Sei sulla pagina 1di 14

43rd AIAA Plasmadynamics and Lasers Conference 25 - 28 June 2012, New Orleans, Louisiana

AIAA 2012-3297

A Novel Set of Unied Maxwell Equations Describing Both Fluid and Electromagnetic Behavior
Richard J. Thompson and Trevor Moeller
University of Tennessee Space Institute, Tullahoma, TN, 37388, USA
This paper reviews a novel theoretical transformation of the two-uid plasma equations into a set of Maxwells equations, where new unied elds supplant the electric and magnetic elds, and contain both the uid and electromagnetic character of the plasma. The challenge to using this framework is that a knowledge of the unied charge and unied current is presumed, which is a superposition of gasdynamic charge and current and electromagnetic charge and current. While electromagnetic charge and current are a familiar physical quantity, the idea of gasdynamic charge more foreign; therefore, this paper explores two numerical simulations where the unied charge and current is postprocessed and examined; this reveals some preliminary knowledge of the structure of the charge and current in the unied Maxwell equations.

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

Nomenclature
A B E 0
m 0

Electromagnetic vector potential (Subscript) species Magnetic induction eld Electric eld Permittivity of free space Permeability of free space Electric scalar potential Electrical conductivity Unied charge Mass density Unied current

je P u a c0 e/m h

Electric current Canonical momenta, P = U + (e/m)A Fluid velocity Generalized vorticity, = P = + (e/m)B Vorticity Generalized Lamb vector, = ( P) u = u = ( + (e/m)B) u Sonic speed Speed of light Charge-to-mass ratio Enthalpy

I.

Introduction

revious work has revealed a novel theoretical framework wherein the equations of plasma dynamics (specically, P the multiuid model including the full Maxwell equations) is shown to comprise a more general set of Maxwell equations, where the new analogous electric and magnetic elds are composed of both the uid and electromagnetic behavior.1 This framework has been previously recognized and written about for incompressible uid ow2 and compressible ow,3 and has recently been extended to plasmas.1 The major challenge imposed within this framework is a conceptual problem: a generalization of the idea of charge and current must be introduced, since the solution of the Maxwell equations insists on a knowledge of these charges and currents. Ideally, some intuitive, physical understanding of these source terms must be developed in analogy to how a conceptual understanding of electric charge and current was arrived at through nineteenth-century experimental science. The incompressible and compressible Maxwell equations involve discovering a form of uid charge and uid current which drives the vorticity (magnetic eld) and uid Lamb vector (electric eld). A plasma broadens these source terms to plasma charge and plasma current, which are superpositions of the uid charges and currents and the electromagnetic charges and currents. We previously constructed the unied Maxwell set specically for two-uid plasmas.1 In this paper, we introduce the simplied Maxwell equations describing the unied behavior of the hydrodynamic and electrodynamic character of the plasma under assumption of a strongly magnetized ow. This further simplies the form of the unied charge and
Graduate Assistant

Research Assistant, Dept. of Mechanical, Aerospace & Biomedical Engineering, 411 B. H. Goethert Pkwy, AIAA Student Member Professor, Dept. of Mechanical, Aerospace & Biomedical Engineering, 411 B. H. Goethert Pkwy MS24, AIAA Associate Fellow

1 of 14
Copyright 2012 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

American Institute of Aeronautics and Astronautics

unied current. We then present numerical simulations of compressible ow and MHD plasma to reveal the structure of the unied source terms in the context of strongly magnetized plasmas and gasdynamics. It will be seen that the unied charge density comprises a superposition of a few charges that propagate with the eigenvalues of the system, which indicates that simple charge models can be constructed. The data further suggests that a rst approximation to the convective current might fashion a suitable model for the unied current in relation to the unied charge. In future work, we hope to develop a numerical algorithm in which the Maxwell equations are solved with injected forms of these charges and currents. A. Magnetohydrodynamic and Two-uid plasma models

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

The successful understanding, prediction and modeling of an engineering plasma relies on a sound physical model for the coupled uid dynamics and electrodynamics occurring in the plasma, as well as any additional behavior of importance (for example, radiative heat transfer, laser-plasma interactions or ablation physics). The coupling between uids and electrodynamics occurring in plasmas is known to be a very challenging mathematical problem, both analytically and numerically. Often, the physics of the plasma can be simplied by invoking the magnetohydrodynamic (MHD) approximation. In the MHD framework, the electromagnetics are described by a limited subset of the Maxwell equations wherein only diusion behavior is permitted (i.e., no electromagnetic waves are permitted). Furthermore, an ad hoc assumption of the macroscopic current is introduced (namely, that the conduction current dominates the convection current), and usually the magnetic force is assumed to dominate over the electric force, since most of MHD concerns itself with the study of strongly magnetized ows. MHD permits a variety of important waves to result (such as Alfvn and magnetoacoustic modes), and allows for simplied analytical and computational solutions to be determined. However, the MHD approximation is only valid for high-conductivity plasma. This can be seen by examining the magnetic telegrapher equation, which describes the full behavior of the magnetic eld, B 1 2 B + 0 + 2 B = 0 2 t t c2 0 (1)

