1. Introduction
Magnetohydrodynamics (MHD) is a combined study of hydrodynamic flows and electromagnetic fluid flow through coupling forces and their interactions. These equations consist of highly coupling equations with well-known Maxwell equations and the hydrodynamic equations for fluid flows. Alfven first proposed the field of MHD for single- and multi-fluid flows [
1]. More recently, this field has become important because of its utilization and practical applications in geophysics, astrophysics, and many other engineering fields, like cooling metal, MHD propulsion [
2], MHD pumps [
3], process metallurgy [
4], controlled thermonuclear fusion and seawater propulsion [
5], electromagnetic casting of metals, MHD power generation, MHD ion propulsion [
6,
7] etc. Moreover, the hydrodynamical behavior of conducting fluids, e.g., electrolytes, liquid metal cooling in nuclear reactors [
8], and plasmas are usually formulated via MHD models [
9]. Furthermore, the theoretical and numerical investigations can be further seen in the following articles and are the given references [
10,
11,
12].
Because these equations are not easily solvable for the analytical solution, the alternative way to solve the complex model equations are the numerical solutions where inf-sup or Ladyzhenskaya–Babuska–Brezzi (LBB) conditions need to satisfy for the well-posedness of the models. The complications caused by the well-known inf-sup condition have prompted the introduction of various stabilization techniques intended to circumvent these conditions [
13], a Galerkin method, Galerkin least square method, penalty method,
-inner products, different stabilization techniques and many modified methods are abundantly used to find the numerical solutions [
14,
15,
16,
17], and the well-posedness of the stationary equations with the inf-sup conditions have been given in [
18]. MHD-modeled equations are highly non-linear coupling equations which must have to accept the complicated structure of the solution. Therefore, it does require an attractive, efficient numerical scheme which plays a key role to find the solution of such types of complicated structured models. In MFE literature, the least square methods are utilized for such types of complex and non-linear models to find the approximate solution of higher-order PDEs by defining some norms of functional spaces as Hilbert-type to solve [
19,
20,
21,
22,
23], singular solutions of flows related to viscoelastic behaviours [
24], heat transfer and flows [
25], and fluids for which temperature-dependecy or viscosity-dependency are critical [
17]. Standard least square methods are efficient, effective, and smart enough to give approximate solutions of the non-linear model equations in the linear algebraic form, which would be symmetric and positive-definite. During the last decade, many methods and theories of least square methods have been discussed in the literature. For further details on such applications, we refer interested readers to [
26,
27,
28] among others.
Similarly, a quasi-least-square scheme (QLSFES) is developed [
23] to solve the problems in the
-inner product. Typically, stabilization relies on some form of modification of the discrete continuity equation. For example, in [
29] the continuity equation is modified to penalize the stability issue with the addition of a single parameter to make the bilinear form coercive for some specific finite elements. The central point of the scheme is to study the problems in the
-norm to the Oseen type (linearized) forms of given coupled nonlinear problems. It has several benefits; first, only
-inner product with norms is used in these methods which are convenient for the computer programming sense. Second, by using the linearizing way (Oseen type), someone can get iterative methods with symmetric positive-definite coefficient metrics. However, this simplified iterative method is easily convergent in a complex domain of initial guess [
23,
30,
31]. For the MHD-based models, common theories and algorithms for approximation of the solutions of such nonlinear problems are not always applicable because of the functional spaces, i.e., Hilbert space for the velocity and pressure. These are not convenient to utilize the Hilbert space, which will ultimately create some stability issue in the analysis. The QLSFES can be easily applied to circumvent this deficiency of the main conditions (inf-sup or LBB). Thus, the prime intention here is to develop a scheme for a coupled branch of non-linear problems to analyze the existence and convergence of model equations. The least-square MFE schemes are inadequate to address the local convergent properties of coupled non-linear problems. However, this method can be utilized as an effective and sufficient way for the well-posedness of the incompressible MHD models without the inf-sup conditons.
In this contribution, the practicality principle can always be met by transforming the given model equations into a transformed first-order system and forming least-squares functionals that use only -norms. We focus attention on quasi-least-squares methods for which the discretization step is invoked after the quasi-least-squares functional has been defined. The key point to utilize this setting is that it allows one to point out the variational interpretation of least-squares principles as projections in a Hilbert space with respect to problem-dependent inner products. From this point of view, the principal task in the formulation of the method becomes setting up a least-squares functional that is norm-equivalent (-norms) in some Hilbert space. This in turn allows one to work in the variational setting as continuity and coercivity (well-posedness) without the LBB or inf-sup conditions. A quasi-least-squares (QLS) mixed finite element method (MFE) based on the -inner product is easily convinced to solve a coupled and nonlinear incompressible magnetohydrodynamic (MHD) model. Here, first, these nonlinear models are made linear around a particular state wherein linearized first-order equations represent an algebraic system of equations with symmetric matrices. Secondly, a direct iteration technique is applied to couple up the nonlinearities and obtain a theoretical convergent rate for general initial guess. As far as we know from the existing literature, this method has never been considered to find the solution of the MHD model equations without the LBB/or inf-sup conditions.
