} \left( \lvert DW \rvert^2+a^2\lvert W \rvert^2 \right) {\rm d}z, \quad \mathcal{I}_2=\int_{0}^{1} \left( \lvert D^2W \rvert^2+2a^2\lvert DW \rvert^2+a^4\lvert W \rvert^2 \right) {\rm d}z, \\ \mathcal{I}_3= \int_{0}^{1} \lvert \varTheta \rvert^2 \, {\rm d}z, \quad \mathcal{I}_4= \int_{0}^{1} \left( \lvert D\varTheta \rvert^2+a^2\lvert \varTheta \rvert^2 \right) {\rm d}z, \\ \mathcal{I}_5=\int_{0}^{1} \lvert \varPsi \rvert^2 \, {\rm d}z, \quad \mathcal{I}_6=\int_{0}^{1} \left( \lvert D\varPsi \rvert^2 +a^2 \lvert \varPsi \rvert^2 \right) {\rm d}z. \end{gathered}\right\} \end{align}

The integrals $\mathcal {I}_1$, $\mathcal {I}_2$, $\mathcal {I}_3$, $\mathcal {I}_4$, $\mathcal {I}_5$ and $\mathcal {I}_6$ are positive. So, subtracting the product of $a^2$ with (5.11) from the sum of the product of $(1-1/G)\gamma$ with (5.9) and the product of $a^2\gamma$ with complex conjugate of (5.10), we get

(5.13)\begin{align} & -\gamma \left(1-\frac{1}{G}\right) \left( \frac{n}{Va}\left[1+\frac{FLPr}{n+LPr}\right] +1 \right) \mathcal{I}_1 - \left(1-\frac{1}{G}\right)\gamma\widetilde{Da}\mathcal{I}_2 \nonumber\\ & \quad + a^2\gamma\left( n^*+\mathcal{H} \right)\mathcal{I}_3 + a^2\gamma\mathcal{I}_4 - a^2 \left( \mathcal{A}n+\gamma\mathcal{H} \right)\mathcal{I}_5 - a^2\mathcal{I}_6 = 0 , \end{align}

where $n^*$ is the complex conjugate of $n$.

The imaginary part of (5.13) gives

(5.14)\begin{equation} \omega \left\{ \left(1-\frac{1}{G}\right) \left( \frac{1}{Va} + \frac{FL^2(Pr)^2}{\left(\sigma+LPr\right)^2+\omega^2} \right)\gamma\mathcal{I}_1 + a^2\gamma\mathcal{I}_3 + a^2\mathcal{A}\mathcal{I}_5 \right\}=0. \end{equation}

The quantity inside the curly brackets is positive definite for $G$ is greater than $1$. Hence, (5.14) suggests that $\omega =0$. This establishes that the principle of exchange of stabilities is valid (Chandrasekhar Reference Chandrasekhar1981).

Now, keeping in mind the validation of the principle of exchange of stabilities, (5.6)–(5.8) can be rewritten as

(5.15)\begin{gather} \sqrt{Ra}a^2\varTheta + \left( D^2-a^2 \right)W - \widetilde{Da}\left( D^2-a^2 \right)^2W=0, \end{gather}
(5.16)\begin{gather}\left( D^2-a^2 \right)\varTheta + \mathcal{H}\left( \varPsi-\varTheta \right) + \sqrt{Ra}\left(1-\frac{1}{G}\right)W=0, \end{gather}
(5.17)\begin{gather}\left( D^2-a^2 \right)\varPsi-\mathcal{H}\gamma \left( \varPsi-\varTheta \right)=0. \end{gather}

The set of equations (5.15)–(5.17), together with the BCs (4.25), constitutes the eigenvalue problem for linear analysis.

6 Method of solution

The eigenvalue problems for both nonlinear as well as linear analyses have been solved using the single-term Galerkin method. In this method, the weighted functions are the same as the base (trial) functions (Yadav et al. Reference Yadav, Bhargava and Agarwal2013). Accordingly, we define $W$, $\varTheta$ and $\varPsi$ in the following form:

(6.1a-c)\begin{equation} W=\sum_{j=1}^{N}{{{A}_{j}}{{W}_{j}}}, \quad \varTheta=\sum_{j=1}^{N}{{{B}_{j}}{{\varTheta}_{j}}}, \quad \varPsi=\sum_{j=1}^{N}{{{C}_{j}}{{\varPsi}_{j}}}, \end{equation}

where ${{A}_{j}}$, ${{B}_{j}}$ and $C_j$ are unknown coefficients, $j= 1,2,3,\ldots, N$ and the base functions ${{W}_{j}}$, ${{\varTheta }_{j}}$ and $\varPsi _j$ are assumed to be in the following forms for free–free, rigid–free and rigid–rigid bounding surfaces (Chandrasekhar Reference Chandrasekhar1981; Yadav et al. Reference Yadav, Bhargava and Agarwal2013), respectively:

(6.2)\begin{gather} {{W}_{j}}=\sin (j{\rm \pi} z)={{\varTheta}_{j}}={{\varPsi }_{j}}, \end{gather}
(6.3a,b)\begin{gather}{{W}_{j}}=z^2\left(1-z\right)\left[ \left(j+2\right)-2z^j \right], \quad {{\varTheta}_{j}}=z^j-z^{j+1}={{\varPsi }_{j}}, \end{gather}
(6.4a,b)\begin{gather} {{W}_{j}}=z^{j+1}-2z^{j+2}+z^{j+3}, \quad {{\varTheta}_{j}}=z^j-z^{j+1}={{\varPsi }_{j}}, \end{gather}

such that $W_j$, $\varTheta _j$ and $\varPsi _j$ satisfy the corresponding BCs (4.25). By substituting the expressions for $W$, $\varTheta$ and $\varPsi$ into (4.22)–(4.24) and (5.15)–(5.17), and then multiplying the first equation by $W_j$, the second equation by $\varTheta _j$ and the third equation by $\varPsi _j$ and integrating the resulting equations over the interval from zero to unity, we obtain a set of linear homogeneous equations. This set of equations admits a non-trivial solution only if its determinant is equal to zero, which gives the characteristic equations of the eigenvalue problems in terms of the Rayleigh–Darcy number $Ra$.

6.1 Rayleigh–Darcy number for free–free bounding surfaces

For nonlinear analysis, $Ra$ has been found as a function of $a$, $\lambda$, $\mathcal {H}$, $\gamma$ and $G$. The expression for $Ra$ is given by

(6.5)\begin{equation} Ra=\frac{4G^2\left(a^2+{\rm \pi}^2\right)^2 \left(1+a^2\widetilde{Da}+{\rm \pi}^2\widetilde{Da}\right) \left( a^2+{\rm \pi}^2+\mathcal{H}+\gamma\mathcal{H} \right)\lambda}{a^2 \left( a^2+{\rm \pi}^2+\gamma\mathcal{H} \right)\left(G-1+G\lambda\right)^2}. \end{equation}

The optimal value of $\lambda$ has been found using the condition $\textrm {d}Ra/\textrm {d}\lambda =0$, which yields $\lambda =(G-1)/G$. Using this value, the expression for $Ra$ becomes

(6.6)\begin{equation} Ra=\left(\frac{G}{G-1}\right)\frac{\left(a^2+{\rm \pi}^2\right)^2 \left(1+a^2\widetilde{Da}+{\rm \pi}^2\widetilde{Da}\right) \left( a^2+{\rm \pi}^2+\mathcal{H}+\gamma\mathcal{H} \right)}{a^2 \left( a^2+{\rm \pi}^2+\gamma\mathcal{H} \right)}. \end{equation}

For linear analysis, the value of $Ra$ found by solving the eigenvalue problem (5.15)–(5.17) using the Galerkin method has been the same as (6.6).

6.2 Rayleigh–Darcy number for rigid–free bounding surfaces

For nonlinear analysis, $Ra$ has been found as a function of $a$, $\lambda$, $\mathcal {H}$, $\gamma$ and $G$. The expression for $Ra$ is given by

(6.7)\begin{align} Ra=\frac{112G^2\left(10+a^2\right) \left( 19a^4\widetilde{Da}+216\left(1+21\widetilde{Da}\right)+a^2\left(19+432\widetilde{Da}\right)\right) \left( 10+ a^2+\mathcal{H}+\gamma\mathcal{H} \right)\lambda}{507a^2 \left( 10+a^2+\gamma\mathcal{H} \right)\left(G-1+G\lambda\right)^2}. \end{align}

The optimal value of $\lambda$ has been found using the condition $dRa/d\lambda =0$, which yields $\lambda =(G-1)/G$. Using this value, the expression for $Ra$ becomes

(6.8)\begin{align} Ra=\frac{28G\left(10+a^2\right) \left( 19a^4\widetilde{Da}+216\left(1+21\widetilde{Da}\right)+a^2\left(19+432\widetilde{Da}\right)\right) \left( 10+ a^2+\mathcal{H}+\gamma\mathcal{H} \right)}{507a^2 \left(G-1\right)\left( 10+a^2+\gamma\mathcal{H} \right)}. \end{align}

For linear analysis, the value of $Ra$ found by solving the eigenvalue problem (5.15)–(5.17) using the Galerkin method has been the same as (6.8).