Here we can see that the diusive behavior of the magnetic eld only dominates if the rst time derivative quantity vastly exceeds the second time derivative quantity. Otherwise, the magnetic eld behavior will be inherently hyperbolic, no matter how small the second time derivative is. MHD fundamentally does away with this diculty by assuming that the limit of the diusion behavior can be reached that is, that the conductivity is large enough to be considered innite. Of course, no true plasma possesses an innite conductivity, but the approximation works rather well for many laboratory and engineering plasmas. There exists a wide variety of plasma that cannot be adequately described using the MHD model, even for engineering cases. The major deciencies of the MHD model are as follows: 1. Although the MHD framework permits the existence of Alfvnic and magnetoacoustic waves, an abundance of other modes are completely excluded; for example, the removal of the displacement current excludes the possibility of electromagnetic waves. An example where this could be important is a plasma thruster operating in vacuum; the vacuum region is a low-conductivity region, so the second time derivative in equation 1 becomes signicant, and wave propagation dominates the behavior of the electromagnetics in the vacuum region. Hence, plasma propulsion thrusters involve transitions from high-conductivity to low-conductivity (vacuum) regions, and the successful modeling of this disparity demands the capability to resolve both the diusion and wave limits of the plasma. Previous computational work has circumvented this problem by injecting eld-carrying uid or by using experimental data in the vacuum region to correct for the wave behavior.4, 5, 6 2. The MHD framework eectively reduces all phenomena to a single time scale. This removes unwanted microscopic behavior specic only to certain species from playing any major role in the development of the overall plasma. However, engineering plasmas exist where this behavior actually may inuence the overall plasma behavior. Mathematically, this becomes a singular perturbation problem7 the MHD theory allows only a limited part of the available phenomena to play a role, and hence cannot fully accommodate cases where the plasma may see two-uid eects develop. 3. MHD assumes a single-uid behavior, which intrinsically limits the approximation to low-frequency plasmas. If higher-frequency phenomena is encountered, the MHD model will not provide sucient delity to the physics of when ions and electrons may react dierently. Seeking a new physical model that resolves the above mentioned deciencies of the MHD model, we turn our attention to the two-uid (or, synonymously, multiuid) model of plasma dynamics. This model involves separate uid
2 of 14 American Institute of Aeronautics and Astronautics

models for each species (usually ions and electrons, although other species are permitted), which behave independently of each other except for common electromagnetic elds, which are described using the full Maxwell equations. This model does not suer from the same above deciencies as the magnetohydrodynamic model; however, the introduction of the speed of light and the required timestep to properly resolve two-uid waves incurs major computational limitations, particularly due to the Lorentz force, which behaves as a source term in the uid dynamic equations. This drives the two-uid equations to become very sti, making it generally very dicult to solve numerically. Hence, we can see that major problems are presented on both sides; the MHD model simplies our treatment of the problem, but sometimes at the expense of the true underlying physics, whereas the two-uid model provides a superior model for certain engineering plasmas, but the computational constraints imposed on solving the equations produces a very sti problem that generally can only be solved with a great deal of diculty, usually owing to the presence of large source terms in the equations. B.
Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

A unied approach to treating uids and electromagnetics

In light of the deciencies mentioned in the MHD model, and the present computational limitations of the two-uid model, it is natural to ask if another path to resolving the behavior of engineering plasmas in a general sense and yet with computational simplicity can be achieved. In this paper, we review a new theoretical perspective of the two-uid equations, in which the equations can be written as a set of Maxwell equations; the electric and magnetic elds are instead transplanted by new, more general eld quantities that describe the evolution of both the electromagnetic and uid dynamic behavior of the plasma. Although the unied Maxwell equations lack closure, it is shown that some limiting cases do result in an isomorphism to classical electrodynamics. The major challenge imposed on this new set of Maxwell equations is pointed out by Jackson:8 Namely, that there exist two limits in which the elds described by the Maxwell equations can be solved exactly, which correspond to one in which the sources of charges and currents are specied and the resulting electromagnetic elds are calculated, and the other in which external electromagnetic elds are specied and the motion of charged particles or currents is calculated... . Occasionally... the two problems are combined. But the treatment is a stepwise one rst the motion of the charged particle in the external eld is determined, neglecting the emission of radiation; then the radiation is calculated from the trajectory as a given source distribution. It is evident that this manner of handling problems in electrodynamics can be of only approximate validity. Hence, the new framework insists that if the unied elds are to be determined, some knowledge of the source terms (the unied charge and unied current) must be possessed by the investigator. The implication of developing a knowledge of the source terms seems daunting. Extensive experimental testing and numerical modeling could be demanded before analysis would yield a robust model that works in all applications. In an eort to expose the nature of the source terms, this paper investigates the unied charge and current for two test cases from compressible ow and MHD. While it remains our ultimate goal to develop models robust enough to model two-uid plasmas, developing some knowledge of the compressible ow and MHD charges and currents is a necessary and more feasible step in revealing the nature of these source terms. Our results will show that the charges for these systems is relatively simple, and could potentially be very feasible to model for a numerical scheme.