From a theoretical viewpoint, such a bona-fide (continuous) QLS finite element methods possess a number of significant and valuable properties, as follows.
The weak problems are, in general, coercive.
Conforming discretizations lead to stable and, ultimately, optimally accurate methods.
The resulting algebraic problems are symmetric and positive definite.
Essential boundary conditions may be imposed in a weak sense.
Finite element spaces of equal interpolation order, defined with respect to the same triangulation, can be used for all unknowns.
Algebraic problems can be solved by using standard and robust iterative methods, such as conjugate gradient methods.
Methods can be implemented without any matrix assemblies, even at the element level.
Only a single parameter with a single initial guess is sufficient to establish the well-posedness of the solution. In the existing literature this is the first time we apply for this model.
The rest of the work is arranged as follows. In
Section 2, we demonstrate linear and non-linear incompressible stationary MHD equations with the proposed scheme (QLSFES). In
Section 3, we investigate the well-posedness of the developed QLSFES system for investigation of the initial guess. In
Section 4, we discuss convergence of the proposed scheme in the case of the nonsingular exact solution. In
Section 5, the theoretical is proven and in conclusion the contribution is illustrated to end this work.
2. Model Introduction
This work illustrates the numerical resolution of the stationary-coupled magnetohydrodynamics system of equations [
32,
33]. The unknowns for this problem are supposed symbolically as velocity field
, the magnetic field
, and fluid pressure
p under the connected two-dimensional domain
.
The non-dimensional stationary MHD equations are as follows:
where
signifies the Reynolds number for hydrodynamic,
the Reynolds number for magnetism, and
, with
, coupling number respectively. In the literature of the industrial sense, we know that, the parameters are considered always
,
and
. To acquire the values of the known velocity field
and pressure field
p and magnetic field
are given in a bounded domain
. We consider the load function
represents the external forces or inertial terms. Additionally, for the 2D form, the operator can be defined as
, while the product of given vectors
and
is defined as
.
The system of nonlinear stationary MHD can be solved under the set of boundary conditions within the connected domain
on boundary
is given from the references [
33,
34,
35,
36]:
where
is a normal vector to the domain
[
17]. Here Equation (
5) is representing the viscous nature fluids and is well-known as a no-slip boundary condition, whereas the Equations (6) and (7) are nominated for the perfect conduction wall.
Remark 1. Alternately, some other boundary constraints are utilized in the literature [12,36,37,38,39] are in Equation (6) and in Equation (7) for the unknown . We utilize and , which is given for the nonlinear MHD for a single-fluid flow. 4. QLSFES Convergence and Existence of the Solutions
In this section, we obtain convergence and existence of the solution of proposed scheme with the illustration of four steps.
We demonstrate property of the bilinear function. For any parameter and are introduced in such a way that there exists a bounded function set that is well defined and continuous as and holds in this bounded domain.
In step two, we find out a large domain of bounded functions which holds all solutions of the System (8)–(13). We intend to solve the nonlinear coupled equations by the right choice of and ,and is well-defined as and are in this large domain of functions.
To show the well-posedness of the proposed scheme QLSFES, we establish the nonlinear plan of the scheme in such a way that the solutions under some nominated domain are fixed points. In step one and step two, we summarize that for a specific value of
and
, the system of the nonlinear model is distinctively executable in this specific bounded set (see detail in Lemmas 3 and 4). Moreover, in Theorem 1, by using the fixed point theory [
23], we illustrate briefly the existence of solutions of the scheme.
In the fourth step, Theorem 2 is given to illustrate the convergence and existence of the proposed scheme QLSFES.