6.3 Rayleigh–Darcy number for rigid–rigid bounding surfaces

For nonlinear analysis, $Ra$ has been found as a function of $a$, $\lambda$, $\mathcal {H}$, $\gamma$ and $G$. The expression for $Ra$ is given by

(6.9)\begin{align} Ra=\frac{112G^2\left(10+a^2\right) \left( 12+504\widetilde{Da}+a^4\widetilde{Da}+a^2\left(1+24\widetilde{Da}\right)\right) \left( 10+ a^2+\mathcal{H}+\gamma\mathcal{H} \right)\lambda}{27a^2 \left( 10+a^2+\gamma\mathcal{H} \right)\left(G-1+G\lambda\right)^2}. \end{align}

The optimal value of $\lambda$ has been found using the condition $\textrm {d}Ra/\textrm {d}\lambda =0$, which yields $\lambda =(G-1)/G$. Using this value, the expression for $Ra$ becomes

(6.10)\begin{align} Ra=\frac{28G\left(10+a^2\right) \left( 12+504\widetilde{Da}+a^4\widetilde{Da}+a^2\left(1+24\widetilde{Da}\right)\right) \left( 10+ a^2+\mathcal{H}+\gamma\mathcal{H} \right)}{27a^2 \left(G-1\right)\left( 10+a^2+\gamma\mathcal{H} \right)}. \end{align}

For linear analysis, the value of $Ra$ found by solving the eigenvalue problem (5.15)–(5.17) using the Galerkin method has been the same as (6.10).

Here, $Ra$ has the same value for both nonlinear and linear analyses, indicating a lack of subcritical regions and demonstrating strong global stability. It has been observed that collisional effects contribute to energy decay, but do not affect the value of $Ra$ in either nonlinear or linear analyses.

7 Results and discussion

This section presents graphical representations of neutral stability curves for various parameter values. Numerical calculations are conducted with $G=3$, $\widetilde {Da}$ ranging from $0$ to $1$, $\mathcal {H}$ ranging from $10^{-5}$ to $10^{5}$ and $\gamma$ ranging from $0$ to $10$. The parameter ranges discussed are inspired by the existing literature (Sharma & Sunil Reference Sharma and Sunil1995; Malashetty et al. Reference Malashetty, Swamy and Kulkarni2007, Reference Malashetty, Swamy and Heera2008; Nield & Bejan Reference Nield and Bejan2013; Yadav & Lee Reference Yadav and Lee2015).

It is evident from the expressions of $Ra$ that, as the values of $G$ increase, $Ra$ also increases, suggesting that compressibility exerts a stabilizing influence. Notably, cases where $G=1$ and $G<1$ are not applicable in this study, as they result in infinite or negative $Ra$. The significance of this relationship becomes particularly pronounced when $G$ exceeds $1$, as emphasized in a study by Sharma & Sunil (Reference Sharma and Sunil1995).

Figure 2 compares LTNE and LTE conditions for the distinct combinations of free and rigid bounding surfaces. The values of $Ra$ for respective bounding surfaces have been calculated using $\mathcal {H}\to 0$ in (6.6), (6.8) and (6.10), for LTE conditions and for LTNE conditions using $\vphantom{{1}^{2^{2^{2}}}}\mathcal {H}=10$ and $\gamma =5$. The value of $\widetilde {Da}$ is kept fixed as 0.1 for both LTE and LTNE. It is clear from the figure that the values of $Ra$ in the case of LTNE are more than that of the LTE case for respective bounding surfaces. This indicates that onset convection occurs earlier for the LTE conditions than for the LTNE conditions.

Figure 2. Comparison of Rayleigh–Darcy number $(Ra)$ for LTNE and LTE across various combinations of bounding surfaces. Solid curves depict $Ra$ variations for LTNE, while dotted curves show $Ra$ variations for LTE. Curves labelled aa, bb and cc show the variations of $Ra$ with wavenumber $(a)$ for rigid–rigid, rigid–free and free–free surfaces, respectively, under LTNE conditions. The curves labelled a$'$a$'$, b$'$b$'$ and c$'$c$'$ represent the variations of $Ra$ with $a$ for rigid–rigid, rigid–free and free–free bounding surfaces, respectively, under LTE conditions.

