On the behaviour of hot fluids in the presence of magnetic field

 

Mihir B. Banerjee1, J. R. Gupta1, R. G. Shandil1, Joginder S. Dhiman2 and Vinay Kanwar3

1Former Professor, Department of Mathematics, Himachal Pradesh University, Shimla-171005.

2Department of Mathematics, Himachal Pradesh University, Summerhill, Shimla-171005.

3Department of Mathematics, University Institute of Engineering and Technology, PU, Chandigarh.

*Corresponding Author E-mail:

 

In his masterly contributions to the theory of magnetohydrodynamic thermal convection  (or magnetoconvection) Chandrasekhar (1952, 1961) laid down the guiding principles of the subject matter of the following conjectural assertions of which the first two were obtained by making use of the method of Pellew and Southwell (1940) while the rest were speculated through an extremely simple solution of the governing equations that fails to satisfy any plausible set of boundary conditions on the magnetic field (except in the case of the (vi) for which the latter part of the remark does not apply) and pertains, according to him, to the case of dynamically free boundaries which occurs only in real physical situations. These assertions are;

(i)             The method of Pellew and Southwell (1940) is not quite strong enough to establish the validity or otherwise of the principle of exchange of stabilities for the problem.

(ii)           There exists a parameter regime for the problem defined by the Chandrasekhar number  remaining less than or equal to some fixed value  in which the total kinetic energy associated with a marginal or unstable disturbance exceeds its total magnetic energy and as a consequence the principle of exchange of stabilities  is valid for the problem in this parameter regime.

(iii)          A sufficient condition for the validity of the principle of exchange of stabilities for the problem is that the magnetic Prandtl number  is   less than the thermal Prandtl number .

(iv)         There exists a parameter regime for the problem defined by the Chandrasekhar number  remaining less than or equal to some fixed value  such that the principle of exchange of stabilities is valid for the problem in this parameter regime even if the magnetic Prandtl number  is   greater than the thermal Prandtl number .

(v)           There exists a parameter regime for the problem defined by the Chandrasekhar number  remaining greater than a critical value  such that the overstability is valid for the problem in this parameter regime while the onset of instability will be stationary convection in the parameter regime defined by  remaining less than or equal to  when the magnetic Prandtl number  is   greater than the thermal Prandtl number .

(vi)         If the principle of exchange of stabilities is valid for the problem then the asymptotic behavior of the critical Rayleigh number and the associated wave number for  are given by;    

  (                                                                                                                                     (1)

and it appears that the same power laws with the same constants of proportionality hold for situations with more general boundaries such as any combination of a dynamically free and a rigid boundary.

(vii)        If overstability  is valid for the problem then the asymptotic behaviours of the critical Rayleigh number and the associated wave number  and frequency  for  are given by; 

 

The sufficient condition for the validity of the principle of exchange of stabilities for the problems as given in (iii) is known as the Thompson-Chandrasekhar criterion in the literature on magnetohydrodynamic thermal convection (Thompson (1951) obtained it first in the context of inviscid fluids).

 

Subsequent contributions to the problem were made primarily to strengthen the fundamental steps enunciated by Chandrasekhar and widen their domain of applicability especially with regard to the consideration of more general boundaries in the theoretical formulation.

 

Gibson (1966) emphasized that since no such theorem as that established by Pellew and Southwell for the hydrodynamic case exists in the hydromagnetic case the possibility of overstability must be considered. He further argued that the boundary conditions assumed by Chandrasekhar in his investigations on (ii) and (vii) that pertain, according to him, to the case of dynamically free boundaries are not correct since they do not include the boundary conditions on the magnetic field and himself investigated the problem wherein the thermally perfectly conducting boundaries could be dynamically free or rigid that might be electrically perfectly conducting or insulating. Gibson concluded from his analysis that while the validity of asymptotic forms of  and        as Given in (vii) could be established along with the truth of the Thompson-Chandrasekhar criterion as given in (iii) in the limit when , it was not so in the case of the asymptotic form of   for      as given in (vii). Gibson’s analysis, however, has limited value since it is relevant only in the limit when      and further it is based on approximations on account of which the conclusions obtained by him are not fully justified.

 

Sherman and Ostrach (1966) examined the validity of the principle of exchange of stabilities for the problem wherein the fluid is confined within an arbitrary enclosed region with thermally perfectly conducting or insulating rigid boundary surfaces that are electrically perfectly conducting while the uniform magnetic field is applied in an arbitrary direction and contributed mainly through a restatement of (ii) and the establishment of the truth of Thompson-Chandrasekhar criterion as given in (iii) in the limit when      in his more generalized framework. The analysis of Sherman and Ostrach, however, is not free from assumptions and as a consequence his establishment of the truth of (iii) in the limit when     cannot be relied upon though his restatement of (ii) in the more general context is justified.

 

More recently Kumar et al. (1986) have investigated the problem wherein the thermally perfectly conducting boundaries are rigid and electrically perfectly conducting and making use of a slight variant of the well known Fourier-Chandrasekhar technique concluded that overstable motions exist in the limit when      even if the magnetic Prandtl number  is less than the thermal Prandtl number , a conclusion which invalidates (iii) and obtained the asymptotic forms of, ,  and  for      which are not in accordance with those that are given by (vii). However, a careful analysis of Kumar et al.’s work shows that their solution of the mathematical double eigenvalue problem is not correct since it does not satisfy, in general, the magnetic induction equation which constitutes one of the governing equations of the problem.

 

This was essentially the state of knowledge in the field of linear magnetoconvection that was largely dominated either by oversimplified mathematical calculations which did not conform to reality or by mathematical conjectures that were arrived at through the use of non-rigorous mathematical analysis and left unproven and uninvestigated when we took up the challenges of the problem at Simla in the early eighties.  The need to derive  a sufficient condition for the validity of the ‘exchange principle’ for more general boundary condition on the velocity and magnetic field than considered by Chandrasekhar, in the parameter space of the system alone, which would reduce to Pellew and Southwell’s result in the hydrodynamic case and correlate it to Chandrasekhar’s conjecture concerning the two energies was indeed pressing after Chandrasekhar’s investigations especially since Nature as well as Laboratory were unmistakably pointing towards the existence of a sufficient condition for the validity of ‘exchange principle’  that would be satisfied unless the conditions prevailing were extraordinary. From the very beginning we did not see anything wrong with the method of Pellew and Southwell which leads to equation

as a necessary condition for the existence of the overstability and apparently blocks the way for further progress towards any conclusive result while on the contrary we viewed the seemingly unfavourable sign attached with the magnetic field terms as a natural indicator of the possible fact that a sufficient condition for the validity of the ‘exchange principle’ is going to crucially depend upon the magnitude of , where  is a real independent variable such that ; , ,  and  are the dependent variables which are complex valued functions of z components of velocity, temperature and magnetic field.  This observation of ours fitted partly well with Chandrasekhar’s assertion (iv) and thus supported, it cut in deeper and made the task much easier, and all that remained before us to be completed was to obtain an upper bound of the term  in terms of the term  and successfully compare the magnitudes of this upper bound and the latter term under appropriate restrictions on the parameters of the system.  Little did we know at that time, however, that in effect our prog ramme was an attempt to settle Chandrasekhar’s conjecture concerning the two energies in the parameter space of the system alone and correlate it to the validity of the ‘exchange principle’.  The strategy as laid out above worked and yielded a result that was long sought after in the field of magnetoconvection and applicable for quite general boundary conditions, thus invalidating Chandrasekhar’s and subsequently Lin’s (1955) assertion that Pellew and Southwell’s method does not appear to be powerful enough to establish the validity of the ‘exchange principle’ for the problem on hand. More work has been completed on assertions (v), (v) and (vii) which include in particular a rigorous mathematical proof of  law of Chandrasekhar as given in assertion (vi).

 

Derivation of Governing Equations and Boundary Conditions of the Magnetohydrodynamic Simple Bénard Instability Problem

Consider a liquid that has the property of electrical conduction and suppose that magnetic fields are prevalent. The basis equations that express the interaction between the liquid motions and the magnetic fields are, of course, contained in the equations of Maxwell and appropriately modified hydrodynamical equations, and for our considerations here we shall present them under a fundamental simplification which is possible namely the simplification arising out of the neglect of displacement current in the Maxwell’s equations whose justification is derived from the fact that we are not concerned herein with effects that are related, in any way, to the propagation of electromagnetic waves. A closely related further possibility of ignoring any direct reference to the electrical charge density is offered not because it is small in itself but because its variations affect the equation expressing the conservation of charge only by terms of order (c being the velocity of light) which is of this order we can ignore with justification.

 

Making use of above simplifications the basic magnetohydrodynamical equations that govern the magnetohydrodynamic simple Bénard problem can be written in Cartesian coordinates as (cf. Chandrasekhar (1961))

The equation of continuity

                                                                                                                                                          (0.1)

The equation of motion 

                                                                                                                                          (0.2)

where,  is the square of the modulus of the magnetic field vector while the effect of the field on motions that results from the Lorentz force is taken care of through the last term within the second curly bracket on the right hand side of equation (0.2).

The equation of heat conduction

                                                                                                        (0.3)

The equation of state

                                                                                                                                                    (0.4)

The equation of magnetic induction

                                                                                                                                      (0.5)

where  is assumed to be a constant while the effect of liquid motions on the electromagnetic field is taken care of through the presence of second term within square bracket on the left hand side of equation (0.5).

The solenoidal character of the magnetic field

                                                                                                                                                                                   (0.6)

It may be noted that in writing down the above equations we have defined only those symbols that occur for the first time while those symbols that have occurred earlier in the text shall retain their same meanings and this procedure shall be generally followed in subsequent discussions as well.

 

Applying the usual Boussinesq approximation (1903),  which in essence amounts to neglecting terms which are of order 10-3 at most as compared to 1 for variations in temperature of order 10   (say), we obtain, respectively, from equations (0.1)-(0.6)  the following equations namely;

                                                                                                                                                                                   (0.7)

                                                                                                                        (0.8)

                                                                                                                                                              (0.9)

                                                                                                                                 (0.10)

                                                                                                                                      (0.11)

and                                                                                                                                                                            (0.12)

where,                                                                                                                                                   (0.13)

 

Let us consider a viscous finitely heat and electrically conducting Boussinesq liquid of infinite horizontal extension and finite vertical depth statically confined between two horizontal boundaries  which are maintained at constant temperatures and respectively in the presence of a uniform vertical magnetic field acting parallel and opposite to the direction of gravity.

 

The initial stationary state solution of the governing equations (0.1)-(0.6)

with is then represented by

,

,

,                                                                                                                         (0.14)

,

,

where,  is a constant and   is the value of  at

Let the initial state described by equation (A214) be slightly perturbed so that the perturbed state is given by

 

 

 

 

,

,

,                                                                                                            (0.15)

,

and ,

 

where,    are respective components of the perturbation magnetic field in the  direction and

                                                                                                                                                                           (0.16)

 

Then, the linearized perturbation equations of continuity, motion, heat conduction, magnetic induction, solenoidal character of the magnetic field and state are given by

                                                                                                                                                              (0.17)

,                                                                                                                   (0.18)

,                                                                                                              (0.19)    

,                                                                                                                                   (0.20)

                                                                                                                                                              (0.21)

                                                                                                                                                          (0.22)

                                                                                                                                                          (0.23)

                                                                                                                                                       (0.24)

                                                                                                                                                         (0.25)

and                                                                                                                                                                    (0.26)

 

We now analyze an arbitrary perturbation into a complete set of normal modes and then examine the stability of each of these modes individually. For the system of equations (0.17)-(0.26) the analysis can be made in terms of two dimensional periodic waves of assigned wave numbers. Thus, we ascribe to all quantities describing the perturbation a dependence on  of the form (cf. Chandrasekhar [1961])

                                                                                                                                                      (0.27)

and making use of (0.27) in equations (0.17)-(0.26), we obtain

                                                                                                                                 (0.28)

,                                                                                         (0.29)

,                                                                                        (0.30)

,                                                                                                              (0.31)

                                                                                                                                        (0.32)

                                                                                                                                      (0.33)

                                                                                                                                      (0.34)

                                                                                                                                     (0.35)

                                                                                                                                            (0.36)

where,    and  are functions of  only.

Eliminating    ,    and  from equations (0.28)-(0.36) and using the nondimensional quantities defined by

                ,    ,                                                                                            (0.37)

We obtain, upon dropping the double dashes and the asterisks for simplicity in writing, the governing magnetohydrodynamical equations in their non-dimensional forms as follows:

,                                                                                              (0.38)

                                                                                                                                                    (0.39)

                                                                                                                                              (0.40)

 

We now seek solutions of equations (0.38)-(0.40) which satisfy the appropriate boundary conditions on  which in turn depend on the nature of the horizontal boundaries under consideration. The boundary conditions on  which continue to be valid in this more general situation also,  and have been described in details in Chandrasekhar (1961)  and therefore it remains to consider the boundary conditions on  which depend on the electrical properties of the medium adjoining the liquid and in the following we shall consider two cases separately namely; (i) when the medium adjoining the liquid is electrically perfectly conducting, and  (ii) when the medium adjoining the liquid is electrically non conducting.