II.
A. Background

Theoretical framework of the unied Maxwell equations

Often two physical theories can illustrate a remarkable degree of similarity in their mathematical structure. An example is the wave equations of acoustics and electromagnetism; although these equations describe waves of totally dierent physical character, the equations can be seen to be nearly identical in many cases. Another simple example is the relationship between linear and angular kinematics; analogues may be constructed between each quantity in these systems. We borrow the language of Towne9 to refer to such a similarity as an isomorphism. The utility in constructing an isomorphism between two physical theories is that well-established theorems and techniques of one eld may be correlated to the other. Another important result is that the physics of one theory may be illuminated in a novel and insightful way by describing its analogous conditions in relation to another theory. In this section, we expose the similarity between the two-uid plasma model and the Maxwell equations of classical electrodynamics. This allows us to reformulate the behavior of a plasma in terms of generalized electric and magnetic elds, which unify the behavior of both the electromagnetics and uid dynamics of the plasma. This results in a set of equations remarkably similar to the Maxwell equations that describes the new unied elds, and introduces generalized charge densities and current densities that include contributions from both the electromagnetics and uid
3 of 14 American Institute of Aeronautics and Astronautics

dynamics. In the limiting cases of strongly magnetized plasmas, magnetohydrodynamics, and gasdynamics, these Maxwell equations become completely isomorphic. B. The Maxwell equations for a plasma

Consider an isentropic, ideal plasma. Viscous and entropic terms can be added in later,1 which results in an analogous polarization of the generalized electric eld. Our derivation will largely follow that presented in Ref 1. The continuity and momentum equations describing such a plasma are t
m,

h + u h + a2 u = 0 t e u + u u = h + E + u B t m +
m, u

(2) (3)

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

Isentropic thermodynamic relations have been used in equation 2 to manipulate it in terms of species enthalpy h , species speed of sound squared, a2 , and the divergence of the species velocity, u . Similar isentropic properties have also been introduced in equation 3 to replace the pressure with the enthalpy. A To simplify equation 3, we introduce the vector and scalar electromagnetic potentials, E = t and B = A. Next, we introduce the Lamb vector identity to break up the nonlinearity, u u = u + ((1/2)u u ), and combine the kinetic energy gradient with the enthalpy gradient. This gives us e e e u + A + + B u = (H + ) t m m m (4)

where H = h + (1/2)u u is the stagnation enthalpy. Recognizing that the rst bracketed quantity is the masse specic canonical momenta, P = u + m A, and the second bracketed term is the curl of the canonical momenta, e e called the generalized vorticity, = + m B, and dening the mass-specic total energy = H + m , and nally simplifying our notation by using = u , we have the following compact momentum equation: P + = (5) t It is fruitful to consider the derivatives of equation 5. We consider briey the divergence, curl and time derivative of this Euler equation in order to develop an analogue to classical electrodynamics: P + = 2 t Curl + = 0 t 2 P + Time derivative = t t t 2 It can be shown (for details, see the appendix) that equation 8 can be rewritten as Divergence e 2 + j a2 B) = 0 c ( t m with the vector j dened as j = 2 u 1 1 e 2 ( u u ) + a2 je u u h + t 2 t 2 0 m (10) (6) (7) (8)

(9)

If we replace equation 8 with its modied form 9, and include an equation exposing the divergencelessness of the generalized vorticity, = P = 0, then our set of equations describing the derivatives of equation 5 swells to Div of Divergence Curl Time derivative = 0 P + = 2 t + = 0 t e 2 + j a2 B) = 0 c ( t m
4 of 14 American Institute of Aeronautics and Astronautics

(11) (12) (13) (14)

By simply rearranging equations 11 through 14, we end up with the following set of equations describing the generalized quantities of the plasma: Twouid plasma unied Maxwell equations No monopoles Gauss law Faraday law Ampere law =

= 0

(15) (16) (17) (18)

e 2 B) + j = a 2 + c ( t m

+ = 0 t

P = 2 H t

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

where j is given in equation 10. These equations strongly resemble the Maxwell equations of classical electroe dynamics. Here the generalized vorticity, = + m B, and the generalized Lamb vector, = u = e u + m B u , supplant the usual magnetic and electric elds, respectively, and hence the new generalized elds include both contributions from the uid-dynamical and electrodynamical character of the plasma. Note that in the context of our earlier discussion of an isomorphism, these equations are not rigorously isomorphic to the Maxwell equations (they only resemble them). This is due to a disparity in the propagation speeds of the uids and electromagnetics; the true Maxwell equations expose a single speed of propagation. Therefore, equations 15 through 18 do not constitute a rigorous isomorphism, but only a similarity. However, some limiting cases can be explored that expose complete isomorphisms to the Maxwell equations, which we discuss next. C. Limiting case: Magnetized plasma equations and Magnetohydrodynamic equations