In figure 3, the variation of $Ra_c$ with $\widetilde {Da}$ for different combinations of bounding surfaces at $\mathcal {H}=100$, $\gamma =5$ and $G=3$ is displayed. This figure illustrates the stabilizing effect of medium permeability for all three different combinations of bounding surfaces. This behaviour occurs because permeability adds resistance to the flow of PIP, leading to enhanced mixing of PIP and more efficient temperature distribution. Consequently, the convective heat transfer increases, delaying the onset of convection and increasing $Ra_c$.

Figure 3. Variation of critical Rayleigh–Darcy number ($Ra_c$) with Darcy–Brinkman number ($\widetilde {Da}$) for the distinct combinations of bounding surfaces.

In figure 4, the relationship between $Ra_c$ and the scaled inter-phase heat transfer coefficient ($\log _{10}\mathcal {H}$) for different bounding surface combinations is illustrated. It is observed that, as $\mathcal {H}$ increases, $Ra_c$ also increases, indicating a stabilizing effect. For small $\mathcal {H}$ values, $Ra_c$ shows minimal variation and is independent of $\gamma$. However, for $\mathcal {H} > 1$, the variation becomes significant as $Ra_c$ becomes dependent on $\gamma$. This behaviour is a result of the negligible heat transfer between the phases at low $\mathcal {H}$ values, making the critical value unaffected by the solid phase properties. Conversely, at high $\mathcal {H}$ values, the temperatures of the phases nearly equalize, allowing them to be treated as a single phase. In between these extremes, $\mathcal {H}$ introduces strong non-equilibrium effects. Additionally, an increase in $\gamma$ leads to a decrease in $Ra_c$, highlighting the destabilizing influence of $\gamma$. This trend is further illustrated in figure 5, which is plotted for a constant $\mathcal {H}$. Higher $\gamma$ values imply that heat is transported through both the solid and PIP phases, whereas lower values indicate that heat is transported primarily through the PIP phase. Thus, convection is more readily initiated for higher $\gamma$ values when all the other parameters are held constant.

Figure 4. Comparison of critical Rayleigh–Darcy number $(Ra_c)$ for various values of scaled inter-phase heat-transfer coefficient $(\log _{10}\mathcal {H})$, for different values of porosity-modified conductivity ratio $(\gamma )$, for distinct combination of bounding surfaces. The blue curves represent rigid–rigid bounding surfaces, the red curves indicate rigid–free bounding surfaces and the black curves correspond to free–free bounding surfaces. The dashed curves are for $\gamma = 0.5$, the solid curves for $\gamma = 1$ and the dotted curves for $\gamma = 5$.

Figure 5. Variation of critical Rayleigh–Darcy number ($Ra_c$) with porosity-modified conductivity ratio ($\gamma$) for the distinct combinations of bounding surfaces.

Additionally, based on these graphical results, we infer that PIP exhibits greater thermal stability when confined between rigid–rigid bounding surfaces, while it shows less thermal stability when the plasma layer is confined between free–free bounding surfaces.

8 Effect of magnetic field

To study the effect of the magnetic field on the problem of instability described in § 2, we consider a horizontal porous layer of PIP, subjected to a magnetic field $\boldsymbol {H}$. We have adopted the quasi-static magnetohydrodynamic approximation proposed by Galdi & Straughan (Reference Galdi and Straughan1985). The key modification is to include a term representing the Lorentz force ($\boldsymbol {\mathcal {L}}=\boldsymbol {j}\times \boldsymbol {B}_0$) in the equation of motion (2.3), where $\boldsymbol {j}$ is the current and $\boldsymbol {B}_0=(0,0, B_0)$ is the magnetic field with only vertical component (Chandrasekhar Reference Chandrasekhar1981). The non-dimensional perturbed equations (3.11)–(3.13) remains the same but (3.16) becomes

(8.1)\begin{gather} \frac{1}{Va} \left[1+\frac{FLPr}{n+LPr}\right] \frac{\partial \boldsymbol{q}}{\partial t} ={-}\nabla p + \sqrt{Ra}\theta \hat{\boldsymbol{k}} - \boldsymbol{q} + \widetilde{Da}\nabla^2 \boldsymbol{q}+ \frac{M^2\widetilde{Da}}{Vr} \left[\left(\boldsymbol{q}\times\hat{\boldsymbol{k}}\right)\times\hat{\boldsymbol{k}}\right], \end{gather}

where $M^2(=\sigma _1\boldsymbol {B}^{2}_{0}d^2/\epsilon \rho _m\nu )$ is the Hartmann number, which accounts for the effect of the magnetic field. Here, $\sigma _1$ is the electrical conductivity.