 

(i)       In the first case when the medium adjoining the liquid is a perfect conductor, it follows that no magnetic field can cross the boundary and we must require that

     on a plane boundary adjoining a perfect conductor.                                                                                   (0.41)

Resolving into normal modes, we have from equation (0.41) that

   on a plane boundary adjoining a perfect conductor.                                                                                    (0.42)

and non dimensionalizing equation (0.42) with the help of the transformations (0.37) and dropping the asterisk for convenience in writing, we obtain

 on a plane boundary adjoining a perfect conductor.                                                                                       (0.43)

We thus have the boundary conditions;

                                                                                                                                                        (0.44)

 

(ii)     In the second case when the medium adjoining the liquid is electrically nonconducting, the equation governing the magnetic field vector    outside the liquid layer is given by;

                                                                                                                                               (0.45)

where,       is the magnetic diffusivity of the region outside the liquid layer,     is its magnetic permeability and    is its electrical conductivity. We shall first derive the boundary conditions for finite values of    and then obtain the boundary conditions when    is infinite which is presently the case.

Resolving into normal modes, we have from equation (0.45) that

                                                                                                                                                 (0.46)

and nondimensionalizing equation (0.46) with the help of the transformations (0.37) and dropping the asterisk for convenience in writing, we obtain

                                                                                                                                                   (0.47)

where, .

 

The general solution of equation (0.47) is given by

                                                                                                (0.48)

and since we require that there are no sources of perturbations at ‘infinity’, we have from equation (0.48) that

             for                                                                                                                 (0.49)

and      for                                                                                                                 (0.50)

where  represent the lower and the upper horizontal boundaries.

Now, continuity of the magnetic field vector    ,   defined within the liquid layer, at the upper and the lower boundaries implies the continuity of     and     at the two boundaries. Further, the normal mode resolution of    and  implies the continuity of    and  at the two boundaries which together with equation (0.25) in its normal mode resolution form namely equation (0.36) yields that

                                                                                                                                                                                         (0.51)

is continuous on a plane boundary adjoining an electrically finitely  conducting medium.     

Nondimensionalizing equation (0.51) with the help of the transformations (0.37) and dropping the asterisk for convenience in writing we obtain

                                                                                                                                                                         (0.52)

is continous on a plane boundary adjoining an electrically finitely conduction medium. Therefore,

 at                                                                          (0.53)

and   at                                                                    (0.54)

since  is continuous at the boundaries.

In the particular case when the medium adjoining the liquid is electrically nonconducting, the boundary conditions given by equations (0.53) reduce to

    at                                                                                                                                                       (0.55)

    at                                                                                                                                                       (0.56)

 

Remark 1: If we take , the system of equations (0.38)-(0.40) together with relevant boundary conditions on yields the eigenvalue problem governing the Simple Bénard Convection Problem.

1.        Review of the Simple Bénard Instability Problem

Consider a double eigenvalue problem for p described by a system of ordinary linear and homogeneous differential equations with linear and homogeneous boundary conditions, namely (see Remark 1 and Chandrasekhar (1961))

                ,                                                                                                     (1.1)

and         ,                                                                                                                                 (1.2)

with                                                                             (1.3)

where z is the real independent variable such that , D is the differential operator  is a constant,  is a constant,  is a constant,  is a complex constant in general such that  and  are real constants and as a consequence the dependent variables  and  are complex valued functions of the real variable z such that  and , are real valued functions of the real variable z.  Let us briefly explain the sense in which we call it a double eigenvalue problem for p.

 

Operating on both sides of equation (1.1) by the operator  and eliminating  from the resulting equation by making use of equation (1.2), we obtain the following single equation in w, namely

                .                                                                     (1.4)

Further, since equations (1.1) and (1.2) are valid everywhere in  which includes the end points  and  also, we derive the following boundary conditions on w by making use of equations (1.3), namely

               

.                                                                                                                                          (1.5)

Equation (1.4) is an ordinary linear and homogeneous differential equation of order six with constant coefficients and consequently its general solution can be written as

                ,                                                                                                                     (1.6)

where, ci’s are arbitrary constants that are to be determined by the appropriate boundary conditions on w and wi’s constitute a fundamental set of solutions of equation (1.4).

 

Applying the six linear and homogeneous boundary conditions given by equations (1.5) on the above solution for w we obtain six algebraic linear and homogeneous equations for the determination of the six unknowns ci’s and since we are interested in non-trivial solutions of equation (1.4) that satisfy the boundary conditions specified by equations (1.5), we must have the coefficient determinant of the above system of algebraic equations equation to zero.  This leads to the characteristic equation of the problem which can be written in the form

                ,                                                                                                                                        (1.7)

where f is an appropriate complex valued function in general (since p is complex in general) of the arguments as shown and therefore equation (1.7), in general, has a real part and an imaginary part which can be written as

                ,                                                                                                                                        (1.8)

and         ,                                                                                                                                         (1.9)

where fr and fi are real and imaginary parts of f.  Now, if  and R are considered as prescribed and pr and pi are regarded as unknowns and further, if fr and fi  have favorable functional forms then we can, in principle, determine pr and pi as

                ,                                                                                                                                                                (1.10)

and         ,                                                                                                                                                 (1.11)

which satisfy equations (1.8) and (1.9).

 

Since two quantities are to be determined instead of a single one for non-trivial solutions of appropriate differential equations and boundary conditions, we call it a double eigenvalue problem.  One could have thought that the problem should still be called an eigenvalue problem and not a double eigenvalue problem since pr and pi are the real and imaginary parts of a single quantity p which is to be determined but a little consideration would yield that it cannot be done so uniformly because in the domain of applications of such system of equations and boundary conditions we have occasions when the problem is regarded as an eigenvalue problem for (say) R and pi for prescribed values of  and pr rather than as an eigenvalue problem for pr and pi for prescribed values of  and R.  We wish to solve the above double eigenvalue problem for pr and pi and thereby obtain, as a first step, sufficient conditions for the validity or otherwise of the result that  implies  for all  a proposition that will play a crucial role in the subsequent discussions in this article.

 

A few remarks at this point that concern the general mathematical nature of the problem will be helpful for the later development of the arguments.  One could obtain from equations (1.1), (1.2) and (1.3) the following single equation (that is similar to equation (1.4) but with  replacing w) and appropriate boundary conditions in .  These are

                ,                                                                     (1.4a)

with        ,                 (1.5a)

and it is clear, of course, that the eigenvalues for p, as functions of  and R and determined by the differential system (1.4) and (1.5) are the same as determined by the differential system consisting of equations (1.4a) and (1.5a).  Further, equations (1.4) or (1.4a) are self-adjoint while the differential system consisting of equations (1.4a) and (1.5) provided the bracketed alternatives in equations (1.5) or (1.4a) and (1.5a) are self-adjoint provided, only the bracketed alternatives in equations (1.5) and (1.5a) are taken into account.  Furthermore, the differential systems consisting of equations (1.4) and (1.5) or (1.4a) and (1.5a) are not self-adjoint provided the bracketed alternatives in equations (1.5) and (1.5a) are not taken into account.  If we consider the differential system consisting of equations (1.1), (1.2) and (1.3) as such then it is not self-adjoint although it could be made self-adjoint with the help of the transformation  and  or  and , and in both these cases equations (1.1), (1.2) and (1.3) reduce to

                ,                                                                                                (1.1a)

and         ,                                                                                                                     (1.2a)

with                                                                                        (1.3a)

and thus showing that by a different scaling a non-self-adjoint differential system may be transformed into a self-adjoint one.  If now we obtain from equations (1.1a), (1.2a) and (1.3a) a single equation and appropriate boundary conditions for W or , we fall back on equations (1.4) and (1.5) or (1.4a) and (1.5a) and thus showing that combining differential system of equations to reduce them to a higher order one can destroy a property like self-adjointness.

 

A complete solution of the eigenvalue problem, which effectively is self-adjoint, consists in the determination of the countable infinity of eigenvalues  and the corresponding eigen-functions wj and  and although the problem is self-adjoint and the associated linear equations have constant coefficients, the solution of it is not entirely trivial and requires an investigation.  An attempt to solve the problem through equations (1.4) and (1.5) from the very fundamentals as indicated earlier does not succeed although equation (1.4) has exponential and sinusoidal solutions, since the resulting characteristic equation is transcendental and thus making it impossible to obtain a simple explicit solution for   which in turn blocks the way for a settlement of the proposition that we intend to seek.  One the other hand, if we consider the problem through equations (1.1), (1.2) and (1.3), express w and  in terms of known functions and adjustable constants (which are inherently capable of doing so) such that the appropriate boundary conditions are satisfied and choose the adjustable constants so as to satisfy the relevant differential equations, we end up with a characteristic equation in the form of a determinantal equation wherein the determinant that contains p is of order infinity and thus making it impossible to obtain a simple explicit solution for .  Furthermore, an attempt to characterize the eigenvalues through equations (1.1), (1.2) and (1.3) (say) by associating with them a variational principle, which is always possible for a linear problem, also fails to serve the purpose since although a stationary property of the eigenvalues can be established, the minimum or the maximum property of the eigenvalues loses its meaning in view of the eigenvalues being complex in general.

 

A precise calculation of the eigenvalues of linear and homogeneous differential systems or the development of meaningful characterizations of the eigenvalues of such systems in the absence of a precise calculation, are mathematical problems in their own right and thus from the point of view of a mathematician there is obviously a problem contained in equations (1.1), (1.2) and (1.3).  Several important mathematical methods have been constructed in the recent past and the present, for example the Rayleigh- Ritz method, the Galerkin method, the Chandrasekhar method, the method of weighted residuals (Finlayson (1972)) and etc. to investigate such problems and a challenging problem in this context is a welcome since it always has the potential to become the starting point of a new mathematical method which could possibly be exploited in other allied domains of enquiry.  From the point of view of a physicist, however, equations (1.1), (1.2) and (1.3) have a special charm in that contained in these equations are motions in the form of cellular structures that manifest in a class of hydrodynamical systems which had defied a theoretical explanation for a long time.  These equations were, in fact, written by Lord Rayleigh (1916), a Noble Prize winning physicist, in a pioneering contribution to the Philosophical Magazine when he was trying to understand and theoretically explain some of the experimental observations of the French experimentalist Henery Bénard (1900).  Bénard was experimenting in a definitive manner to demonstrate and quantitatively characterize the onset of first thermal motions in liquids, and for a appreciation of his findings We pose and analyze the consequences of the following thought experiment.  Consider a horizontal layer of liquid in which an adverse temperature gradient is maintained by heating it from below.  We qualify the temperature gradient thus maintained as adverse since on account of thermal expansion the liquid at the bottom will be lighter than the liquid at the top and this will give rise to a top heavy arrangement which is potentially unstable.  On account of this latter instability there will be a natural tendency on the part of the liquid to redistribute it-self and remedy the weakness in its arrangement.  This tendency of the liquid, however, will be inhibited by its own viscosity and as a consequence, we expect that the temperature gradient which is maintained will have to exceed a certain value before the first thermal motions can manifest themselves. One of the findings of Bénard who subjected horizontal liquid layers bounded below by metallic plate with the upper surface left free to be in contact with the ambient air to uniform vertical adverse temperature gradients is precisely the above but in addition to it he discovered another fact that is as fundamental as his above finding and cannot be easily derived by making use of a common sense analysis of the type employed above, and more relevantly in the present context it is this latter finding that has a dominant bearing on the main theme of the present article.  It will suffice here to summarize the principal findings of Bénard and other contemporary experimentalists and these are, first, a certain critical adverse temperature gradient must be exceeded before the onset of first thermal motions, and second, the motions that ensure on surpassing the critical temperature gradient have a stationary cellular character.  What actually happens at the onset of motions is that the layer becomes reticulated and reveals its dissection into cells and if the experiment is performed with sufficient care the cells become equal, hexagonal and properly aligned.

 