Before proceeding to the actual contribution, we intend to understand several existing results which are utilized in the immanent sections. By the theory of embedding and the Poincaré’s inequality, the positive constants e, , and always depend on the fact that the domain can be recalled as
According to Green’s formula and integration by parts [
42], we can state
The bilinear form
is well known as an inner product in space
or
. It is always known to be a fact that
holds the boundary condition
if
such that
Lemma 1. Let constants and . There exists two constants β and such that for each satisfying and , and for each ,where the constants and are defined as followsand . Proof. From Young’s inequality and the given conditions, we can demonstrate
and
Let us consider
holds (
23), and we obtain the following estimation
By utilizing Equation (
23), we can state
in a similar way,
By substituting Equations (
30) and (
32) in Equation (26), we have
Considering the constants
we can state
It is further estimated that
is:
From the Equations (
32) and (
34), we understand that the conclusion of the condition (
24)(
i) holds for the
defined by (
24) and (
25), whereas by (
24)(
), (
28), (
30) and (
35) also satisfies. □
Remark 3. Lemma 1 deals three positive constants , η, and γ where and η are nominated constraints for the gradient and divergence of function illustrated below. We would be able to indicate that is playing a key role to control the side terms in right-hand. By employing the constant , we may figure out the bounded function set which restrains all solutions of the given MHD model equations. Otherwise, it could not be easy to handle the divergence-free conditions in the theoretical formulation. Here, we manage the (divergence-free) and . The Λ is a coefficient which is actually free from the constat and γ but it depends strongly on η and π. It is worth noting that Λ becomes smaller if η is larger. Hence, Lemma 1 illustrates the way by which the parameter Λ can be formulated. This will be utilized to execute γ in practical usage. The value of γ is a penalty factor which is used to control the boundary condition for the velocity field divergence. It is important to note that the parameter has acting no such type of usage in the mathematical formulation at all.
Remark 4. Because a magnetic field is a solenoidal field, it may not be considered as compressible or incompressible [11]. To penalize the magnetic effect as divergence or of the field there is no coefficient so far discussed in the literature. It means the well-psedness of the MHD is still open for the researchers. This will be a challenging problem for our future work. In active usage, the well-defined bilinear form
for all
is not compulsory and a similar remedy is considered for the
is applied. However, one can find the approximate numerical values of the unknown functions in the particular domain, which have all the exact solutions. We are suppose to seek this bounded function set. To this end, some functions are supposed in
as:
The given lemma illustrates that for some positive parameters
, Equations (8)–(13) holds solutions in the bounded set
.
Lemma 2. Consider . Suppose that Λ and holds ((24)i,ii) and satisfies All the possible solutions of (8)–(13) are in the function set .
Proof. Let
be the only one solution of the given system Equations (8)–(13). For all
, it can be seen that
This states:
We know the term for the incompressible conditon
and
, so
including
Therefore,
where
. Accordingly, it can be easily seen that
. Because
By taking
and by using Lemma 1, one can see that
is in
. □
We acquire the approximate solution of the System (8)–(13). Let us consider the nonlinear map based upon the defined spaces
into
as
such that for each
∈
,
Now, it is certain that the System (
44) is linear with respect to
. Now the non-linear MHD can be estimated through the following results given below.
Lemma 3. Suppose the Lemmas 1 and 2 satisfy. Then the bilinear form of the system from to is distinctly defined.
The key point is that the Lemma 3 is the straightforward result of Lemma 1.
Lemma 4. Let us consider that the results of Lemmas 1 and 2 satisfies all the conditions for γ and it satisfies Now the bilinear operator shows to itself.
Proof. Hence from Equation (
44)
because we have
and
so Equation (
24) tends to
Furthermore,
Therefore,
. This proves completed. □
Theorem 1. Let , K holds (38), γ holds (45), Λ and holds ((24)i,ii). Therefore the scheme QLSFES satisfies and holds a minimum one solution in bilinear form . However, all the illustrated unknown solutions of the nonlinear model Equation (16) are in . Proof. Lemma 4 states that, the operator
relates the bounded domain
into itself under the specific conditions of Theorem 1. It satisfies the fixed point Browners theory that the nonlinear System (
16) has minimum one solution in
. Indeed it is true that all the solutions of the unknowns of model nonlinear Equation (
16) are in
. Hence the proof of Theorem 1 is completed. □
Moreover, the solutions of QLSFES convergence is demonstrated below as:
Theorem 2. Suppose that Theorem 1 holds and is a unique sequence of the solutions of the proposed scheme as . Then the sequence of solution can be further subdivided into many subsequences which will converge weakly to the different solutions of the first-order MHD (8)–(13). Particularly, components of the MHD are weakly convergent into the subsequences and are strongly convergent to the similar limit components in for specific condition .
Before proceeding to the prove of the Theorem 1 we entail some lemmas for the understanding, which are given as
Lemma 5. We suppose the inequality . For a specific Oseen-type function holds the relation and a particular load function . At present the problem with boundary values can be stated aswhich holds a unique solution in . The reader can see [
23] appendix 1. Hence, the embedding theory between Sobolev spaces and some results reported in [
23,
40] are directly utilized here as in the following lemmas given below.
Lemma 6. Assume that is a Hilbert space in which F might be a bounded function set within , i.e., there is a domain-dependent constant in such a way that for every . If T is a function set which is weakly compact in the given Hilbert space [23]:Here denotes the inner product of two functions in space . Lemma 7. Let T be a set of bounded functions in Hilbert space , such that there exists such that for each . Then the T set is compact strongly in space for every interval , i.e., the , which is strongly convergent in the Hilbert space as , which may be deduced from T.