The eigenvalue problem for nonlinear analysis incorporating the effect of magnetic field is

(8.2)\begin{gather} \frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] a^2\varTheta + 2\left(D^2-a^2 \right)W - 2\widetilde{Da}\left(D^2-a^2\right)^2W + \frac{2M^2\widetilde{Da}}{Vr}D^2W=0,\end{gather}
(8.3)\begin{gather}\frac{\sqrt{Ra}}{\sqrt{\lambda}} \left[ \lambda+\frac{G-1}{G} \right] W + 2 \left( D^2-a^2 \right)\varTheta - 2\mathcal{H}\left( \varTheta -\varPsi \right)=0, \end{gather}
(8.4)\begin{gather}\frac{2}{\gamma} \left( D^2-a^2 \right)\varPsi + 2\mathcal{H} \left( \varTheta -\varPsi \right)=0. \end{gather}

The eigenvalue problem for linear analysis incorporating the effect of magnetic field is

(8.5) \begin{gather} \sqrt{Ra}a^2\varTheta + \left( D^2-a^2 \right)W - \widetilde{Da}\left( D^2-a^2 \right)^2W+\frac{M^2\widetilde{Da}}{Vr}D^2W=0, \end{gather}
(8.6)\begin{gather}\left( D^2-a^2 \right)\varTheta + \mathcal{H}\left( \varPsi-\varTheta \right) + \sqrt{Ra}\left(1-\frac{1}{G}\right)W=0, \end{gather}
(8.7)\begin{gather}\left( D^2-a^2 \right)\varPsi-\mathcal{H}\gamma \left( \varPsi-\varTheta \right)=0. \end{gather}

The BCs associated with both eigenvalue problems remain the same as (4.25).

The Rayleigh–Darcy numbers for both nonlinear and linear analyses were determined by solving the eigenvalue problems using the Galerkin method as detailed in § 6. The value of $Ra$ for nonlinear analysis has been a function of $a$, $\lambda$, $\mathcal {H}$, $\gamma$, $M^2$ and $G$. The optimal value of $\lambda$ is $(G-1)/G$ for all combinations of bounding surfaces. Using this value of $\lambda$, we have found that $Ra$ for nonlinear analysis has been the same as $Ra$ for linear analysis. The Rayleigh–Darcy numbers ($Ra$) for distinct combinations of bounding surfaces are as follows.

For free–free bounding surfaces

(8.8)\begin{equation} Ra=\left(\frac{G}{G-1}\right)\frac{\left(a^2+{\rm \pi}^2\right) \left(a^2+{\rm \pi}^2+\left(a^2+{\rm \pi}^2\right)^2\widetilde{Da}+{\rm \pi}^2\widetilde{Da}M^2\right) \left( a^2+{\rm \pi}^2+\mathcal{H}+\gamma\mathcal{H} \right)}{a^2 \left( a^2+{\rm \pi}^2+\gamma\mathcal{H} \right)}. \end{equation}

For rigid–free bounding surfaces

(8.9)\begin{align} Ra& = \frac{28G\left(10+a^2\right) \left( 10+ a^2+\mathcal{H}+\gamma\mathcal{H} \right)}{507a^2 \left(G-1\right)\left( 10+a^2+\gamma\mathcal{H} \right)} \nonumber\\ & \quad \times\left[ 19a^4\widetilde{Da}+216\left(1+21\widetilde{Da}\right)+a^2\left(19+432\widetilde{Da}\right) +216\widetilde{Da}M^2\right]. \end{align}

For rigid–rigid bounding surfaces

(8.10)\begin{align} Ra& = \frac{28G\left(10+a^2\right) \left( 10+ a^2+\mathcal{H}+\gamma\mathcal{H} \right)}{27a^2 \left(G-1\right)\left( 10+a^2+\gamma\mathcal{H} \right)} \nonumber\\ & \quad \times \left[ 12+504\widetilde{Da}+a^4\widetilde{Da}+a^2\left(1+24\widetilde{Da}\right) +12\widetilde{Da}M^2\right]. \end{align}

For the limiting case of the magnetic field i.e. $M^2=0$, clearly the Rayleigh–Darcy number for the respective bounding surfaces has been the same as given in (6.6), (6.8) and (6.10).

Figure 6 shows the variation in $Ra_c$ with the square of the Hartmann number ($M^2$) for different boundary surface combinations. The figure shows that, with the increase in $M^2$, $Ra_c$ also increases, indicating that magnetic fields delay the onset of thermal convection, thus exerting a stabilizing influence. This occurs because the magnetic field impedes the motion within the PIP. When the PIP attempts convective motion, the Lorentz force generated by the magnetic field counteracts this movement, resisting the flow and suppressing convective instabilities, thereby delaying convection onset. This suppression of plasma motion is essential in postponing or preventing the onset of thermal convection. Additionally, the figure reveals that PIP confined within rigid–rigid bounding surfaces exhibits greater thermal stability compared with PIP confined within rigid–free or free–free bounding surfaces.