Rayleigh described these Bénard motions as the motions that would manifest on account of the hydrodynamical instability suffered by the initial stationary state in which a static layer of liquid of infinite horizontal extension and finite vertical depth is under the action of a uniform adverse vertical temperature gradient so that the temperature decreases linearly with height while the density increases linearly with height with an appropriate supporting hydrostatic pressure in the force field of gravity.  It can be easily checked that the above purely conducting state of the liquid constitutes an exact solution of the equations that express the conservation of mass, momentum and energy (with constant thermal conductivity) and an appropriate equation of state together with the required boundary conditions in a relevant framework.  To analyze the hydrodynamic instability of the above stationary state of the liquid Rayleigh added infinitesimally small perturbations to it which are inevitably present in any experiment, argued that such a disturbed state should also be a solution of the same governing equations of which the initial stationary state is a solution which is justified, simplified the equations governing the perturbations on the basis of the Boussinesq approximation which is relevant to the context, obtained the linearized equations governing the perturbations which is also justified since one is interested in the characterization of the onset of first thermal motions only and not beyond it, analyzed the perturbations with the help of the usual normal mode analysis which for the problem under consideration is in terms of two dimensional periodic waves of assigned wave numbers and is in accordance with the Fourier series methods which is justified in the present context, and with ent as the           t-dependence of the perturbations, where  is a complex constant in general so that nr and ni are real constants and which is allowed since the resultant linear and homogeneous system of perturbation equations contains t-only through their derivatives with respect to t (the initial stationary state solution being independent of t),  combined the governing equations of the problem into equations (1.1), (1.2) and (1.3) which are in the non-dimensional forms (see § 0).  The mathematical meanings of the various symbols used in these equations have been defined earlier, however, from the physical point of view they represent the following quantities, namely, t is the time, z is the vertical coordinate, z = 0 and z = 1 are the lower and upper horizontal boundaries, D is the differential operator  is the square of the wave number,  is the Prandtl number where  is the kinematic viscosity and  is the thermometric conductivity;  where pr is the amplification factor, pi is the circular frequency, and d is the depth of the layer;  is the Rayleigh number where g is the gravity,  is the thermal expansion,  is the uniform adverse temperature gradient which is maintained; w is the vertical velocity and  is the temperature.  Rayleigh posed a double eigenvalue problem for pr and pi as a consequence of the validity of equations (1.1), (1.2) and (1.3) and keeping in mind the experimental requirements for the manifestation of the Bénard motions enquired about that locus  in the space which represent the envelope of the loci  in the space obtained from equation (1.10) by equating by equating  with respect to all admissible values of .  Such a locus would separate in the space the regions of  for all  (complete stability) from the regions of  for at least one  (instability on account of the linearity of the problem) and would present case) on which  for some  and further that when the system becomes unstable as we cross  at some particular point the mode of perturbation which will be manifested (at onset) will be the one whose locus  touches  at that point.  Rayleigh made another enquiry in this connection, namely whether  implies  for all  in which case a stationary pattern of motions prevails at neutral stability and one says (following Jeffreys (1926)) that the principle of exchange of stabilities is valid otherwise oscillatory motions prevail at neutral stability and then one says (following Eddington (1926)) that one has a case of overstability in which case pi  will have determinate values on .  The principal result established by Rayleigh is that  implies  for all  and  is a line segment of infinite extent parallel to the axis of R such that  where  corresponds to the value of R given by  for .  The above result when interpreted in the context of the experiments of Bénard means that if the temperature of the lower boundary is gradually increased so that R increases then so long as R remains less than  we should not observe any motions because all the inevitable infinitesimally small perturbations decay in time and are therefore stable for  while on the other hand we should observe the onset of first thermal motions in the form of a stationary pattern that is characterized by a wave number  at an instant when R becomes  because this particular aperiodic perturbation of wave number  is the only perturbation that neither grows nor decays in time and is neutrally stable at the least value of R which is  while all other perturbations decay in time and are therefore stable at , and further when R is just greater than  this particular perturbation grows in time and is therefore unstable.  Rayleigh associated this perturbation with the first thermal motions at the onset in the experiments of Bénard and subsequent experiments performed since then bear ample testimony of the fundamental correctness of his ideas, notwithstanding the experimental investigation of Block (1956) and the theoretical analysis of Pearson (1958), which raise it almost to a level of certainty that most of the motions observed by Bénard (being in very thin layers with a free surface) were driven by the variation of surface tension with temperature and not by thermal instability of light liquid below heavy liquid, since Rayleigh’s theory is in accord with layers of liquid with rigid boundaries and on thicker layers with a free surface because the significance of the variation of surface tension as compared to that of buoyancy diminishes as the thickness of the layer increases.  Indeed, it was a masterly contribution that opened up new chapters in theoretical hydrodynamics and is chiefly responsible for the profound developments in the theories of hydrodynamic instability and bifurcation.  However, Rayleigh had made a simplifying assumption while solving the mathematical problem, that both the boundaries are dynamically free which implies that tangential viscous stresses do not act on them though a viscous liquid is allowed to slip on them and as a consequence the cases wherein  vanishes on any one of the boundaries or both drop out from the considerations of equations (1.3) thus leaving the resulting eigenvalue problem in terms of w self-adjoint (in an appropriate function space with a suitable scalar product defined as earlier mentioned) and from which Rayleigh could prove the result that  implies  for all  or equivalently the ‘exchange principle’, and carry out the subsequent calculations.  Such boundaries, however, are unreal in any physical situation except possibly in relation to meteorology and thus arose the crucial need for proving the ‘exchange principle’, which is so necessary for taking a first step towards an explanation of the stationary pattern of the Bénard motions as was demonstrated by Rayleigh, for more frequently occurring real boundaries such as rigid boundaries on which a viscous liquid does not slip though tangential viscous stresses may act on them and this leads to the consideration of precisely those conditions of equations (1.3) which were left out in Rayleigh’s analysis and for which the resulting eigenvalue problem in terms of w is a non self-adjoint one.  Harold  Jeffreys in an important paper in 1926 in the Philosophical Magazine examined this assumption in some detail showing that in effect it disregards the possibility that two exponential time factors may coalesce and become complex conjugate quantities so as to have finite imaginary parts when their real part vanishes and cites Rayleigh as having disposed of this possibility in relation to boundary conditions of a special type and himself makes appeal to experiments to confirm that it is not the type of motion that in fact arises when equilibrium breaks down.  The first decisive step in this direction was taken by Anne Pellew and Richard Southwell (1940) in a pioneering paper in the Proceedings of the Royal Society of London wherein they established the validity of the ‘exchange principle’ in a comprehensive manner which include all cases considered under boundary conditions given by equations (1.3).  The technique utilized by Pellew and Southwell to prove the ‘exchange principle’ is essentially simple and we present this ‘proof’ below not only for the sheer delight of its elegance and simplicity but also for its bearing on the main theme of this article.

Multiplying both sides of equation (1.1) by (*denoting the complex conjugate) and integrating the resulting equation over the vertical range of z, we obtain

               

                               

                                .                                                                                                                              (1.12)

Taking the complex conjugate of both sides of equation (1.2), we have

                .                                                                                                                           (1.13)

 

Multiplying both sides of equation (1.13) by  and integrating the resulting equation over the vertical range of z, we get

               

                                                                .                                                                                                       (1.14)

Substituting for  in equation (1.12) from equation (1.14), it follows that

               

                                .                                                                                (1.15)

Evaluating the definite integrals  in the desired forms by integrating them by parts for an appropriate number of times in each case and making use of the boundary conditions given by equations (1.3), we derive

                ,                                                                                                                    (1.16)

                ,                                                                                                                    (1.17)

and         .                                                                                                                         (1.18)

It is important to note here that equations (1.16), (1.17) and (1.18) hold good whether the boundaries are both dynamically free or both rigid or any one of them dynamically free and the other rigid as can be easily seen.

Making use of equations (1.16), (1.17) and (1.18) we can write equation (1.5) as

               

                                .                                                                                 (1.19)

Equating the imaginary part of both sides of equation (1.19) and rearranging the resulting equation, we obtain

                                .                                                                                 (1.20)

Equation (1.20) implies that  since  irrespective of  and this establishes the validity of the ‘exchange principle’ for quite general boundary conditions.  The technique of Pellew and Southwell has, since then, become celebrated in the domain of hydrodynamic instability and the above result obtained by them has become the starting point of several research investigations besides inspiring research workers using slight variations of their technique to obtain results which would otherwise have been difficult to derive.

 

The linear theory of hydrodynamic instability of a liquid layer heated underside as propounded by Rayleigh, together with reality and completion brought to it by Pellew and Southwell, has served the purpose of a foundation on which scores of experimental and theoretical investigations stand, and it would be fairly accurate to state that this line of thought is complete in almost all its aspects notwithstanding its limitations.

 

At this point we leave the considerations of the simple Bénard instability problem and the associated Pellew and Southwell’s ‘method of proof’ to join with it again a little later in the context of a more general physical problem wherein the liquid has the property of electrical conduction and magnetic fields are prevalent.

 

2.  Magnetohydrodynamics and magnetohydrodynamic simple Bénard instability problem

Hans Alfvén (1942) a Swedish physicist had made a discovery in the dynamics of electrically conducting liquids pervaded by magnetic fields.  The motion of electrically conducting liquids across the magnetic lines of force generates electric currents and the associated magnetic fields modify the original magnetic fields and thus the motions contribute to the changes in the original magnetic fields.  On the other hand, additional mechanical forces are generated which act on the current carrying liquid elements when they traverse magnetic lines of force and thus the magnetic fields contribute to the changes in the liquid motions.  This twofold interaction between the motions and the fields which are responsible for pattern of behavior that are often unexpected and striking constitute essentially the subject matter of magneto hydrodynamics.  With a masterly physical insight Alfvén chose Faraday’s method of regarding magnetic action as represented by the magnetic lines of force; realized that conducting liquids of large dimensions would behave as perfect conductors; derived for such situations his famous theorem of frozen-in-fields which implies that in a moving perfectly conducting liquid the magnetic lines of force are frozen in the liquid; argued that the magnetic lines of force must then behave as taut material strings possessing both inertia and tension with a mass per unit length being equal to the density  of the liquid and under tension  (since in a magnetic field of intensity H pervading a medium of permeability  the magnetic stresses are equivalent to a tension  along the magnetic lines of force and a hydrostatic pressure  which can be balanced by a diminution in the liquid pressure), anticipated by analogy with transverse vibrations of taut strings that when the liquid is slightly displaced from rest the magnetic lines of force would perform transverse vibrations, the phase velocity of the waves generated being  = ; and gave a neat mathematical demonstration of it all in one single paper in 1942 which announced the discovery simultaneously in Nature and Arkiv. f. mat. Astr. O. Fysik. It showed for the first time that transverse waves can be propagated in an electrically conducting incompressible liquid pervaded by a magnetic field and so energy can be transmitted without large scale exchanges of the liquid elements.  Alfvén was awarded the Nobel Prize for physics for this work and several other fruitful ideas that he introduced and worked on in magnetohydrodynamics and plasma physics, and theory of electrically conducting liquids in the presence of magnetic fields grew rapidly in its fundamental as well as applied aspects after 1942.

 

This twofold magnetohydrodynamic interaction was considered by Alfvén (1950) and Walen (1949) in discussing processes in the Sun, by Fermi (1949), Richtmeyer and Teller (1949) in a theory of the origin of cosmic rays, and by Elasser (1946) and Bullard (1949) in an explanation of terrestrial magnetism.  Bullard (1949) proposed a reasonably simple scheme of thermal convection currents and the magnetic fields within the core of the earth which might be able to maintain the observed field at the surface by the self-inductive mechanism originally proposed by Larmor, and showed that the magnetic field is of prime importance in determining the flow.

 

As a first step toward investigating the hydrodynamical aspects of the process suggested by Bullard a study was made by Thompson in 1951 in the Philosophical Magazine, of the slow thermal convection currents set up in a plane layer of an electrically conducting liquid placed in a homogeneous magnetic field and heated from below which in effect amounts to a re-examination of the simple Bénard instability problem for the case when the liquid considered is an electrical conductor and an external magnetic field is impressed on the liquid.  Thompson modified the analysis of Rayleigh and Jeffreys to include the ponderomotive effects of the magnetic field by adding to the hydrodynamical and the heat flow equations Maxwell’s equations for the electromagnetic fields and one of the fundamental questions to which he addressed himself to was whether the ‘exchange principle’ was valid for this more general problem or otherwise.  His analysis which was unlike that of Pelew and Southwell yielded the result that if the uniform applied magnetic field is oppositely aligned with the direction of the non-magnetic body force then a sufficient condition for the validity of the ‘exchange principle’ is this that the thermal diffusivity  (thermometric conductivity as stated earlier) is less than or equal to the magnetic diffusivity , a condition met by a large margin under most terrestrial and laboratory conditions.  Unfortunately, however, Thompson’s analysis is somewhat limited since due attention is not paid either to hydrodynamic or to magnetohydrodynamic boundary conditions, and more importantly, his analysis ignores the effect of viscosity while deriving this condition and as a consequence his results cannot be relied upon.  In the discussion that follows we present an analysis of this problem which is essentially due to Thompson except for deviations on points of minor detail and derive his sufficient condition for the validity of the ‘exchange principle’.

 

3.       The governing equations and Thompson’s condition for the ‘exchange principle’

The corresponding double eigenvalue problem for p obtained as a result of an imposition of a uniform magnetic field acting parallel to gravity on the simple Bénard configuration of an electrically conducting liquid are given by the following system of ordinary linear and homogeneous differential equations with linear and homogeneous boundary conditions.  These are (cf. equations (0.38) - (0.40))

,                                                                                  (3.1)

                                ,                                                                                                                 (3.2)

and                         ,                                                                                                              (3.3)

with         on both the horizontal boundaries which are

                                                                                                                                 (3.4)

              on a rigid boundary,                                                                                                            (3.5)

            on a dynamically free boundary,                                                                                       (3.6)

                    on both the boundaries if the regions outside the liquid are perfectly conducting                                                                                                                                                                  (3.7)

and                 on the upper and lower boundary respectively if the regions outside the liquid are insulating,                                                                                                                                                          (3.8)

where  is a constant and stands for the Chandrasekhar number where  is the magnetic permeability and H is the uniform externally imposed magnetic field,  is a constant and stands for the magnetic Prandtl number,  is a complex valued function of the real variable z in general such that  and  are real valued functions of the real variable z, and  is the vertical component of the perturbation in the initial externally imposed magnetic field.  All other symbols used in the above equations have the same mathematical and physical meanings as given earlier.

For a non-viscous liquid the above equations and boundary conditions become

,                                                                                                  (3.9)

                                ,                                                                                                                 (3.10)

and                         ,                                                                                                               (3.11)

with         on both the horizontal boundaries which are

                                                                                                                                 (3.12)

 on both the boundaries if the regions outside the liquid are perfectly conducting,                                                                                                                                                                                       (3.13)

and         on the upper and lower boundary respectively if the regions outside the liquid are insulating,                                                                                                                                                                                 (3.14)

where  are both positive.

Operating on both sides of equation (3.9) by the operator    and eliminating  and  from the resulting equation by making use of equations (3.10) and (3.11), we obtain the following single equation in w, namely

.                                         (3.15)

Assuming a particular simple solution for w in the form

                constant    ,                                                                                                                               (3.16)

which satisfies the boundary conditions prescribed for w, and substituting this solution for w in equation (3.15), we obtain the characteristic equation

.                                                               (3.17)

Assuming that overstability is valid so that , we can rewrite equation (3.17) in the form

.                                                                                        (3.18)

Separating the real and imaginary parts of equation (3.18), we obtain

,                                                                                            (3.19)

and         .                                                                                                         (3.20)

Rewriting equations (3.19) and (3.20) in the forms

,                                                                                                              (3.21)

and         ,                                                                                                       (3.22)

and substituting for  in equation (3.22) from equation (3.21), we obtain on simplification

                 .                                                                                                      (3.23)

Equation (3.23) enables us to draw at once one conclusion and that is that the solutions describing overstability cannot occur if

                 ,                                                                                                                                                                 (3.24)

Because in that case  would be negative which is contrary to the hypothesis.