Lemma 8. Suppose a positive constant C exists and holds for interval and every Hence Theorem 2 might be resumed as:
Proof. It comes from Theorem 1 that the solutions of QLSFES
are bounded [
23]. Lemmas 6 and 7 concludes that the results are divided into many other subsequences which are weakly convergent in the
. Moreover, the weak convergence of the subsequence of the proposed scheme, might still represent it by
and its weak limitation by
in [
]. However, from Lemma 7, someone may know that components
are strongly convergent to
in
for condition
.
We are supposed to justify that
is the first-order linear system Solution (
8)–(13). As a result,
is a unique solution of the MHD. To this end, tentative functions
are introduced as:
and similarly for
Right now, Systems (
49a) and (
50a) represent two different linear systems, which are independent Lemma 5 concludes that for the
these two free system of equations are distinctly executable. Let
and
. It is true that
is a unique solution for the Equations (8)–(13) if the relation holds as
. We intend to justify this in three steps.
In Step 3, from (
51), (
52a) and (52b) it is satisfied that
Equation (
53a) represents that
because the Equation (
53a) holds a unique but trivial solution. Similarly, one can prove
. □
The proof of (
51) can be written and it is obvious that (
51) is identical to
For every
, one might have a test functions for the variables velocity and pressure as (
v,
q).
From Lemma 5, we conclude that
. By assuming
, one can have
The weak convergence property of
is the unique solution of the MHD System (
8)–(13) for the nominated positive parameter
. This might be further seen equivalently as
Therefore Lemma 8 and the
concludes that
together with
From Equations (
8)–(13), it can be further estimated as
If we replace Equations (
57)–(
59) with (
56), it yields (
54). Now we may proceed the proof of the second step (
52a) and (52b). With the help of weak convergence solution sequence of the FE space approximations, it can be further seen for each
∈
, and we have
Now, this shows a complete proof of the Theorem 2 which is indeed a solution of QLSFES in general cases. Moreover, close to the approximate solutions of a singular solution, someone can only execute approximate solutions of weak convergence of the subsequence. However, in the next section, the proof of the strong convergence of the MHD holding with the uniform convergence rate in non-singular solutions are briefly illustrated.
5. Convergent Rate of Non-Singular Solution of QLSFES
In this section, the non-singular solutions and convergence of the system Equations (
8)–(13) are demonstrated. A solution of (
8)–(13) is considered as the non-singular solution in some special cases i.e., if this solution does not depend on any other solution and the first-order differential approximation of the system is non-singular at this solution [
23]. We consider a linear system
, which states that if
it holds a unique solution
. For all of the above, the constant e holds the relation
We may assume the Stokes equation for further analysis as
and
Here,
is a regular solution for any
so the solution
,
and
Furthermore, we assume that the MFE spaces
satisfies the approximation properties such that there exists an approximation order
r of the MFE space which is defined as
and C such that
here,
r is the order of approximation for spaces defined.
Theorem 3. Let us suppose the conditions of Theorem 2 holds. Assume that () is the unique approximate solution which weakly converges to () (8)–(13). Suppose that () is a non-singular solution of the given model, and then the sequence of solution () converges strongly in as Thus the estimation results priori satisfies the following relation: Proof. Suppose (
) is a solution sequence for a discrete scheme, which is weakly convergent to a unique continuous solution (non-singular) (
) of its lowest order Model (
8)–(13) and belongs to
as
Furthermore, we can represent
as:
Then, we note here
Equation (
69) gives the inequality as
For the error estimations, we may bound right-hand terms of Equation (
79) as
It follows from (
8)–(13) and (
16) we can get
To bind the right-hand side final terms of Equation (
81), we can define the alternative equation as
Hence this System (82)–(84) possesses one unique solution
in
. Let us consider
and
. Then by auxiliary
we have the following relation:
By Substituting (
75), (
76), and (
85) into (
81), we get
Having the Equations (
80) and (
86) into (
79), and the assignment of
leads to the following relation:
where
Consequently, if
is a relation which converges to 0 as
, Equation (
86) leads us to
For
, Equation (
88) leads to (
78) for interval
. This is a complete proof of the theorem. □
Limitation and Future Work
This scheme can be applied in many problems, such as
an error analysis of uasi-least squares finite element method of velocity-pressure-magnetic field formulation for MHD problem [
22,
43];
quasi-least-square method to solve the MHD with four unknowns, i.e., velocity of the fluid , velocity of the magnetic field, pressure of the fluid and pressure for the magnetic field;
quasi-least square method for the Maxwell equations [
44]; and
quasi-least square method for the second order MHD model equations [
45].