Figure 6. Variation of critical Rayleigh–Darcy number ($Ra_c$) with the square of Hartmann number ($M^2$) for the distinct combinations of bounding surfaces.

The stability analysis of PIP in a porous medium with LTNE effects has significant applications across various scientific and engineering domains. In astrophysics, this study enhances the understanding of stellar and planetary atmospheres, contributing to more accurate models of these complex systems. In the field of nuclear fusion, it offers insights into the stability of magnetic confinement systems, which is crucial for the development of efficient fusion reactors. The findings also aid in the design of thermal protection systems for spacecraft, ensuring their integrity during re-entry and other high-temperature conditions. Furthermore, the optimization of heat exchangers and combustion chambers in engineering applications benefits from the improved thermal stability insights provided by this study. Additionally, the study informs the development of plasma-based pollution control technologies and advances in plasma medicine, particularly in the design of stable plasma sources for medical treatments involving porous tissues or materials.

9 Conclusions

This study explored the impact of LTNE and magnetic field on the stability of PIP heated from below and saturating a porous medium. We analysed three combinations of bounding surfaces: rigid–rigid, rigid–free and free–free. The eigenvalue problems resulting from these configurations were numerically solved using the Galerkin method. The key findings are as follows:

  1. (i) Energy decay was examined using the energy method, revealing that the collisional frequency and diffusivity ratios significantly influence the rate of energy decay.

  2. (ii) The principle of exchange of stabilities was confirmed, indicating the absence of oscillatory convection modes.

  3. (iii) Both linear and nonlinear analyses showed identical stability boundaries for all bounding surface combinations, indicating the absence of a subcritical region and the establishment of global stability.

  4. (iv) Factors such as compressibility, medium permeability and inter-phase heat transfer were found to delay the onset of thermal convection, thereby stabilizing the system.

  5. (v) The porosity-modified conductivity ratio exhibited a destabilizing effect on the system stability. At very low heat-transfer coefficients, the critical values were independent of the porosity-modified conductivity ratio.

  6. (vi) The presence of a magnetic field was observed to delay the onset of thermal convection, thus exerting a stabilizing effect.

  7. (vii) Among the bounding surface configurations, rigid–rigid surfaces provided the greatest thermal stability for confining the PIP.

Future research could explore the impact of surface tension on thermal convection in partially ionized plasma using both linear and nonlinear analyses across various bounding surface configurations. Furthermore, investigating the effects of varying ionization levels, Hall currents and finite Larmor radius on stability and convection thresholds would significantly broaden the study's scope. Additionally, the application of machine learning techniques to control Rayleigh–Bénard convection in PIP presents a promising avenue for advanced control and optimization.

Acknowledgments

The authors would like to sincerely thank the esteemed editorial board and reviewers for their invaluable comments and insightful suggestions, which have greatly enriched the quality and depth of our work.

Editor Antoine C. Bret thanks the referees for their advice in evaluating this article.

Funding

This research received no specific grant from any funding agency, commercial or not-for-profit sectors.

Declaration of interests

The authors report no conflict of interest.