Thus, a sufficient condition for the validity of the ‘exchange principle’ is that the thermal diffusivity  is less than or equal to the magnetic diffusivity , a condition met by a large margin under most terrestrial and laboratory conditions as mentioned earlier.  In the literature on magnetoconvection this sufficient condition for the validity of the ‘exchange principle’ is known as the Thompson-Chandrasekhar condition, and constitutes an important building block for the analysis of the problem in the non-linear domain.

 

4.       Extension to Viscous Case and Chandrasekhar’s First Method

Almost during the time that Thompson’s paper appeared, the same physical configuration was also being investigated by one of the great men of science of our times, Chandrasekhar, a Nobel Prize winner in physics, who happened to look at the freshly arrived December 1951 issue of the Philosophical Magazine containing Thompson’s paper while he himself was completing his own paper for the press.  A comparison between the papers of Thompson and Chandrasekhar shows that both of them were seeking answers on two fundamental points of enquiry, namely (i) does the ‘exchange principle’ hold in this more general situation also? And (ii) does the magnetic field postpone the onset of two papers go.  On the other hand, whereas Thompson was primarily motivated from the point of view of terrestrial applications as a consequence of which he was tempted to ignore the viscous terms from the governing equations for the benefit of achieving mathematical simplification while analyzing point (i), Chandrasekhar in addition to terrestrial applications had astrophysical applications also in mind on account of which he retained the viscous terms throughout the analysis as it was known then that viscosity does at times play a rather important role in astrophysical problems.

 

Since the answer to point (ii) crucially depends upon the answer to point (i), Chandrasekhar analyzed point (i) in great detail in his fundamental paper in the Philosophical Magazine in 1952, implementing every possible investigation that he could bear upon it as his paper shown and devised in the process two different methods of attack to the problem, and since this part of his work is closely connected with the subsequent investigations presented in this article we given below a brief description of it in all its essentials.

 

Chandrasekhar’s first method consists in combining the equations (3.1), (3.2) and (3.3) in terms of the single dependent variable w and then assuming a particularly simple solution for it for the case when both the boundaries are dynamically free.  This leads him directly to the characteristic equation of the problem and subsequently to a sufficient condition for the validity of the ‘exchange principle’.  The various steps of this scheme of calculations are shown below.

 

Operating on both sides of equation (3.1) by the operator    and eliminating  and from the resulting equation by making use of equations (3.2) and (3.3), we obtain the following single equation in w, namely

               

                .                                              (4.1)

Assuming a particularly simple solution for w in the form

                constant    ,                                                                                                                               (4.2)

which satisfies the prescribed boundary conditions for w, for the case when both the boundaries are dynamically free, and substituting this solution for w in equation (1.4.1), we obtain the characteristic equation

.                                                              (4.3)

Assuming that overstability is valid so that , we can rewrite equation (4.3) in the form

.                                                                   (4.4)

Separating the real and imaginary parts of equation (1.4.4), we obtain

,                                                            (4.5)

and         .                                                                  (4.6)

Rewriting equations (1.4.5) and (1.4.6) in the forms

,                                                                                    (4.7)

and         ,                                                                                                (4.8)

and substituting for  in equation (4.8) from equation (4.7), we obtain on simplification

                 .                                                                    (4.9)

Equation (1.4.9) enables us to draw one conclusion at once and that is that solutions describing overstability cannot occur if

                 ,                                                                                                                                                              (4.10)

Because in that case  would be negative which is contrary to the hypothesis. Recalling the definitions of  and , the foregoing condition is equivalent to

                 ,                                                                                                                                                                 (4.11)

and this is precisely Thompson’s condition as mentioned earlier.

However, the solution for w as given by equation (4.2) is not correct mathematically (and Chandrasekhar is aware of it) since it would then imply from equations (3.3) and (4.2) that

                                on both the boundaries,                                                                                     (4.12)

which in turn would imply from equations (3.3) and (4.2) that

                                on both the boundaries,                                                                                                     (4.13)

since .  Equations (3.3), (3.7) or (3.8) and (4.13) would then lead us to conclude that  satisfies

                (constant) ,                                                                                         (4.14)

and         .                                                                                                          (4.15)

The general solution of equation (4.14) is given by

                ,                                                                            (4.16)

where A and B are constants of integration, and equation (4.15) then implies

                                ,                                                                                                      (4.17)

                ,                                                                                  (4.18)

                                ,                                                                                                                                        (4.19)

                .                                                                                                                      (4.20)

 

Equations (4.19) and (4.20) yield that A = 0 = B while equations (4.17) and (4.18) imply that A = 0 = B cannot be allowed.  In other words, equations (4.14) and (4.15) do not admit of a solution for  and this shows that the assumed solution for w as given by equation (4.2) cannot be correct (Banerjee et al 1988).  In fact, the above analysis shows that the correct solution for w must be such that  must not vanish on one of the boundaries at least (Bhattacharjee 1987).  Furthermore, it is clear from the above analysis that the assumed solution for w as given by equation (3.16) in the non-viscous case is also not correct and thus the derivation of the Thompson Chandrasekhar condition in both the cases suffers from deficiencies in mathematical analysis even if the condition itself, which does yield Pellew and Southwell’s result when the liquid under consideration is electrically non-conducting and magnetic fields are absent, may be correct.  The construction of the correct solution for w and the answer regarding the validity or otherwise of the ‘exchange principle’ in the present context wherein both the boundaries are dynamically free, has been a subject matter of intense research activity and controversy (Gibson (1966); Sherman and Ostrach (1966); Kumar et. al. (1986); Banerjee et al (1988)) in the recent past, and we leave these considerations here for the present to take them up again at the end of the article  and show the way the matter has been settled by Banerjee et al (1988) in favour of the classic calculations of Chandrasekhar.

 

5.       Chandrasekhar’s Second Method and His Conjecture

Chandrasekhar’s second method, which was designed to remove the deficiencies of the analysis as given under the first method, consists in extending the proof of the ‘exchange principle’ developed by Pellew and Southwell to include the effects of electrical conductivity of the liquid and a uniform applied magnetic field in the vertical direction.  The various steps of this scheme of calculation are shown below.

 

Multiplying both sides of equation (3.1) by  and integrating the resulting equation over the vertical range of z, we obtain

    .                                                      (5.1)

The definite integral on the left hand side and the first definite integral on the right hand side of equation (5.1) can be expressed, with the help of appropriate boundary conditions in the forms exactly as in the Pellew and Southwell’s case.  For the second definite integral on the right hand side of equation (5.1), we have on integrating by parts once and making use of the boundary conditions given by equations (3.4)

.                                                                                             (5.2)

Taking complex conjugate of both sides of equation (3.3), we get

                .                                                                                                                     (5.3)

Substituting for  in the right hand side of equation (5.2) from equation (5.3), we obtain

.                                                                    (5.4)

 

Evaluating the definite integral by integrating by parts once and making use of the boundary conditions (3.7), we obtain

                 .                                                                                                                           (5.5)

Making use of equations (1.16), (1.17), (1.18), (5.2), (5.3), (5.4) and (5.5), we can write equation (5.1) as

.                                                                                                     (5.6)

Equating the imaginary part of both sides of equation (5.6) and rearranging the resulting equation, we get

.                                               (5.7)

 

It is clear from equation (5.7) that one cannot conclude from it, as in the Pellew and Southwell’s case wherein , that  implies  for all , and this is on account of the fact that the terms containing the magnetic field misbehave as regards their signs in equation (5.7).  Chandrasekhar remarked that “it will appear that the method of Pellew and Southwell is not quite strong enough to establish the principle for the problem on hand”.  However, Chandrasekhar did make a conclusion from equation (5.7) and an important one at that undoubtedly, namely that if the total kinetic energy associated with a perturbation (which equals  is greater than or equal to its total magnetic energy (which equals  then  for all .  Since, however, one cannot apriori be certain when this sufficient condition will be satisfied, being in terms of unknown eigensolutions w and  of the problem, Chandrasekhar sought without success the regime in terms of the parameters of the system alone in which the validity of his sufficient condition will be ensured.  Unfortunately, he did not present a rigorous proof of his sufficient condition and as a consequence it came to be known in the literature on magnetoconvection as Chandrasekhar’s conjecture.  A rigorous derivation of this conjecture together with the establishment of its regime of validity in the parameter space of the system alone were much sought after questions to be resolved since this work of Chandrasekhar.  Sherman and Ostrach (1966), in an important paper in the Journal of Fluid Mechanics, successfully answered the first question in a more general context when the liquid is completely confined in an arbitrary region and the uniform magnetic field is applied in any arbitrary direction but failed to resolve the second.  From 1966 to 1984 the state of art in this particular domain of enquiry remained more or less the same and we do not witness any worthwhile contribution that advances the frontiers established by the fundamental investigations of Thompson and Chandrasekhar.

 

6.       A Sufficient Condition for the ‘Exchange Principle’

The need to derive a sufficient condition for the validity of the ‘exchange principle’ for more general boundary conditions on the velocity and magnetic field, than considered by Chandrasekhar, in the parameter space of the system alone which would reduce correlate it to the Chandrasekhar’s conjecture concerning the two energies was indeed pressing after Chandrasekhar’s investigations especially since the Nature as well as the Laboratory were unmistakably pointing towards the existence of a sufficient condition for the validity of the ‘exchange principle’ that would be satisfied unless the conditions prevailing were extraordinary.

 

This was essentially the state of knowledge in the field of linear magnetoconvection that was largely dominated either by oversimplified mathematical calculations which did not conform to reality or by mathematical conjectures that were arrived at through the use of non-rigorous mathematical analysis and left unproven and uninvestigated when we took up the challenges of the problem at Shimla in the early eighties.  From the very beginning we did not see anything wrong with the method of Pellew and Southwell which leads to equation (5.7) and apparently blocks the way for further progress towards any conclusive result while on the contrary we viewed the seemingly unfavourable sign attached with the magnetic field terms as a natural indicator of the possible fact that a sufficient condition for the validity of the ‘exchange principle’ is going to crucially depend upon the magnitude of .  This observation of ours fitted partly well with another conjecture of Chandrasekhar in this field of enquiry and made at a later date, 1961, with the help of his specialized solution of the problem as mentioned in the first method, namely, for , there exists a  such that for  the onset of instability will be as stationary convection, while for  it will be as overstability.  The observation thus supported, cut in deeper, and made the task easier and all that remained before us to be completed was to obtain an upper bound for the term  in terms of the term  and successfully compare the magnitudes of this upper bound and the latter term under appropriate restrictions on the parameters of the system.  Little did we know at that time, however, that in effect our programme was an attempt to settle Chandrasekhar’s conjecture concerning the two energies in the parameter space of the system alone and correlate it to the validity of the ‘exchange principle’.  The strategy as laid out above worked and yielded a result that was long sought after in the field of magnetoconvection and applicable for quite general boundary conditions, thus invalidating Chandrasekhar’s and subsequently Lin’s (1955) assertion that Pellew and Southwell’s method does not appear to be powerful enough to establish the validity of the ‘exchange principle’ for the problem on hand.  The various steps of this scheme of calculations are shown below.

 

(a) Case of perfectly conducting boundaries (Banerjee et al 1984)

In this case, since both the horizontal boundaries are perfectly conducting, boundary conditions given by equations (3.7) hold good and not those given by equations (3.8).  Thus the double eigenvalue problem for p is given by the following system of equations and boundary conditions:

,                                                                                  (6.1)

                                ,                                                                                                                 (6.2)

and                         ,                                                                                                              (6.3)

with         .                                                                              (6.4)

 

Multiplying both sides of equation (6.4) by , integrating the resulting equation over the vertical range of z and evaluating the various definite integrals by making use of the remaining equations (6.2) and (6.3), and boundary conditions given by equations (6.4), as shown earlier, we get equation (5.6) which is given by

- .                                                                                                   (6.5)

Equating the imaginary part of both sides of equation (6.5) and rearranging  throughout by assuming , we obtain

  .                                                   (6.6)

Multiplying equation (6.3) by  and integrating the resulting equation over the vertical range of z, we get

.                                                                                                     (6.7)

Integrating the left hand side of equation (6.7) by parts for a suitable number of times and making use of the boundary conditions on  given by equations (6.4), we obtain

.                                                                            (6.8)

Equating the real part of both sides of equation (6.8), we have

.                                                      (6.9)

Further, Real part of

                                                                    

                                                                    

                                                                    

                                                                     .                                                                                (6.10)

Assuming , we derive from equation (6.9) and inequality (6.10) that

.                                                                                         (6.11)

Combining inequality (6.11) with the Rayleigh-Ritz inequality (Schultz (1973)), namely

                (since ),                                                             (6.12)

we derive

,                                                                                                          (6.13)

and further, combining inequality (6.11) with inequality (6.13) we obtain

.                                                                                    (6.14)

Making use of equation (6.6) and inequality (6.14), we get

,                                 (6.15)

and therefore we must have

,                                                                                                                                (6.16)

or            ,                                                                                                        (6.17)

and which cannot hold good if

                 .                                                                                                                                               (6.18)

 

Thus, if  and , then  must vanish for all .  In other words, a sufficient condition for the validity of the ‘exchange principle’ is that  is less than or equal to , a condition that was long sought after since the fundamental papers of Thompson and Chandrasekhar.  However, the mathematical analysis employed in case (a) is no longer applicable when both the horizontal boundaries are insulating, and this is on account of the fact that the Rayleigh-Ritz inequality (6.12) involving the function , which was utilized in the analysis of case (a), may not be valid when the boundary conditions given by equations (3.8) instead of (3.7) are taken into consideration.  We overcame this difficulty by constructing an alternative proof and showed that this not only yields a sufficient condition for the validity of the ‘exchange principle’ which is uniformly applicable to both the cases under boundary conditions given by equations (3.7) or (3.8) but also improves upon the contents of inequality (6.18) that is applicable for case (a).  The various steps of this scheme of calculations are shown below.