Now that the full multiuid equations have been revealed in equations 15 through 18, some simplications can be achieved for certain cases which result in complete isomorphisms. One of the immediate simplications we can make is by assuming that the electric eld term in the Lorentz force does not signicantly aect the body force acting on the plasma, or, synomously, the plasma is strongly magnetized. In such a case, only the je B term is retained in the Lorentz force. Then equation 5 reduces to u + = H = h + k (19) t If we follow the same process as outlined in the previous section, we arrive at the following simplied equations for a strongly magnetized plasma: Strongly magnetized plasma unied Maxwell equations No monopoles Gauss law Faraday law Ampere law where the current j now takes the simplied form j = 2 u 1 2 a2 ( u u ) + u u h t 2 t 2
e

= 0 =
2

(20) (21) (22) (23)

+ j = a2 t

+ = 0 t

= H u t

(24)

Here is the uid vorticity, and = u = ( + m B) u is still the generalized Lamb vector for the plasma. Notice that we have not neglected the displacement current, E/t, here (no simplications of EM Maxwells equations were introduced; only the eect of the electric eld on the uid was neglected); therefore, this model still admits a large number of waves and modes not seen in MHD. This Maxwell set is also fully isomorphic, with the speed of sound now corresponding to the speed of light in the EM Maxwell set. These equations provide a simpler formulation of the plasma in the special case that the electric eld does not signicantly inuence the uid behavior. The magnetohydrodynamic equations may be further recovered from these equations if we take the limit of a single species in the plasma (a single-uid model), and fashion the electromagnetic current from an Ohms law.
5 of 14 American Institute of Aeronautics and Astronautics

D.

Limiting case: Gasdynamic equations

Removing both electric and magnetic elds reduces equation 5 back to the usual Euler equation for isentropic gasdynamics, u + u = H = h + k (25) t In this case, there is no electromagnetic contribution, and the usual compressible uid ow limit is restored. We can once again apply the process described above to determine the Maxwell equations describing gasdynamics, Gasdynamic Maxwell equations No monopoles Gauss law
Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

=0 ( u) =
2

(26) (27)

+ ( u) = 0 (28) t ( u) + j = a2 (29) Ampere law t Notice that the charge in equation 27 and current in equation 29 are identical to those in equations 21 and 24; we call these the gasdynamic charge and gasdynamic current. Since the electric charge is much more familiar, part of the task of this paper is to expose the nature of the gasdynamic charge. Notice that we have reduced the number of species to a single-uid approximation as well. This set of equations describes the evolution of the uid vorticity and the uid Lamb vector u. A very similar set of equations was introduced previously by Kambe.3 Faraday law

= H ( u) t

III.

Analysis of unied source terms

The three Maxwell equation sets we introduced in the previous section reveal how magnetized and two-uid plasmas and gasdynamics may be evolved in time in terms of their unied elds. Two-uid plasmas admit the greatest number of modes, and the simpler case of a strongly magnetized plasma admits most of the two-uid modes if multiple species are retained. The MHD model may be recovered by reducing the strongly magnetized case to a single-uid case and neglecting the displacement current. Finally, removing all plasma modes and only retaining the gasdynamic modes reduces the elds to just the vorticity and Lamb vector of the uid. In this research, we have taken advantage of the simplicity of the strongly magnetized form of the equations and the gasdynamic form to study the unied charges and currents. To study the particular form of the source terms, two investigations were undertaken. In the rst, nite volume numerical solutions of the Euler equations in a shock tube were determined, and the uid contributions of the gasdynamic charge and current were postprocessed. In the second investigation, nite volume numerical solutions of the Brio and Wu electromagnetic plasma shock problem were executed and the uid and electromagnetic contributions to the source terms were determined. Although there remains much more analysis that can be done (particularly for locating the dierence in the unied charge by including or excluding two-uid eects), this provides an initial investigation into the structure of the charges. This analysis provides a rst step into exploring the nature of unied charges and currents for the above sets of Maxwell equations. A. Gasdynamic charge and current

A shock-capturing nite volume solver was used to solve the approximate Riemann problem at each volume interface in a one-dimensional mesh using an explicit Roe scheme. The Euler equations were solved in their conservation form, u + u uu + P t E (E + P) u x 0 = 0 0

(30)

The initial conditions were set up to imitate the Sod shock tube problem,10 1 1/8 = , = u 0 u 0 P 1 P 1/10 Left Right
6 of 14 American Institute of Aeronautics and Astronautics

(31)