References

Arnone, G., Capone, F. & Gianfrani, J.A. 2024 Stability of penetrative convective currents in local thermal non-equilibrium. Proc. R. Soc. Lond. A 480 (2287), 20230820.Google Scholar
Ballai, I. 2019 Linear waves in partially ionized plasmas in ionization non-equilibrium. Front. Astron. Space Sci. 6, 39.CrossRefGoogle Scholar
Ballester, J.L., et al. 2018 Partially ionized plasmas in astrophysics. Space Sci. Rev. 214 (2), 58.CrossRefGoogle Scholar
Ballester, J.L., Soler, R., Carbonell, M. & Terradas, J. 2021 The first adiabatic exponent in a partially ionized prominence plasma: effect on the period of slow waves. Astron. Astrophys. 656, A159.CrossRefGoogle Scholar
Bansal, A. & Suthar, O.P. 2022 A study on the effect of temperature modulation on Darcy–Bénard convection using a local thermal non-equilibrium model. Phys. Fluids 34 (4), 044107.CrossRefGoogle Scholar
Bansal, A. & Suthar, O.P. 2024 Temperature modulation effects on chaos and heat transfer in Darcy–Bénard convection using a local thermal non-equilibrium model. Nonlinear Dyn. 112 (18), 1647516493.CrossRefGoogle Scholar
Chandel, V. & Sunil, 2024 Influence of magnetic fields and bounding surface configurations on thermal convection in partially ionised plasmas: nonlinear and linear stability analyses. Pramana – J. Phys. 98 (3), 107.CrossRefGoogle Scholar
Chandel, V., Sunil, & Sharma, P. 2024 Study of global stability of rotating partially-ionized plasma saturating a porous medium. Spec. Top. Rev. Porous Media: Intl J. 15 (6), 2746.CrossRefGoogle Scholar
Chandrasekhar, S. 1981 Hydrodynamic and Hydromagnetic Stability. Dover.Google Scholar
Galdi, G.P. & Straughan, B. 1985 Exchange of stabilities, symmetry, and nonlinear stability. Arch. Rat. Mech. Anal. 89 (3), 211228.CrossRefGoogle Scholar
Ingham, D.B. & Pop, I. (Ed.) 2005 Transport Phenomena in Porous Media III. Elsevier.Google Scholar
Kaothekar, S. 2018 Thermal instability of partially ionized viscous plasma with Hall effect FLR corrections flowing through porous medium. J. Porous Media 21 (8), 679699.CrossRefGoogle Scholar
Krishan, V. 2022 Different representations of a partially ionized plasma. J. Astrophys. Astron. 43 (2), 43.CrossRefGoogle Scholar
Kumar, S., Poser, A.J., Schöttler, M., Kleinschmidt, U., Dietrich, W., Wicht, J., French, M. & Redmer, R. 2021 Ionization and transport in partially ionized multicomponent plasmas: application to atmospheres of hot Jupiters. Phys. Rev. E 103 (6), 063203.CrossRefGoogle ScholarPubMed
Kuznetsov, A.V. 1998 Thermal nonequilibrium forced convection in porous media. In Transport Phenomena in Porous Media (ed. D.B. Ingham & I. Pop), pp. 103–129. Pergamon.CrossRefGoogle Scholar
Mahajan, A. & Raj, M. 2024 The impact of internal heating on natural convection in a rectangular porous container. Chin. J. Phys. 90, 651663.CrossRefGoogle Scholar
Maheshwari, S.L. & Bhatia, P.K. 1976 Frictional effects with neutrals and Rayleigh–Taylor instability of a compressible Hall plasma. Beitr. Plasmaphys. 16 (4), 251261.CrossRefGoogle Scholar
Malashetty, M.S., Swamy, M. & Heera, R. 2008 Double diffusive convection in a porous layer using a thermal non-equilibrium model. Intl J. Therm. Sci. 47 (9), 11311147.CrossRefGoogle Scholar
Malashetty, M.S., Swamy, M. & Kulkarni, S. 2007 Thermal convection in a rotating porous layer using a thermal nonequilibrium model. Phys. Fluids 19 (5), 054102.CrossRefGoogle Scholar
Nield, D.A. & Bejan, A. 2013 Convection in Porous Media. Springer.CrossRefGoogle Scholar
Postelnicu, A. 2008 The onset of a Darcy–Brinkman convection using a thermal nonequilibrium model. Part II. Intl J. Therm. Sci. 47 (12), 15871594.CrossRefGoogle Scholar
Postelnicu, A. & Rees, D.A.S. 2003 The onset of Darcy–Brinkman convection in a porous layer using a thermal nonequlibrium model-part I: stress-free boundaries. Intl J. Energy Res. 27 (10), 961973.CrossRefGoogle Scholar
Qin, Y. & Kaloni, P.N. 1995 Nonlinear stability problem of a rotating porous layer. Q. Appl. Maths 53 (1), 129142.CrossRefGoogle Scholar
Rees, D.A.S. & Pop, I. 2005 Local thermal non-equilibrium in porous medium convection. In Transport Phenomena in Porous Media III (ed. D.B. Ingham & I. Pop), pp. 147–173. Elsevier.CrossRefGoogle Scholar
Sharma, R.C. 1972 Finite Larmor radius and Hall effects on thermal instability of a rotating plasma. Phys. Fluids 15 (10), 18221826.CrossRefGoogle Scholar
Sharma, R.C. & Sharma, K.C. 1978 Thermal instability of a partially ionized plasma. Austral. J. Phys. 31 (2), 181188.CrossRefGoogle Scholar
Sharma, R.C. & Sharma, Y.D. 1989 Taylor instability of partially-ionized plasma in porous medium in the presence of variable magnetic field. Astrophys. Space Sci. 155 (2), 295300.CrossRefGoogle Scholar
Sharma, R.C. & Sunil, 1995 Thermal instability of a compressible finite Larmor radius, Hall plasma in porous medium. Phys. Plasmas 2 (6), 18861892.CrossRefGoogle Scholar
Sharma, R.C. & Sunil, 1996 Thermal instability of a compressible finite-Larmor-radius Hall plasma in a porous medium. J. Plasma Phys. 55 (1), 3545.CrossRefGoogle Scholar
Sharma, S., Sunil, & Sharma, P. 2024 Stability analysis of thermosolutal convection in a rotating Navier–Stokes–Voigt fluid. Z. Naturforsch. A 79 (7), 689702.CrossRefGoogle Scholar
Shivakumara, I.S., Lee, J., Mamatha, A.L. & Ravisha, M. 2011 Boundary and thermal non-equilibrium effects on convective instability in an anisotropic porous layer. J. Mech. Sci. Technol. 25 (4), 911921.CrossRefGoogle Scholar
Soler, R. & Ballester, J.L. 2022 Theory of fluid instabilities in partially ionized plasmas: an overview. Front. Astron. Space Sci. 9, 789083.CrossRefGoogle Scholar
Spiegel, E.A. & Veronis, G. 1960 On the Boussinesq approximation for a compressible fluid. Astrophys. J. 131, 442447.CrossRefGoogle Scholar
Straughan, B. 2004 The Energy Method, Stability, and Nonlinear Convection. Springer.CrossRefGoogle Scholar
Straughan, B. 2006 Global nonlinear stability in porous convection with a thermal non-equilibrium model. Proc. R. Soc. Lond. A 462 (2066), 409418.Google Scholar
Straughan, B. 2008 Stability and Wave Motion in Porous Media. Springer.Google Scholar
Sunil, , Sharma, P. & Mahajan, A. 2010 Nonlinear ferroconvection in a porous layer using a thermal nonequilibrium model. Spec. Top. Rev. Porous Media: Intl J. 1 (2), 105121.CrossRefGoogle Scholar
Thakur, A., Kumar, S. & Devi, R. 2024 The effect of rotation on ferroconvection in the presence of couple stress forces in porous medium: a nonlinear analysis. Eur. Phys. J. Plus 139 (3), 236.CrossRefGoogle Scholar
Yadav, D., Bhargava, R. & Agarwal, G.S. 2013 Thermal instability in a nanofluid layer with a vertical magnetic field. J. Eng. Math. 80 (1), 147164.CrossRefGoogle Scholar
Yadav, D. & Lee, J. 2015 The effect of local thermal non-equilibrium on the onset of Brinkman convection in a nanofluid saturated rotating porous layer. J. Nanofluids 4 (3), 335342.CrossRefGoogle Scholar
Figure 0