 

(b) Case of insulating boundaries (Banerjee et al 1985a)

In this case, since both the horizontal boundaries are insulating, boundary conditions given by equations (3.8) hold good and not those given by equations (3.7).  Thus, the double eigenvalue problem for p is given by the following system of equations and boundary conditions:

                ,                                                                                  (6.19)

                ,                                                                                                                                 (6.20)

and         ,                                                                                                                              (6.21)

with        ,                                                                   (6.22)

and         .                                                                              (6.23)

 

Multiplying both sides of equation (6.19) by  and integrating the resulting equation over the vertical range of z, we obtain

               

                                                .                                                                               (6.24)

Evaluating the definite integral on the left had side and the first definite integral on the right hand side of equation (6.24) as before, we get

               

                                                ,                                                                                (6.25)

and        

                                   .                                                                           (6.26)

 

For the second definite integral on the right hand side of equation (6.24), we have integrating by parts once and making use of the boundary conditions given by equations (6.22)

                .                                                                                              (6.27)

Taking complex conjugate of both sides of equation (6.21), we get

                ,                                                                                                                         (6.28)

and substituting for  in the right hand side of equation (6.27), we derive

               

 

                                .                                                                   (6.29)

Evaluating the definite integral  by integrating by parts once and making use of the boundary conditions given by equations (6.23), we obtain

                 .                                                                              (6.30)

Making use of equations (6.25), (6.26), (6.27), (6.29) and (6.30), we can write equation (6.24) as

               

                 .                                                                 (6.31)

Equating the imaginary part of both sides of equation (6.31), rearranging the resulting equation and cancelling  throughout by assuming , we get

  

.                                                                                              (6.32)

Multiplying both sides of equation (6.21) by  and integrating the resulting equation over the vertical range of z, we have

.                                                                                                     (6.33)

Integrating by parts once the definite integral on the right hand side of equation (6.33) and making use of the boundary conditions given by equations (6.22), we get

.                                                                                                                                    (6.34)

Making use of equations (6.3) and (6.34), equation (6.33) can be written as

.                                                                                                     (6.35)

Equating the real part of both sides of equation (6.35), we obtain

                                .                                                                                             (6.36)

Further, - Real part of

                                                                               

                                                                               

                                                                               

                                                                                .                                                                  (6.37)

Assuming , we derive from equation (6.36) and inequality (6.37) that

,                                                            (6.38)

and hence

                                .                                                                           (6.39)

Making use of the boundary conditions on w given by equations (6.22), we have by the Rayleigh-Ritz inequality

                .                                                                                                                       (6.40)

Combining inequality (6.32) with inequalities (6.39) and (6.40), we obtain

,                                                                (6.41)

and therefore, we must have

,                                                                                                                                                  (6.42)

and which cannot hold good if

.                                                                                                                                                         (6.43)

 

Thus, if  and , then  must vanish for all .  In other words, a sufficient condition for the validity of the ‘exchange principle’ is that  is less than or equal to , a condition that was long sought after since the fundamental work of Chandrasekhar in 1961 wherein he anticipated something similar through his .

 

Another important conclusion that can be derived from the foregoing analysis is that the same sufficient condition is valid also for the case when both the horizontal boundaries are perfectly conducting, since in this case boundary conditions given by equations (3.7) yield that

,                                                                                                                                (6.44)

and thus the proof given in case (b) is uniformly applicable to perfectly conducting as well as insulating boundaries and improves upon the contents of inequality (6.18) as shown by inequality (6.43).

 

However, surprisingly, Banerjee et al did not see the connection between their own work and Chandrasekhar’s conjecture concerning the two energies and thus the second question raised in section 1.5 remained partially answered only.  This gap in the literature on magnetoconvection has been recently completed by Banerjee et al (1985b) who presented a simple mathematical proof to establish that Chandrasekhar’s conjecture is valid in the regime  and further that this result is uniformly applicable for any combination of a dynamically free or a rigid boundary when the regions outside the liquid are perfectly conducting or insulating.  We present this proof below for the case of insulating boundaries from where follows the proof for the case of perfectly conducting boundaries.

 

7.       Resolution of Chandrasekhar’s Conjecture Concerning the Two Energies (Banerjee et al 1985b)

Consider the case when both the horizontal boundaries are insulating so that the double eigenvalue problem for p is given by the system of equations (6.19), (6.20) and (6.21) and boundary conditions (6.22) and (6.23).

Multiplying both sides of equation (6.21) by , integrating the resulting equations over the vertical range of z and proceeding exactly as in section 1.6(b) we get inequality (6.39), namely;

                                .                                                                           (7.1)

It follows from inequality (7.1) that

.                                                                                                                (7.2)

Making use of boundary conditions on w given by equations (6.22), we have by Rayleigh-Ritz inequality

.                                                                                                                               (7.3)

Combining inequalities (7.2) and (7.3), we obtain

,                                                                                                       (7.4)

and hence

.                                                                                   (7.5)

Therefore, if , it follows from inequality (7.5) and

.                                                                               (7.6)

 

We thus have the result that the total kinetic energy associated with a neutral or unstable perturbation is greater than or equal to its total magnetic energy in the regime , since the left hand side of inequality (7.6) represents the total kinetic energy associated with a perturbation while the right hand side represents its total magnetic energy.

In the case when both the horizontal boundaries are perfectly conducting so that the boundary conditions given by equations (3.7) hold good and not those given by equations (3.8), we have

                                                                                                                                 (7.7)

and it then follows from the above analysis that inequality (7.6) holds good in the regime .

 

The foregoing result when considered in the context of the result given in section (1.6) shows the link between Chandrasekhar’s conjecture and the validity of the ‘exchange principle’ which was missing in the literature, and settles the conjecture in the regime .  Surprisingly, Banerjee et al (1984, 1985a) did not pursue their investigation in this direction and consequently did not see this connection which was implicitly contained in their work.  The mathematical analysis presented in this section, however, has surely an edge over that given in the two previous sections, for reasons of (i) its striking simplicity and directness (ii) showing that the validity of Chandrasekhar’s conjecture is a consequence of only the equation of magnetic induction together with the magnetic boundary conditions and the fact that the horizontal boundaries are fixed and (iii) being the first in this context to explicitly claim a long sought after result.

 

Banerjee et al. (1989) and Dhiman (1995) further extended the results of Banerjee et al. (1985b) respectively for hydromagnetic thermohaline convection and rotatory magneto thermohaline convection problems of Veronis’ and Stern’s types.

 

8.       Solutions for the Case When ‘Exchange Principle’ is Valid

In his pioneering work on the initiation of magnetoconvection wherein the boundaries are dynamically free and the ‘exchange principle’ is valid Chandrasekhar (1961) derived analytically, consequences of importance regarding the inhibiting effect of the magnetic field and the asymptotic dependence of the critical Rayleigh number  on the Chandrasekhar number Q as  through a correct guess, in its simplest and closed form, of the exact solution of the relevant governing equations.  The various steps of his scheme of calculations are presented below.

 

When both the boundaries are dynamically free and the ‘exchange principle’ is valid the neutral state will be characterized by  and hence the relevant governing equations and boundary conditions (3.1) – (3.8) reduce to

                ,                                                                                                          (8.1)

                ,                                                                                                                                         (8.2)

and         ,                                                                                                                                          (8.3)

with                                                                                                          (8.4)

(both the boundaries dynamically free)                                                                        

or                                                                                                                  (8.5)

(both the boundaries rigid)             

,                                                                                                                                          (8.6)       

(if the regions outside the liquid are perfectly conducting )

and                                                                                                                                                                   (8.7)

(if the regions outside the liquid are insulating)                                                                         

Eliminating  from equation (8.1) by making use of equation (8.3), we obtain

                ,                                                                                                              (8.8)

while eliminating  from equation (8.8) by applying the operator  and making use of equation (8.2), we obtain the following single equation in w, namely

                .                                                                                         (8.9)

We must seek solution of equations (8.9) which satisfy the boundary conditions

,                                                                            (8.10)

in deriving which we have made use of equation (8.8) and the boundary conditions given by equations (8.4).

If follows from equation (8.9) and (8.10) that

,

and then by differentiating equation (8.9) with respect to z for an even number of times we can likewise successively conclude that all even derivatives of w must vanish .  The proper solution for w appropriate for the lowest mode is therefore given by

                ,                                                                                                                                                  (8.11)

where A is a constant.

Substituting the above solution for w in equation (8.9), we obtain the characteristic equation

                 ,                                                                                                    (8.12)

which clearly shows the inhibiting effect of the magnetic field on the onset of instability.  As a function of , R given by equation (8.12) attains its minimum when

                ,                                                                                                                             (8.13)

and with  determined as a solution of equation (8.13) which gives , equation (8.12) will give the critical Rayleigh number

                ,                                                                                                                        (8.14)

and thus the asymptotic behaviours of  and  are given by

 and                                                                                                       (8.15)

of which the first result constitutes Chandrasekhar’s celebrated law.  It may be noted that since neither equation (8.9) nor boundary conditions (8.10) involve , it is clear that the solution of the underlying eigenvalue problem can be carried out independently of boundary conditions on the magnetic field though these are needed for a complete solution of the problem.  However, in the absence of such a simple solution for w as that given by equation (8.11) pertaining to the considerations of more general boundaries like the rigid boundaries etc. That equation governing  and Q can no longer be derived in a closed form and thus it is extremely difficult to derive results through analytical methods.  For such cases that concern more general boundaries, Chandrasekhar analyzed the equation governing  and Q by the method of successive approximations and on the basis of the results of the first approximation which were obtained through numerical computations, conjectured that the magnetic field exercises an inhibiting effect on the onset of instability and the law is valid provided the ‘exchange principle’ is satisfied.  A rigorous mathematical demonstration of the validity or otherwise of the above predictions of Chandrasekhar is missing in the literature on the subject till the recent times to the best of our knowledge except a general result of Banerjee et al (1988) which states that a necessary condition for the validity of the ‘exchange principle’ in magnetoconvection with quite general boundaries is that R exceeds , a result which shows the inhibiting effect of the magnetic field on the onset of instability and is in accordance with Chandrasekhar’s prediction.  We present Banerjee et al’s investigation below.

 

The relevant governing equations and boundary conditions for the situation under consideration are given by equations (8.1)-(8.10) with .

 

In that case, eliminating  from equation (8.1) by making use of equation (8.3), we obtain

                .                                                                                                              (8.16)

Multiplying equation (8.16) by  and integrating the resulting equation over the vertical range of z, we get

                .                                                                   (8.17)

Evaluating the various definite integrals by making use of the remaining equation (8.2), namely

                ,                                                                                                                                        (8.18)

and boundary conditions (8.4) – (8.7), we get

                                                .                                                                                              (8.19)

Multiplying equation (8.18) by  and integrating the resulting equation over the vertical range of z, we get

                .                                                                                                              (8.20)

Integrating the left hand side of equation (8.20) by parts once and making use of the boundary conditions on  given by equations (8.4), we obtain

               

                                                          

                                                          

                                                          

                                                            .                                                 (8.21)

                                                                                                (Cauchy-Schwartz inequality)

It follows from inequality (8.21) that

 

or            .                                                                                                                 (8.22)

Combining inequalities (8.21) and (8.22), we get

 .                                                                                               (8.23)

Making use of the boundary conditions on w given by equations (8.4), we have by the Rayleigh-Ritz inequality

,                                                                                                                     (8.24)

so that inequality (8.23) reduces to

 .                                                                                     (8.25)

Combining equation (8.19) with inequality (8.25), we obtain

     ,                                                                          (8.26)

and therefore, we must have

                ,

or            ,                                                                                                                                                          (8.27)

and this establishes the result of Banerjee et al.  However, the result  as  which was also predicted by Chandrasekhar for the case of quite general boundaries and constitutes his law cannot be derived from the present analysis and thus remained an open problem.

 

Motivated by the above problem, Dhiman and Kumar (2012) reinvestigated the problem of onset of thermal convection in an electrically conducting fluid layer heated from below in the presence of magnetic field and tried to validate the Chandrasekhar’s conjecture regarding  law  for all combinations of rigid and dynamically free boundaries using Galerkin technique.

 

Following Finlayson (1972), and applying the Galerkin’s method to find the critical value of Rayleigh number by taking a single term in the expansion for  

 

Therefore,   taking

   and    

where   and are the suitably chosen trial functions which satisfy the respective boundary conditions given in (8.4)-(8.7) and  and  are arbitrary constants. Further, multiplying the resulting equations (obtained after substituting the above trial functions in (8.8) and (8.2) by   and respectively, integrating each of the resulting equations by parts, using the relevant boundary conditions (3.4)-  (3.7) and eliminating  constants  and  ,we obtain following expression for Rayleigh number as (after dropping the subscripts);

                                                                                                  (8.28)

 

The various suitable trial functions chosen for the different combinations of boundary conditions are as below:

Case I:    and                                                                                                               (8.29)

( both dynamically free boundaries)

Case II:   and                                                                                                             (8.30)

( both rigid boundaries)   

Case III:  and                                                                                                         (8.31)

(Lower rigid and upper free boundary)

               

Now, considering each of the case of boundary conditions above and find the value of the critical Rayleigh number as follows;               

Case I: When both boundaries are dynamically free

Evaluating the value of  by substituting (8.29) in (8.28), for the present case of boundary conditions, we have

                                                                  (8.32)

 

The minimum of  i.e.   (the critical value of Rayleigh number corresponding to the onset of stationary convection) for the present case exist for  the positive root of the cubic given by;

.                                                                                            (8.33)

 

Now, for , equation (8.33) yields

,

which implies that for  a2 = 4.92

 

= 664.5251.