The initial discontinuity was centered in the computational mesh. The gas was assumed to be calorically perfect and ideal, and the gasdynamic energy was taken as E = P/( 1) + (1/2) m u2 . The gas properties were taken as air at STP. Simulations were run using a much lower Courant number than usual (CFL = 0.05), and used 500 volumes. This provided a means of computing the gasdynamic charge as per equation 27. Using a much smaller Courant number allowed us to construct movies of the charge for analysis. Figure 1 shows the usual solution of the Sod shock tube problem after a nondimensional time of t = 1.6. Sod Gasdynamic Shock Tube Solution Several nonlinear shock structures have resolved at difDensity 1 ferent speeds, which correspond to the eigenvalues of Velocity (X) Pressure the system of conservation equations written in equa0.8 tion 30. Figure 2 shows the results of the gasdynamic charge at nondimensional time t = 1.6; individual con0.6 tributions from each term (enthalpic and spatiotemporal) from equation 26 are visible. In Figure 3, the left subg0.4 ure shows a pseudocolor contour plot of the gasdynamic charge against nondimensional position and nondimen0.2 sional time, and the right subgure shows a detail of the initial dispersion of the gasdynamic charge. Three pri0 -4 -2 0 2 4 mary charge structures are visible in the contour plots of Position (nondimensional) Figure 3. There is a compression charge moving to the right; this charge is narrow and strongly negative. The contact discontinuity is associated with a smeared dipole Figure 1. The numerical solution of the Sod shock tube problem. that is much wider than the compression charge. Finally, there is a rarefaction charge moving left corresponding to the expansion seen in the density prole (blue dashed line in Figure 2). The rarefaction charge appears as a series of small charges (much less in magnitude than the contact discontinuity charge or the compression charge), shown in the inset in Figure 2. Examination of the gasdynamic charge exposes a unique characteristic of its nature gasdynamic charge is associated strongly with gasdynamic wave structure. This is anticipated by the analytical form of the gasdynamic charge in equation 27 and the form of the current in equation 24. It is clearly visible in Figure 2 that the gasdynamic charges follow the waves in the shock tube, and dierent waves correspond to dierent charge structure. The rarefaction wave traveling to the left is comprised of a small-magnitude charge comb, or a series of charge teeth, and is shown in more detail in the inset in Figure 2. Figure 4 reveals more detail in the contact discontinuity dipole charge the compression charge. The structure of the shocks exposed by this analysis is encouraging. The entire shock tubes dynamics can be composed of a superposition of a simple number of charges that march through the shock tube with known speeds corresponding to the eigenvalues of the system of equations 30, since their contributions will have nonzero magnitude only near some changing structure in the ow. Since the eigenvalues are prerequisite to constructing a nite volume solution of the equations anyway, the speeds of the charges propagating in the shock tube are already known. The structure of each individual charge is less clear. The salient features of each charge must be determined as some function of known parameters describing the problem at hand. Examination of equation 27 (the form of the gasdynamic charge ) also matches what we expect: magnitudes the enthalpic derivative 2 H and spatiotemporal derivative t u only occur at the waves, so the results presented conrms our intuition that the charge follows the wave structure. The gasdynamic current includes more contributions than the charge, as seen in equation 24, but still reveals a simple form. Figure 5 shows the contributions of each term to the current, and the total gasdynamic current at dierent simulation times, respectively. A signicant aspect of solving the unied Maxwell equations involves having a knowledge of the source terms in order to solve for the unied elds. A rst approximation to the gasdynamic current may be guessed as a convective term, j = u, where is the gasdynamic charge. Figure 5 exhibits a comparison of this simple current model (the thin solid black line) to the actual gasdynamic current (the thick solid red line) determined from the simulations. The magnitudes do not match for the compression and contact surface charges, and the rarefaction current departs signifcantly from the convective model, but the general agreement in curve shape suggests that this simple model might be a valuable rst approach for modeling the current from the charge. The solution of the gasdynamic equations 30 for the gasdynamic charge leads us to an important conclusion: that the structure of the gasdynamic charge and current are coupled to the gasdynamic wave structure. Hence, we can begin to start a foundation of knowledge for gasdynamic charge and current by thinking about the wave structure of the gas. The presence of a gasdynamic charge or current induces a wave structure in the unied Maxwell equations.
Density, Velocity (X), Pressure (nondimensional)

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

7 of 14 American Institute of Aeronautics and Astronautics

Gasdynamic Charge in Shock Tube


600 Density (Scaled) Enthalpic Term Spatiotemporal Term Total Gasdynamic Charge

400 Gasdynamic Charge (nondimensional)

200

0 4 2 0 -2 -4 -2 -1.5 -4 -1 -0.5 -2 0 0.5 0 Position (nondimensional) 2 4

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

-200

-400

-600

Figure 2. Gasdynamic charge present in the Sod shock tube. The waves (visible in the density prole in blue dashes) correspond to the charges in the system. A superposition of strong dipole charges due to separate terms constitutes the compression shock on the far right. The contact surface moves with a dipole charge. The rarefaction charge has been shown in more detail in the inset gure. A series of charge teeth are present, and covers the range over which the rarefaction wave occupies in the gas.