Figure 1. Geometrical representation of the problem.

Figure 1

Figure 2. Comparison of Rayleigh–Darcy number $(Ra)$ for LTNE and LTE across various combinations of bounding surfaces. Solid curves depict $Ra$ variations for LTNE, while dotted curves show $Ra$ variations for LTE. Curves labelled aa, bb and cc show the variations of $Ra$ with wavenumber $(a)$ for rigid–rigid, rigid–free and free–free surfaces, respectively, under LTNE conditions. The curves labelled a$'$a$'$, b$'$b$'$ and c$'$c$'$ represent the variations of $Ra$ with $a$ for rigid–rigid, rigid–free and free–free bounding surfaces, respectively, under LTE conditions.

Figure 2

Figure 3. Variation of critical Rayleigh–Darcy number ($Ra_c$) with Darcy–Brinkman number ($\widetilde {Da}$) for the distinct combinations of bounding surfaces.

Figure 3

Figure 4. Comparison of critical Rayleigh–Darcy number $(Ra_c)$ for various values of scaled inter-phase heat-transfer coefficient $(\log _{10}\mathcal {H})$, for different values of porosity-modified conductivity ratio $(\gamma )$, for distinct combination of bounding surfaces. The blue curves represent rigid–rigid bounding surfaces, the red curves indicate rigid–free bounding surfaces and the black curves correspond to free–free bounding surfaces. The dashed curves are for $\gamma = 0.5$, the solid curves for $\gamma = 1$ and the dotted curves for $\gamma = 5$.

Figure 4

Figure 5. Variation of critical Rayleigh–Darcy number ($Ra_c$) with porosity-modified conductivity ratio ($\gamma$) for the distinct combinations of bounding surfaces.

Figure 5

Figure 6. Variation of critical Rayleigh–Darcy number ($Ra_c$) with the square of Hartmann number ($M^2$) for the distinct combinations of bounding surfaces.