 

which is very close to the critical value of Rayleigh number  obtained by Chandrasekhar (1961) for Bénard problem (657.511) in absence of magnetic field.

 

Table 1: The values of critical Rayleigh numbers , for Case I of boundary conditions,  for  various values of  and the corresponding values of  associated wave numbers given by (8.33)

Chandrasekhar  Results

Results by Galerkin  Method.

 Q

           ac

           ac2

           Rc

          ac

             ac2

            Rc

0

2.233

4.9863

657.511

2.2270

4.9594

664.5249

5

2.432

5.9146

796.573

2.432

5.9408

804.5119

10

2.590

6.7081

923.070

2.5959

6.7387

931.8232

20

2.826

7.9863

11554.19

2.8330

8.0257

1164.3691

50

3.270

10.6929

1762.04

3.2775

10.7420

1775.7072

100

3.702

13.7048

2653.71

3.7098

13.7624

2671.9889

200

4.210

17.7241

4258.49

4.2196

17.8049

4284.2424

500

4.998

24.9800

8578.28

5.0094

25.0936

8622.1088

1000

5.684

32.3079

15207.0

5.6968

32.4533

15273.621

5000

7.585

57.5322

63135.9

7.6275

58.1792

63337.962

6000

7.839

61.4499

74632.1

7.8793

62.0840

74863.275

10000

8.588

73.7537

119832

8.6252

74.3941

120171.78

 

Case II: When both boundaries are rigid.

Evaluating the value of  by substituting (8.30) in (8.28), for the present case of boundary conditions, we have

                                                                                                        (8.34)

 

The minimum of  i.e.  ( the critical value of Rayleigh number corresponding to the onset of stationary convection ) for the present case exist for the positive root of the cubic given by

                                                                                                                            (8.35)

Now, for , equation (8.35) yields

,

which implies that for   

=1749

which is very close to the critical value of Rayleigh number  obtained by Chandrasekhar for Bénard problem (1715.1) in absence of magnetic field.

 

 

Table 2: The values of critical Rayleigh numbers , for Case II of boundary conditions,  for  various values of  and the corresponding values of  associated wave numbers given by (8.35).

Chandrasekhar  Results

Results by Galerkin  Method.

Q

ac

ac2

Rc

ac

ac2

Rc

0

3.13

9.7969

1715.1

3.1165

9.7127

1749

10

3.25

10.5625

1953.7

3.2599

10.6270

1996

50

3.68

13.5424

2811.4

3.6689

13.4615

2906.8351

100

4.00

16

3767.6

4.0068

16.0547

3950.0581

200

4.45

19.8025

5499.9

4.4553

19.8501

5886.1131

500

5.16

26.6256

10122

5.2098

27.1424

11199.0

1000

5.80

33.64

17116

5.8954

34.7560

19425.0889

2000

6.55

42.9025

30139

6.6782

44.5984

34993.4982

4000

7.40

54.76

54712

7.5617

57.1787

64752.959

6000

7.94

63.0436

78405

8.1279

66.0634

100934.2345

8000

8.34

69.5556

101622

8.5531

73.1572

126786.3153

10000

8.66

74.9956

124523

8.8970

79.1578

150292.1626

 

 

Case III: When one boundary is rigid and the other is free

Evaluating the value of  by substituting (8.31) in (8.28), for the present case of boundary conditions, we have

                                                                                       (8.36)

The minimum of  i.e.  ( the critical value of Rayleigh number corresponding to the onset of stationary convection ) for the present case exist for  the positive root of the cubic given by

                                                                                                             (8.39)

Now, for , equation (8.39) yields

 

which implies that for  a2 =7.13         =1138.7

which is very close to the critical value of Rayleigh number  obtained by Chandrasekhar for Bénard problem (1112.7) in absence of magnetic field.

 

It is to note that the Case IV (Lower boundary free and upper boundary rigid) of the boundary conditions has not treated separately, since the results of Case III of the boundary conditions are same as that of Case IV.  Further, this combination of boundary conditions is also difficult to realize physically.

 

Table 3: The values of critical Rayleigh numbers , for Case III of boundary conditions,  for  various values of  and the corresponding values of  associated wave numbers given by (8.39).                                                                       

Chandrasekhar  Results

Results by Galerkin  Method

Q

ac

ac2

Rc

ac

ac2

Rc

0

2.68

7.1824

1112.7

2.6698

7.1276

1138.7003

2.5

2.75

7.5625

1179.4

2.7356

7.4835

1209.341

12.5

2.97

8.8209

1428.3

2.9518

8.7132

1475.8114

25

3.17

10.0489

1712.7

3.1567

9.9647

1784.4809

50

3.45

11.9025

2231.3

3.4556

11.9401

2354.6292

125

4.00

16.000

3600.2

4.0012

16.00987

3886.9737

250

4.50

20.2500

5627.5

4.5226

20.4535

6193.4502

500

5.10

26.0100

9318.5

5.1317

26.3345

11350.548

1000

5.75

36.0625

16143

5.8274

33.9585

18381.22

2000

6.50

42.2500

28893

6.6124

43.7231

33368.889

5000

7.65

58.5225

64861

7.7981

60.8107

73901.542

10000

8.65

74.8225

122155

8.8194

77.7820

144167.93

 

From the results shown in Tables 1-3 for various cases of boundary conditions, it is apparent that all three cases exhibit the same general features. With a2 determined as a solution of cubic equations (8.35), (8.37) and (8.39), equations (8.32), (8.34) and (8.36) respectively give the required critical Rayleigh number   for each case of boundary conditions.   The inhibiting effect of the magnetic field on the onset of instability is clear from the results and for sufficiently large Q, the asymptotic behaviour of  and  , namely        and      is  verified. The above conclusions clearly establish the   -Law for each of the Cases of boundary conditions. Thus, the present analysis establishes that the  -Law is true for all types of boundary conditions and validates the claim of Chandrasekhar. Further, the obtained results are in good agreement with the numerical results of Chandrasekhar.

 

Chandrasekhar while analyzing the magnetoconvection problem by following usual variational techniques predicted that the lowest characteristic value of  (i.e. ) is indeed a true minimum. But critical examination of his analysis reveals that his claim regarding the minimum property of the functional is not valid, since one cannot establish the necessary orthogonal property of the eigen functions for general nature of boundary conditions (Cases III-IV of boundary conditions).

 

The celebrated  –Law for the magnetoconvection thermal instability problem was also validated by Dhiman et al. (2014) following the route of variational Principle. To prove this result they followed the analysis of Chandrasekhar (1961) with a different expression for  in terms of ;   

 

Following the variational method adopted by Chandrasekhar for thermal convection problem and proceeding analogously, one can easily prove the stationary property of the functional  (see equation (140), Chandrasekhar (1961)) for the Cases I to IV of  boundary conditions, when the quantities on right hand side are evaluated in terms of true characteristic functions. Also the quantity on the right hand side of above mentioned equation (140) attains its true minimum when  belongs to , i.e. the lowest  characteristic value of , namely , is indeed a true minimum. i.e.

                                                                                                          (8.40)

where ,                                                                                                                         (8.41)

Further, it is remarkable to note that the above result is uniformly valid for all cases of boundary conditions. This establishes the variational principle for the magnetoconvection of a fluid layer heated from below.

 

We shall now prove -Law for the present problem for each case of boundary conditions.

In view of inequality (8.27), it suffices to prove that

         for large .

To prove above inequality, let us consider a trial function

                                                                                                                                                               (8.42)

which obviously satisfies the boundary conditions

                                                                                                      (8.43)

where the origin has been shifted to the midway for convenience in computation.

Now, using the above defined value of in equation (8.41), we obtain

                                                                                                                               (8.44)

whose general solution is obtained as

                                                                                                          (8.45)

where,                                                                                                                                   (8.46)

and ,  are constants given below. Further, and are the roots of the auxiliary equation of equation (8.44). Now, using the various boundary conditions on , we have

                                                    (8.47)

and

                                                    (8.48)

 

where,       

and         .                                                    

Now, evaluating integrals and by using the expression for  and  given in (8.42) and (8.45) respectively, we have

                                                                                                             (8.49)

                                         (8.50)

Substituting the above values of integrals and  in inequality (8.40), we get

                                                                 (8.51)

Inequality (8.51) can be written in a convenient form as;

                                                                                                                                             (8.52)

where

                                                                                                                 (8.53)

Now, using the values of and  from (8.47) and (8.48) in (8.53) respectively for each case of boundary conditions (I)-(IV) and then simplifying the resulting expressions, we get

                    (8.54)

Taking . Further, and (using their definitions) are of the order of (i.e. ). Using this asymptotic dependence of  in (8.54) for sufficiently large values of , we get

for each case of boundary conditions.

Therefore, inequality (8.51), in view of above value of , yields that

,                                

which for sufficiently large values of  yields that

                              (for large values of ).                                                                                                      (8.55)

So, combining inequalities (8.27) and (8.55), we have

;      for large values of .

This establishes the -Law for each case of boundary conditions (I)-(IV) and thus, validates the claim of Chandrasekhar (1961) and Dhiman and Kumar (2012). Further, for sufficiently large values of , the asymptotic behaviour of  , namely   for given has also been verified by Dhiman et. al.(2014) by computing the values of  using the right hand side expression of inequality (8.51) for various values of  and the associated wave numbers  Moreover, the obtained values for  are in good agreement with the values obtained by Chandrasekhar [3] for the problem.

 

9.       Solutions for the Case when Overstability is Valid.  Settlement of the Controversy

As pointed out in the concluding part of Section 1.4, the construction of the corrects solution for w and the answer regarding the validity or otherwise of the ‘exchange principle’ in magnetoconvection wherein both the boundaries are dynamically free, has been a subject matter of intense research activity and controversy in the recent past.  In this section we take-up the problem again and show the way the matter has been settled by Banerjee et al (1989) in favour of the classic calculations of Chandrasekhar.  The various steps in stepson Banerjee et al’s scheme of calculations are presented below.

 

The governing equations and boundary conditions for the magnetoconvection problem wherein the dynamically free boundaries are thermally insulating and electrically perfectly conducting and a uniform magnetic field acting parallel to gravity is impressed upon the system are given by (cf. Section 3)

                ,                                                                                  (9.1)

                ,                                                                                                                                 (9.2)

                ,                                                                                                                              (9.3)

with        ,                                                                                  (9.4)

 

wherein the various symbols used in the above equations have the same meanings as given in section 1.3 with the difference that the origin of z is translated to be mid way between the two horizontal boundaries for the sake of convenience, and boundary conditions that are relevant to thermally insulating boundaries (Normand et al 1977) replace the corresponding ones that that are relevant to thermally perfectly conducting boundaries.

 

Combining the above equations and boundary conditions in an appropriate manner we derive the following systems of equations and the associated boundary conditions in terms of w alone,  alone and  alone, namely

               

 

                ,                                                                                                                                                              (9.5)

,                                                     (9.6)

                ,                                                                                                                                                               (9.7)

,                                                        (9.8)

                ,                                                                                                                                                               (9.9)

,                                                        (9.10)

where    

                                ,                                        (9.11)

                ,                                                             (9.12)

               

,                                                    (9.13)

               

                                                                ,                                                        (9.14)

                ,                               (9.15)

and        

                                                                                .                                              (9.16)

 

We may first observe that it follows from the eveness of the operator L that occurs in equations (9.5), (9.7) and (9.9), and the identity of the boundary conditions that have to be satisfied at  as given respectively by equations (9.6), (9.8) and (9.10), that the proper solutions of equations (9.5), (9.7) and (9.9) fall into two non-combining groups of even and odd solutions.  Further, it follows from equation (9.2) that proper solutions for w and  must either be both even or both odd while equation (9.3) implies that the proper solutions for w and  must neither be both even nor both odd.  From these considerations and the considerations of the corresponding hydrodynamic problem with dynamically free and thermally insulating boundaries it follows that the proper solutions for w and  must be odd while that for  must be even.  Therefore if  and  are constants the function  is even and since it is required to vanish at , we can expand it in a Fourier cosine series in the form

                .                                                                  (9.17)

With  given by equation (9.17), equation (9.3) becomes

               

,                                                    (9.18)

which upon integration yields

               

   ,                                              (9.19)

where  is a constant of integration.

The requirement that the above solution for w satisfies the boundary conditions as specified by equation (9.4) leads to a unique determination of ,  and  which are given by

,                                               (9.20)

,                                               (9.21)

and         .                                                                                                                                                                                (9.22)

Making use of equations (9.19) – (9.22), we obtain the proper solution for w as

 .                                                       (9.23)

With w given by equation (9.23), equation (9.5) becomes

 .                                           (9.24)

where

                ,                                        (9.25)

                ,                                       (9.26)

                ,                                                                                                           (9.27)

and        

                               

                                               

                                                                .                                                     (9.28)

Multiplying equation (9.24) by  (since the first derivative with respect to z of the left hand side of equation (9.24) vanishes at ) and integrating the resulting equation over the vertical range of z, we obtain

                                                                                     (9.29)

               

where  is the Kronecker’s delta.