The structure of each charge varies for each wave. B. MHD charges and currents in strongly magnetized ows

The previous section explored the case of gasdynamic charge and current. In the case of a plasma, Alfvn and magnetoacoustic modes are frequently important. In order to analyze the unied charge and current associated with these modes, the Brio and Wu electromagnetic plasma shock problem11 was considered. The equations of a singleuid plasma can be rewritten in a strong conservation term, which couples the electromagnetics to the uid entirely through the ux and conservation vector instead of introducing source terms.6 This allows to write: ui 0 em + p em 2 u + S / c u u 0 xx x 0 em 2 em / c u v v + S 0 y x y 0 em 2 em 0 w + S / c u w z xz 0 em 0 ( ) E + u E + p u + S em x = + (32) 0 b 0 x t x 0 by ez 0 b e z y j e 0 x x j e b y y z jz ez by
8 of 14 American Institute of Aeronautics and Astronautics

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

Figure 3. (Left) A visualization of the time dependency of the charge. The far right line corresponds to the compression. The dipole travels with a relatively constant strength. The rarefaction charge looks like ripples, and widens over time. (Right) A closer look at the initial onset of the charges resolved in the computational domain.

Gasdynamic Charge (nondimensional)

150 100 50 0 -50

Gasdynamic Charge (nondimensional)

Structure of Contact Discontinuity Charge

Structure of Compression Charge


100 50 0 -50

-100 -150 -200 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4
Position (nondimensional)

-100 -150 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5


Position (nondimensional)

Figure 4. A closer look at the charge structures in the Sod shock tube in Figure 1. (Left) The contact surface charge. (Right) The compression shock charge.

em is the Maxwell stress tensor, and u is the electrowhere Sem is the Poynting electromagnetic propagation vector, em 6 magnetic energy density. Taking the limit as c0 will reduce this system of equations to a magnetohydrodynamic simulation with magnetic diusion. The initial conditions posed by Brio and Wu for a single species are 1 0.125 u 0 u 0 v 0 v 0 w w 0 0 P P 1 0 . 1 = , = (33) b 0 . 75 b 0 . 75 x x by 1 by 1 b 0 b 0 z z 0 0 e e x x 0 0 ey ey e e 0 0 z z Left Right

9 of 14 American Institute of Aeronautics and Astronautics

Gasdynamic Current in Shock Tube


600 Density (Scaled) Term 1 Term 2 Term 3 Term 4 Total Gasdynamic Current Predicted Current (rho*u)

400 Gasdynamic Current (nondimensional)

200

0 4 2 0 -2 -4 -2 -1.5 -4 -1 -0.5 -2 0 0.5 0 Position (nondimensional) 2 4

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

-200

-400

-600

Figure 5. Gasdynamic current in the Sod shock tube. The current closely follows the charge in structure. The black line indicates a possible model for the current using a convective form, j = u, where is the gasdynamic charge. Although some scaling issues are clearly visible, the convective model seems to resolve the structure in the correct locations. The current model departs for the rarefaction current.

The system of equations 32 was solved using a nite volume Roe scheme implemented in an explicit solver. For Analytical Density 1 1 Alfvnic and magnetoacoustic charge, a single species and the Calculated Density Analytical Pressure FR full Maxwell equations were solved. Movies of the source Calculated Pressure 0.8 0.5 SC terms were postprocessed for analysis. Figure 6 presents the simulation results after ten light transit times in the computa0.6 0 CD tional domain. Figure 7 presents calculations of the unied 0.4 -0.5 charge (equation 21) at the same xed time. Figure 8 shows a comparison of the x component of current (equation 24) in the 0.2 -1 shock tube at this time. Figure 8 also presents a comparison of SS FR a convective current model and the actual current. -1.5 0 -0.4 -0.2 0 0.2 0.4 The same characteristic relationship between the charges Position (nondimensional) and waves are observed for the results here, except that more waves are admitted to this sytem of equations, and so more Figure 6. The solution to the Brio and Wu shock problem after ten light transit times. The density and pressure proles charges appear. In the single-uid magnetohydrodynamic sysare shown; they match with good agreement to the analytical tem, we expect seven (two fast, two Alfvn, two slow, and a solution. The reason for the discrepancy is that the numerical gasdynamic) waves to be captured. The magnetoacoustic (fast scheme solved the equations including the displacement curand slow) waves can manifest as shocks or as rarefactions in rent, which MHD neglects. the ow. In Figure 6, we see (from left to right) a fast rarefaction (FR), a slow compound wave (SC), a contact discontinuity (CD), a slow shock (SS), and a fast rarefaction wave (FR) still propagating in the shock tube. Figure 7 makes clear the connection between the dierent waves captured at this time and their correlation to the unied charges. Since the basic physics involved is fundamentally dierent, new structures for the charge emerge.
Brio and Wu Plasma Shock Solution
Magnetic field in Y (nondimensional) Density (nondimensional)

10 of 14 American Institute of Aeronautics and Astronautics

MHD Charge in Shock Tube


600 Density (Scaled) Enthalpic Term Spatiotemporal Term Total MHD Charge

400 MHD Charge (nondimensional)

200

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

0 4 2 0 -2 -4 -0.2 -0.15 -0.1 -0.05 -0.2 0 0 Position (nondimensional) 0.2 0.4

-200

-400

-600 -0.4

Figure 7. The plasma charge in the computational domain at the same time as shown in Figure 6. The charge is considerably more complicated than the gasdynamic form for the Sod shock tube. Still, similar charge structures are observed in the plasma case.