Equations (9.29) provide a set of linear and homogeneous equations for the constants  and the requirement that the determinant of this system of equations must vanish provides the characteristic equation for the determination of R and pi when .  We thus obtain

 .                                                                                                    (9.30)

The nth approximation to the eigenvalues of R and pi is obtained by setting the nth order determinant consisting of the first n rows and columns in the left hand side of equation (9.30) to zero, and this corresponds to the retention of first n terms only in the Fourier expansion of  as given by equation (9.17).  The corresponding result is

                ,                                                (9.31)

from which it follows uniquely that the lowest eigenvalue of R and the associated value of pi are given by

                ,                                                                                                                                                                 (9.32)

Since   and  are non-zero numbers for every permissible value of n except  and  respectively which  does not vanish for any permissible value of n.  Further, since equation (9.32) is valid whatever by the value of n it follows that it is the unique solution that provides the lowest eigevalue of R and the associated value of pi as given by the characteristic equation (9.30).

 

With w given by equation (9.23),  can be determined in accordance with equation (9.2) together with the relevant boundary conditions on  as specified by equation (9.4).  Thus, we obtain

 .                                                     (9.33)

 

We complete the solution of the problem by demonstrating that w,  and  which are respectively given by equations (9.23), (1.933) and (9.17) and satisfy equations (9.2) and (9.3) along with the boundary conditions (9.4) also satisfy equation (9.1).

To prove this we consider equation (9.5) which can be written in an alternative form as

                ,                                                                                           (9.34)

where     ,                                                                                                                                                 (9.35)

and         ,                                                       (9.36)

and for w,  and as given respectively by equations (9.23), (9.33) and (9.17), we have

                .                                                                                                (9.37)

Multiplying equation (9.34) by , integrating the resulting equation over the vertical range of z by parts a suitable number of times and making use of equations (9.37), we get

               

                                                .                                                                                            (9.38)

Equating the imaginary part of both sides of equation (9.38), we obtain

.                                                                                              (9.39)

Since, , it follows from equation (9.39) that

                ,

which in turn implies that equation (9.1) is also satisfied.

 

The characteristic equation of the problem that we have derived herein is identical with the characteristic equation (1.4.3) which was first derived by Chandrasekhar through an extremely simple solution for w (cf equation (1.4.2)) of equation (9.5) that does not satisfy all the boundary conditions as specified by equation (9.6) and leading to a solution for hz that fails to satisfy any plausible set of boundary conditions required of a magnetic field.  However, the present analysis shows that Chandrasekhar’s characteristic equation (1.4.3) and the subsequent conclusions based on it was derived by him are valid for the problem with dynamically free and thermally insulating boundaries and therefore this problem stands completely solved.  However, the problem of finding a necessary and sufficient condition for the validity of the ‘exchange principle’ for the relatively simpler case of magnetohydrodynamic simple Bénard instability problem with dynamically free and thermally perfectly conducting boundaries continues to remain unresolved and its solution is awaited with great interest.

 

Gupta et al (1984, 85, 86, 88) have shown that the scheme of calculations presented in Section 6 are of much wider generality than the simple context in which they are given and derived important conclusions on the characterization of the neutral states for these more general instability problems.

 

Kumar et al. (1986) have investigated the problem wherein the thermally and electrically perfectly conducting boundaries are rigid. However, a careful analysis of Kumar et. al.’s work shows that their solution of the mathematical double eigen value problem is not correct since it does not satisfy, in general, the magnetic induction equation which constitutes one of the governing equations of the problem. In the above analysis Banerjee et. al. (1989) have obtained a exact solution of this problem wherein the electrically perfectly conducting and dynamically free boundaries are thermally insulating and proved for this allied problem (on which  not vanishes;  is the perturbation in temperature) that the eigenvalues calculated by Chandrasekhar through a extremely simple solution of the governing equations that fails to satisfy any plausible set of boundary conditions on the magnetic field are correct ones.

 

However, since this was not the problem to which Chandrasekhar addressed himself, therefore Banerjee and  Bhowmick (1992) examined the magnetohydrodynamic thermal stability problem with dynamically free and thermally and electrically perfectly conducting boundaries and attempted to construct a exact solution of the problem by appropriately modifying their earlier analysis (1989) of the problem with dynamically free and thermally insulating and electrically perfectly conducting boundaries. However, a close and critical examination of their analysis reveals a flaw in their analysis and invalidates their claim to have obtained a correct exact solution of the problem.

 

Dhiman and Sharma (2013) presented a correct solution of the problem by appropriately rectifying the flaw in the earlier analysis of Banerjee and Bhowmick (1992). They showed that in the analysis of the incorrectness of the solution may be attributed to the observation of  Banerjee and Bhowmick that the relevant solutions for  and  must satisfy the restriction

                                                         (9.40)

so as to satisfy equation (9.1) and the thermally perfectly conducting boundary condition, namely . However, this restriction on  and  and subsequently solving for  from equation (9.2) with  given by equation (9.19) together with the boundary condition  are mutually inconsistent steps and thus result in an incorrect solution of the problem. In fact the nature of the solutions for  and coupled with equation (9.1) clearly show that solution for   must be such that its odd order derivatives vanish on the boundaries. Thus, instead of imposing the restriction (9.40) on    and  to determine the third constant  introduced by Banerjee and Bhowmick in their mathematical analysis, they should have determined it by satisfying  from a solution for  that takes care of the foregoing remarks. This will obviously alter the values of constants and the solution for  as obtained by them without affecting the nature of their mathematical analysis and yield a correct solution of the problem. A correct solution of the problem in view of the above observation is presented below.

 

Following Banerjee and Bhowmick (1992), the proper solutions for  and  must be odd while that for  must be even. Therefore, if and are constants then the function                                                                                           

 is even and since it is required to vanish at  , we can expand it in a Fourier cosine series in the following form

                                                           (9.41)

With  given by equation (9.41), it follows from equation (9.3) that

                               (9.42)

where

                                                                                                          (9.43)

and  is a constant of integration.

Using the boundary conditions on  as given by equation (9.4), it follows from the resulting four equations that

                                                                                                                                                                                 (9.44)

                                                                                                                   (9.45)

                                                                                                                   (9.46)

where

                                                                                                                                                      (9.47)

Further, with and  as given by equations (9.41) - (9.42), it follows from equation (9.1) that

                                                                                                                                                           (9.48)

Therefore, it follows from equation (9.2) that

                                              (9.49)

where

                                                                                                                              (9.50)

Now, since , it follows that

                                                                                                                     (9.51)

Equations (9.45), (9.46) and (9.51) lead to the following unique values of and

                               (9.52)

                               (9.53)

                                (9.54)

where

                                                                                                                                                      (9.55)

                                                                                                                                                     (9.56)

                                                                                                                                                     (9.57)

Substituting for  from equation (9.42 (note that  ) in equation (9.5), multiplying the resulting equation by , integrating over range of z and following the similar steps as in the above analysis of Banerjee et al. and performing some insipid computations, we get the expression

                                                                                                                                                                                 (9.58)

where  is given by the (9.28).

We now complete the solution of the problem by demonstrating that  and as given by equations (9.17), (9.19) and (9.49), which by virtue of the nature of their determination satisfy equations (9.2) and (9.3), also satisfy equation (9.1).To prove this, equation (9.5) in an alternative form can be written as

                                                                                                                (9.59)

where

            

and

, .

The expressions for  and  as given by equations (9.17), (9.19) and (9.49) clearly show that

.                                                                                                                                     (9.60)

Note that  cannot be zero in Banerjee et al. (1989a) and therefore the proof of their assertion that  and  as obtained by them satisfy equation (1) is not tenable.

Multiply equation (9.60) by E* (the complex conjugate of E) throughout and integrating the resulting equation over the range of z by parts an appropriate numbers of times, using (9.60), equating the imaginary part of the resulting  equation to zero and canceling , we get

                                                                                                                                            (9.61)

Equation (9.61) clearly implies that

This proves our contention, namely  and  as given by equations (9.17), (9.19) and (9.49) also satisfy equation (9.1) and complete our construction of a correct exact solution of the problem by appropriately rectifying the lacuna in the earlier analysis of Banerjee and Bhowmick (1992).

 

10.    Some illustrative examples

Example 1

Consider a double eigenvalue problem for p described by the following system of differential equations and boundary conditions (thermohaline instability problem of the Veronis (1965) type)

                ,                                                                                     (10.1)

                ,                                                                                                                                 (10.2)

                ,                                                                                                                              (10.3)

and

either      ,

or            ,

or             ,                                                                                     (10.4)

where  is a constant representing the thermohaline Rayleigh number,  is a constant representing the Lewis number and  is the concentration.

We prove the following results:

(a)           For the case of dynamically free boundaries, if  and , then a proper solution for w belonging to the lowest mode implies that

                .

Proof. If  and , then

                ,

and equations (10.1) – (10.3) become

                ,                                                                                   (10.5)

                ,                                                                                                                                                (10.6)

                .                                                                                                                            (10.7)

Equations (10.5) – (10.7) for the case of dynamically free boundaries imply that

                ,                                                                                               (10.8)

where .

Therefore a proper solution for w belonging to the lowest mode is given by

                ,                                                                                                                                                  (10.9)

where A is a constant.

Operating on equation (10.5) by  and using equations (10.6) and (10.7), we get

               

                .                                                                        (10.10)

Substituting for w from equation (10.9) in equation (10.10), we get

               

                .                                                                                       (10.11)

Equation (10.11) can be written in an alternative form as

               

                               

                                                .                                                                 (10.12)

Equating the real and imaginary parts of equation (10.12), we get

                ,                                                          (10.13)

and         .                                                                     (10.14)

Substituting the value of  from equation (10.14) in equation (10.13), we get

.                                                         (10.15)

Using the expression for R given by equation (10.15) in equation (10.14), we get

.                                                                                                   (10.16)

Equation (10.16) clearly implies that

                .

This completes the proof of the result.

Results (a) implies that for the case of dynamically free boundaries a necessary condition for the validity of overstability is that  and hence a sufficient condition for the validity of the ‘exchange principle’ is that .

(b)           if and , then

                 (Gupta et al. 1986).

Proof. Multiplying both sides of equation (10.1) by  and integrating over the vertical range of z, we get

                .                                        (10.17)

Substituting for  appropriately from equations (10.2) and (10.3) on the right hand side of equation (10.17), we get

               

.                                                                              (10.18)

Integrating equation (10.18) by parts a suitable number of times and making use of boundary conditions (10.4) and the equality

                ,

,                                                                                                                (10.19)

we may rewrite equation (10.18) in the form

               

.                                                          (10.20)

Equating the imaginary part of both sides of equation (10.20) and cancelling , we get

                .                                                          (10.21)

Multiplying equation (10.3) by its complex conjugate, integrating the resulting equation over the vertical range of z by parts a suitable number of times and using equation (10.19), we get

                                                                .                                                               (10.22)

Since , therefore it follows from equation (10.22) that

                .                                                                                                                       (10.23)

Making use of boundary conditions on w and  given by equation (10.4), we have

,                                                                                                                       (10.24)

and         .                                                                                                                       (10.25)

Combining inequalities (10.23) – (10.25), we get

.                                                                                                            (10.26)

Using inequality (10.26) in equation (10.21), we get

.                                                         (10.27)

Inequality (10.27) clearly implies that

                .

This completes the proof of the result.

Result (b) in particular implies that a necessary condition for the validity of overstability is that  and hence a sufficient condition for the validity of the ‘exchange principle’ is that .  Further, this result is uniformly valid for all combinations of boundaries which may be rigid or dynamically free.

(c)           For the case of dynamically free boundaries, if the ‘exchange principle’ is valid, then a proper solution for w belonging to the lowest mode implies that the critical Rayleigh number  and the critical wave number  are given by

.              (Veronis 1965)

Proof. If the ‘exchange principle’ is valid, then at the marginal state

                ,

and equations (10.1) – (10.3) become

,                                                                                                                  (10.28)

                ,                                                                                                                                         (10.29)

                .                                                                                                                                        (10.30)

Equations (10.28) – (10.30) for the case of dynamically free boundaries imply that

                ,                                                                                                (10.31)

where .

Therefore a proper solution for w belonging to the lowest mode is given by

                ,                                                                                                                                                   (10.32)

where A is a constant.

Operating on both sides of equation (10.28) by the operator  and using equations (10.29) and (10.30), we get

                        .                                                                                               (10.33)

Substituting for w from equation (10.32) in equation (10.33), we get

.                                                                                                                                    (10.34)

It follows from equation (10.34) that

                ,

and         .

Hence,

                .

This completes the proof of the result.

(d)           For the case of dynamically free boundaries, if overstability is valid, then a proper solution for w belonging to the lowest mode implies that the critical Rayleigh number , the critical wave number  and the square of the critical frequency  are given by

                ,                                                              (10.35)

                ,                                                                                                                                                    (10.36)

and         .                                                                                                                         (10.37)  

 

Proof. If overstability is valid, then at the marginal state

               

so that equations (10.15) and (10.16) hold good.

                ,

and         .

Hence,    ,

                ,

and         .

This completes the proof of the result.

 

 

 

Example 2

The hydrodynamic instability that manifests under appropriate conditions in a static horizontal initially homogeneous viscous and Boussinesq liquid layer of infinite horizontal extension and finite vertical depth which is kept under the action of a uniform vertical adverse temperature gradient in the force field of gravity is known as simple Bénard in stability while the problem of a theoretical investigation of the essential aspects of this instability is known as simple Bénard instability problem.  One of the fundamental assumption that constitutes the theoretical framework in which this problem is analyzed is that the liquid under consideration is initially homogeneous which, in general, is not true in any real physical situation although it may happen that the extent of this non-homogeneity is small.  But then the Boussinesq approximation which is made use of in the theory with respect to small extent of non-homogeneity introduced on account of thermal effects should have been applied with respect to the small extent of initial non-homogeneity as well and thus there is a need of an extended theory for the simple Bénard instability problem wherein the liquid is initially non-homogeneous.

 

The hydrodynamic instability that manifests under appropriate conditions in a static horizontal initially continuously vertically non-homogeneous viscous and incompressible liquid layer of infinite horizontal extension and finite vertical depth in the force field of gravity is known as Rayleigh-Taylor instability while the problem of a theoretical investigation of the essential aspects of this instability is known as Rayleigh-Taylor instability problem.  One of the fundamental assumption that constitutes the theoretical framework in which this problem is analyzed is that the upper and the lower horizontal boundaries of the liquid under consideration are at the same temperatures which, in general, is not true in any real physical situation although it may happen that the extent of this temperature difference is small.  But then the continuous vertical non-homogeneity that is introduced on account of this thermal effect should also be taken into account in the theory along with the initial continuous vertical non-homogeneity and thus there is a need of an extended theory for the Rayleigh-Taylor instability problem wherein the liquid is initially continuously vertically non-homogeneous.

 

The hydrodynamic instability that manifests under appropriate conditions in a static horizontal initially homogeneous viscous and Boussinesq liquid layer of infinite horizontal extension and finite vertical depth which is kept under the simultaneous action of a uniform vertical temperature gradient and a gravitationally opposite uniform vertical concentration gradient in the force field of gravity is known as thermohaline instability while the problem of a theoretical investigation of the essential aspects of this instability is known as thermohaline instability problem.  However, there is one particular case of this problem which is important from the point of view of its applicability to real physical situations (such as the oceans) and that is the smallness of the coefficient of mass diffusivity of the liquid under consideration compared to its coefficient of heat diffusivity which makes the resulting Lewis number small compared to the thermal Prandtl number.  Thus, there is a need for working out the consequences of a thermohaline instability problem that is simplified on the basis of the above mentioned hypothesis.

 

These ideas were put forward for the first time in 1971 and 1972 by Banerjee who gave a unified mathematical treatment of the three well known problems of hydrodynamic instability namely the simple Bénard instability problem, the Rayleigh-Taylor instability problem and the thermohaline instability problem by showing that if these problems are modified on the basis of the ideas as given above they give rise to the same problem which non is known in the literature as the generalized Bénard instability problem.  Some of the consequences derived by Banerjee from this unified problem later found relevance in the context of more general problems both from physical as well as mathematical points of view (Banerjee et al 1978, 1981; Gupta et al 1985) and in the following we present a brief mathematical investigation of the generalized Bénard instability problem.

 

Consider a double eigenvalue problem for p described by the following system of differential equations and boundary conditions (generalized Bénard instability problem):

                ,                                                                              (10.38)

                ,                                                                                                                                 (10.39)

and

either      ,

or            ,

or           

                ,                                                                                                                  (10.40)

where , and  is a constant representing the initial non-homogeneity Rayleigh number.

We prove the following results:

(a)           If  then .

 

Proof.  Multiplying equation (10.38) by , utilizing equation (10.39), integrating the resulting equation over the vertical range of z by parts a suitable number of times and making use of boundary conditions given by equation (10.40), we get

                                . (10.41)

Equating the real parts of both sides of equation (10.41), we get

      .                                               (10.42)

If , then equation (10.42) gives

.                                           (10.43)

Equation (10.43) clearly implies that

                .

This completes the proof of the result.

Result (a) implies that for the generalized Bénard instability problem the ‘exchange principle’ is not valid.  Further, this result is uniformly valid for all combinations of boundaries which may be rigid or dynamically free.

(b)           If , then .

 

Proof.  Since , therefore dividing equation (10.41) throughout by p, equating the imaginary parts of both sides of the resulting equation and cancelling , we have

                .                                            (10.44)

Equation (10.44) can be written as

.                                                      (10.45)

Equation (10.44) clearly implies that

                , i.e.,

                .

This completes the proof of the result.

Result (b) implies that the complex growth rate of an arbitrary oscillatory perturbation whether stable, neutral or unstable for the generalized Bénard instability problem must lie inside the circle given by

                .                                                                                                                                                       (10.46)

Further, this result is uniformly valid for all combinations of boundaries which may be rigid or dynamically free.

(c)           For the case of dynamically free boundaries, a proper solution for w belonging to the lowest mode implies that the critical Rayleigh number , the critical wave number  and the square of the critical frequency  are given by

                ,

               

and         .

Proof. Follows by proceeding exactly as in result (d) of Example 1.  We note that the critical frequency as given in result (c) above lies inside the circle given by equation (10.46).  Further, the result contained herein also follow from the corresponding results in result (d) of Example 1 by putting             .

 

ACKNOWLEDGEMENT:

This research paper is dedicated to Rev. Fr. Goreux and contains the excerpts of Key Note Address ‘Rev. Fr. Goreux Memorial Lecture’, by Prof. Mihir B. Banerjee on UGC sponsored, Two-day National Seminar held at Sidharth Govt. College Nadaun, Dist. Hamirpur (HP) on March 12-13, 2015.

 

REFERENCES:


1.        Alfvén, H. (1942), Existence of electromagnetic-hydromagnetic waves, Nature, 150, 405.

2.        Alfvén, H. (1950), Cosmical electrodynamics, International Series of Monographs on Physics, Oxford, England.

3.        Banerjee, M.B. (1971), Thermal instability of non homogeneous fluids: a mathematical theory-I, Ind. J. Pure and Appl. Math., 2, 257.

4.        Banerjee, M.B. (1972a), Thermal instability of non homogeneous fluids: a mathematical theory-II, Ind. J. Pure and Appl. Math., 3, 226.

5.        Banerjee, M.B. and Bhowmick, S. K. (1992), Salvaging the Thompson-Chandrasekhar Criterion:  A tribute to S. Chandrasekhar, J. Math. Anal. Appl., 167, 57-65.

6.        Banerjee, M.B., and Gupta, J.R.  (1985b), A sufficient condition for the validity of Chandrasekhar’s conjecture in magnetoconvection, Unpublished work.

7.        Banerjee, M.B., and Jain R. K. (1972b), A unified instability criterion of heterogeneous shear flows, Math. Student, XL, 111.

8.        Banerjee, M.B., Gupta, and Banerjee, B. (1984), A characterization of non oscillatory of motions in magnetohydrodynamics, Ind. J. Pure and Appl. Math., 15, 377.

9.        Banerjee, M.B., Gupta, J.R., and Katyal, S. P. (1989), A characterization theorem for magnetothermohaline convection,J. Math. Anal. and Applns., 114, 141.

10.     Banerjee, M.B., Gupta, J.R., Shandil, R.G., Sood, S. K., and Banerjee, B. (1988a), On the mathematical incorrectness of the derivation of Thompson-Chandrasekhar Criterion in Terrestrial Physics,  J.Math. and Phys. Sci., 22, 547.

11.     Banerjee, M.B., Gupta, J.R., Shandil, R.G., Sood, S. K., Banerjee, B., andBanerjee, K. (1985a),  On the principle of exchange of stability in the magnetohydrodynamic simple Bénard problem, J. Math. Anal. and Applns., 108, 216.

12.     Banerjee, M.B., Katoch, D. C., Dube, G. S., and Banerjee, K. (1981), Bounds for growth rate of a perturbation in thermohaline convection, Proc. Roy. Soc. London, A 387, 301.

13.     Banerjee, M.B., Roy, R., Gupta, J. R., and Banerjee, A. (1978a), A generalized biharmonic equation and its applications to hydrodynamic stability, J. Math. and Phys. Sci., 12, 19.

14.     Banerjee, M.B., Shandil, R. G., and Gupta, J.R. (1978b), On further reducing Howard’s semi-circle, J. Math. and Phys. Sci., 12, 1.

15.     Bénard, H. (1900), Les tourbillons cellulaires dans une nappe liquide, Revue générale des Sciences pures et appliqués, 11, 1261, 1309.

16.     Bhattacharjee, J.K. (1987), Convection and chaos in fluids, World Scientific Publishing Co. Singapore, 9128.

17.     Block, M. J. (1956), Surface tension as the cause of Bénard cells and surface deformation in a liquid film, Nature, 178, 650.

18.     Boussinesq, J.E. (1903), Theorieanalytique de la chaleur, 2, 172, Gauthier Villars, Paris.

19.     Bullard, E. C. (1949), The magnetic field within the earth, Proc. Roy. Soc., A 197, 433.

20.     Chandrasekhar, S. (1952), On the inhibition of convection by a magnetic field, Phil. Mag., 43, 501.

21.     Chandrasekhar, S. (1961), Hydrodynamic and Hydromagnetic Stability, Oxford: Clarendon Press.

22.     Dhiman, J.S. and Sharma, P. K. (2013), On the Exact Solution for the Magnetohydrodynamic Thermal Stability Problem, Journal of the Calcutta Mathematical Society, 9(2), 101-110.

23.     Dhiman, J.S. (1995), Some Characterization Theorems in Rotatory Magneto Thermohaline Convection, Proc. Indian Acad. Sci. (Math. Sci.), 105(4), 461.

24.     Dhiman, J.S., and Kumar V. (2012), On -Law in Magnetoconvection Problem for General Nature of Boundaries Using Galerkin Method, Research J. Engineering and Technology, 3(2), 186.

25.     Dhiman, J.S., Sharma, P. and Singh G. (2014), On the validation of Chandrasekhar’s -Law in Magnetoconvection by Variational Method, Recent Trends in Algebra and Mechanics (Ed. book), Indo American Books, New Delhi, pp.

26.     Eddington, A. S. (1926), The internal constitution of the stars, Cambridge, England.

27.     Elsasser, W. M. (1946), Induction effects in terrestrial magnetism, Physical Rev., 69, 106.

28.     Fermi, E. (1949), On the origin of cosmic radiation, Phys. Rev., 75, 1169.

29.     Finlayson, B.A. (1972), The Method of Weighted Residuals and Variational Principles; Academic Press, New York.

30.     Gibson, R.D. (1966), Overstability in magnetohydrodynamic Bénard problem at large Hartmann numbers, Proc. Camb. Philos. Soc., 62,257.

31.     Gupta, J. R.,    Sood, S.K Shandil, R. G. and Banerjee, M. B.  (1985b), A generalized biharmonic equation and its applications to hydrodynamic stability, J. Math. Anal. and Applns., 110, 31.

32.     Gupta, J. R.,   Sood, S.K., Shandil, R. G. and Banerjee, M. B.  (1985b), A generalized biharmonic equation and its applications to hydrodynamic stability, J. Math. Anal. and Applns., 110, 31.

33.     Gupta, J. R., and Kaushal, M. B. (1988a), Rotatory hydromagnetic double-diffusive convection with viscosity variations, J. Math. and Phys. Sci., 22, 301.

34.     Gupta, J. R., Sood, S. K. and Bhardwaj, U. D. (1984b), A note on semicircle theorems in thermohaline convection with rotation and magnetic field, Ind. J. Pure and Appl. Math., 15, 203.

35.     Gupta, J. R., Sood, S. K ., and Bhardwaj, U.D . (1985a), Double-Diffusive convection in completely confined fluids, J. Math. and Phys. Sci., 19, 283.

36.     Gupta, J. R., Sood, S.K., and Bhardwaj, U.D. (1986), On the characterization of non-oscillatory motions in rotatory hydromagneticthermohaline convection, Ind. J. Pure and Appl. Math., 17, 100.

37.     Jeffreys, H. (1926), The Stability of a layer of fluid heated below, Phil. Mag., 2, 833.

38.     Kumar, K., Bhattacharjee, J. K. and Banerjee, K. (1986), Overstability in magnetohydrodynamic convection revisited, Phys. Fluids, 29, 4032.

39.     Lin, C. C. (1955), The theory of hydrodynamic stability, Cambridge University Press.

40.     Normand, C., Pomeau, Y. and Velarde, M. (1977), Convective instability: A physicists approach, Rev. Mod. Phys., 49, 581.

41.     Pearson, J. R. A. (1958), On convection cells induced by surface tension, J. Fluid Mech., 4,489.

42.     Pellew, A and Southwell, R.V. (1940), On the maintained convective motion in a fluid heated from below, Proc. Roy. Soc. London, A176, 312.

43.     Rayleigh, L. (1916), On the convective current in a horizontal layer of fluid when the higher temperature is on the upper side, Phil. Mag., 32, 529.

44.     Ritchmeyer, R. D., and Teller, E. (1949), Phys. Rev., 75, 1729

45.     Schultz, M. H. (1973), Spline Analysis, Prentice-Hall, Englewood Cliffs, NJ.

46.     Sherman, M. and Ostrach, S. (1966), On the principle of exchange of stabilities for the magnetohydrodynamic and thermal stability problem in completely confined fluid, J. Fluid Mech. , 23, 661.

47.     Thomson W. B. (1951), Thermal convection in a magnetic field, Phil. Mag. Ser., 7, 42, 1417.

48.     Veronis, G.  (1965), On finite amplitude instability in thermohaline convection, J. Mar. Res., 23, 1.

49.     Walen, C. (1949), On the vibratory rotation of the sun, Henrik Lindstahls Bokhandel, Stockholm.

 

 

 

 

Received on 18.04.2015       Modified on 30.04.2015

Accepted on 28.05.2015      ©A&V Publications All right reserved

Research J.  Science and Tech. 7(2): April-June. 2015; Page 55-103

DOI: 10.5958/2349-2988.2015.00009.1