It is still clear that the charges are tied to the waves. The slow compound wave corresponds to a charge structure shown in the inset of Figure 7. While the charges and waves are dierent from the gasdynamic case, the correlation between structure and wave remains similar. The unied current presented in Figure 8 demonstrates similarity again to the gasdynamic case, except that since the waves of the system are dierent, the resulting current is as well. An immediate question is if the same simple convective model for the gasdynamic current, j = u, is still valid. Figure 8 presents a comparison between the predicted plasma current using the simple model j = u for the x direction. Strong currents (contact discontinuity and shocks) matches the shape well, although not always the magnitude; the smaller currents due to rarefaction waves (the two insets in Figure 8) depart from the simple convective model.

IV.

Conclusions

A theoretical recasting of the equations for a two-uid plasma reveal a set of equations remarkably similar to the Maxwell equations. A mathematical diculty presents itself here, since the equations given in 15 through 18 are not strictly isomorphic to the Maxwell equations. We hope to explore this issue in closer detail in the future. However, simplications of the plasma will reduce the set to a system of equations that is truly isomorphic to the Maxwell equations. The primary simplied models we have endeavored to study in this paper are that of a strongly magnetized plasma where the electric eld eect on the uid is small, and pure gasdynamics that is not inuenced by the presence of electromagnetic elds.

11 of 14 American Institute of Aeronautics and Astronautics

MHD Current in Shock Tube


600 Density (Scaled) Total MHD Current Predicted (rho*u) Current 4 2 0 -2 MHD Current (nondimensional) 200 -4 0.3 0.32 0.34 0.36 0.38 0.4

400

0 4 2 0 -2 -4 -0.2 -0.15 -0.1 -0.05 -0.2 0 0 Position (nondimensional) 0.2 0.4

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

-200

-400

-600 -0.4

Figure 8. The plasma current in the computational domain after ten light transit times. The current features much of the same structure as the charge here. The black solid line represents a simple convective model for the current, j = u, where is the plasma charge. This simple model shows some agreement to the current, although there are clearly scaling issue and the convective model departs for rarefaction cases.

For both strongly magnetized plasma and gasdynamics, the charge and current reduces to the same form. Two particular cases were investigated to reveal the structure of the charge and current: a gasdynamic Sod shock tube and a Brio and Wu shock tube. The resulting solutions were postprocessed for the unied charge and current. An important conclusion of this analysis is that the charges are directly tied to the waves in the system; therefore, they propagate as single charges (or, in some cases like rarefaction waves, collections of similar charges) with the eigenvalues of the system of equations. Since nite volume methods generally rely on a knowledge of the eigenvalues, this is encouraging, because the charges could be modeled using the same information needed for nite volume solutions. The structure of the charges for these two test cases has been illustrated and examined. Dipole charges correspond to discontinuities, and rarefaction waves exhibited combs of charges. The current behaved in a very similar manner to the charge. An initial test model of a convective current using the unied charge was compared to the actual current; there is still more work needed in studying the magnitude of this model, but as a preliminary rst step, it does seem to indicate the proper structure of the current. A better understanding of scaling, and particularly the rarefaction charges, is required before this model could suciently describe the current to an accurate degree. We ultimately hope that this theoretical framework might allow a new approach to predicting plasmas for engineering utility. A substantially better understanding of the charge and current is prerequisite to this goal. Robust models of the charge and current which depend on functions of the known parameters of the problem must be developed before a numerical solution of the unied Maxwell equations is attempted. If the isomorphism between the plasma equations and Maxwell equations is rigorous, then solution techniques common in electromagnetism can be applied with equal validity to the solution of the plasma equations. Particularly, since the eigenvalues of the Maxwell set will correspond to the speed of wave propagation, nite volume methods can be implemented directly. Charges and currents can be applied using empirical models. At present, these solution techniques have not been tested yet, since models of the charge and current must be known; that is why this paper has focused on this development. Also, it should be pointed out that equations 15 through 18 are not rigorously isomorphic; this problem needs to be solved as well before Maxwell
12 of 14 American Institute of Aeronautics and Astronautics

solvers can be applied to these equations for full two-uid plasmas.

Appendix: Derivation of the unied Ampere law


We have postponed the mathematical details of the derivation of the unied Ampere law until now. This derivation reviews the approach given in Ref. 1. Starting with equation 8, we want to manipulate it to achieve the form of equation 9. To do so, we must re-introduce e e the denitions of the canonical momenta, P = u + m A and the total energy, = H + m . This yields

e 2 A e 2 u + + = H 2 2 m t t t m t t e 2 u 2 A + = H + t t m t t 2 t 2
Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297
2

(34) (35)

A , then where we have just shued the terms around in equation 35. If we substitute the electric eld, E, for t 2 the last term becomes (e/m)E/t; we can substitute the electromagnetic Ampere law in for the time derivative of E here to obtain

e 2 u j = H + e + c2 + 0 B 2 t t m t 0

(36)

We have succeeded so far in recovering a curl of the magnetic eld; now we are left with the task of introducing a curl of the vorticity in order to realize an eective Ampere law for both the vorticity and magnetic eld. This can be done by breaking up the gradient of the stagnation enthalpy, H = h + k . We have e 2 u j = h k + e + c2 + 0 B 2 t t t m t 0 h = u h a2 u t (37)

If we take the gradient of the continuity equation in the form given in equation 2, we have the following identity: (38)

and applying the vector identity = u = u 2 u , we recover the curl of the vorticity, e e 2 u j 2 2 + = u h + a2 k + e + c2 B u + a 0 2 t t m m t 0 Simply moving the remaining terms to the left-hand side, we have e j e 2 u 2 2 + k + e = a2 B a2 + c0 u u h + t t m 0 m t 2 or, more simply, just as e 2 + j = a 2 B + c0 t m with j = e j 2 u 2 a2 k e u u h + t m 0 t 2 (42) (41) (40) (39)

Thus, we have equation 9 from some manipulation of equation 8. It is important to realize that this is not rigorously an Ampere law, since two curls appear that cannot be combined to yield the unied vorticity, , due to the dierence in propagation speeds, a and c0 . A true set of Maxwell equations would close this term to a curl of the unied vorticity, , with a corresponding speed that would allow a separate treatment per species. This is an important observation, because constructing a nite volume numerical scheme capable of solving these equations demands a knowledge of the eigenvalues of the system. If the unied Maxwell equations are not rigorously isomorphic to their electromagnetic form, then we cannot directly apply nite volume techniques for the electromagnetic equations with equality on the unied set. We hope to explore this issue in greater detail in a future paper. Notice that the system of equations 20 through 23 and 26 through 29 do constitute rigorous isomorphisms to the Maxwell equations, since only a single speed is present.
13 of 14 American Institute of Aeronautics and Astronautics

References
1 Thompson, R. J. and Moeller, T. M., A Maxwell formulation for the equations of a plasma, Physics of Plasmas, Vol. 19, No. 1, 2012, pp. 010702. 2 Marmanis, H., Analogy between the Navier-Stokes equations and Maxwells equations: Application to turbulence, Physics of Fluids, Vol. 10, No. 6, 1998, pp. 14281437. 3 Kambe, T., A new formulation of equations of compressible uids by analogy with Maxwells equations, Fluid Dynamics Research, Vol. 42, 2010, pp. 455502. 4 Cramer Kenneth, R. and Pai, S.-I., Magnetouid Dynamics for Engineers and Applied Physicists, Scripta, 1973. 5 Rooney, D., Moeller, T., Keefer, D., Rhodes, R., and Merkle, C., Experimental and Computer Simulation Studies of a Pulsed Plasma Accelerator, AIAA Joint Propulsion Conference, 2007. 6 Li, D., Merkle, C., Scott, W. M., Keefer, D., Moeller, T., and Rhodes, R., Hyperbolic Algorithm for Coupled Plasma/Electromagnetic Fields Including Conduction and Displacement Currents, AIAA Journal, Vol. 49, No. 5, 2011, pp. 909920. 7 Yoshida, Z., Mahajan, S. M., and Ohsaki, S., Scale hierarchy created in plasma ow, Physics of Plasmas, Vol. 11, No. 7, 2004, pp. 3660 3664. 8 Jackson, J. D., Classical Electrodynamics, John Wiley & Sons, Inc., 3rd ed., 1999. 9 Towne, D. H., Wave Phenomena, Addison-Wesley Publishing Company, 1967. 10 Sod, G., A survey of several nite dierence methods for systems of nonlinear hyperbolic conservation laws, Journal of Computational Physics, Vol. 27, No. 1, 1978, pp. 131. 11 Brio, M. and Wu, C. C., An Upwind Dierencing Scheme for the Equations of Ideal Magnetohydrodynamics, Journal of Computational Physics, Vol. 75, 1988, pp. 400422.

Downloaded by UNIVERSITY OF TENNESSEE on August 7, 2013 | http://arc.aiaa.org | DOI: 10.2514/6.2012-3297

14 of 14 American Institute of Aeronautics and Astronautics

Potrebbero piacerti anche