AIAA-2001-0795, AIAA 39th Aerospace Sciences Meeting &

Exhbit, Reno, NV, Jan. 8-11, 2001

AIAA-2001-0795

POTENTIAL PERFORMANCE OF SUPERSONIC MHD POWER

GENERATORS
Sergey O. Macheret,* Mikhail N. Shneider,** and Richard B. Miles***
Princeton University, Department of Mechanical and Aerospace Engineering,
D-414 Engineering Quadrangle, Princeton, NJ 08544
Macheret@princeton.edu
Abstract
A novel concept of hypersonic cold-air MHD power generators is discussed. Ionization of the cold air is
shown to be a critical issue, determining overall design, geometry, operating conditions, and performance
envelope. Analysis of various ionization methods, from electron beams to high-voltage repetitive pulses to
alpha particles, shows that keV-class electron beams injected along magnetic field represent the most
efficient and credible method of ionization. The requirement that the energy cost of ionization be lower
than the extracted electric power reduces the conductivity in electron beam sustained MHD channels
compared with that in conventional MHD generators. This restricts performance and calls for very strong
magnetic fields and high Hall parameters, making ion slip and near-anode processes first-order issues.
Possible problems that could be caused by hypersonic boundary layers and electrode sheaths, including
anode sheath instability and ways to avoid it, are discussed. Calculations of hypersonic power generators
for flight Mach numbers between 4 and 10 and altitudes of 15-40 km are quite promising. With 3-meter
long channel of inlet cross section of 2525 cm2 and magnetic field of 7 Tesla, several megawatts of
electric power could be generated, while spending only a few hundred kilowatts of that power on sustaining
ionization.
1.

Introduction

One of the most promising applications of magnetohydrodynamic (MHD) technology to

hypersonics is a MHD power generator at the inlets of airbreathing hypersonic engines. The generated
electricity could be used either to power various electromagnetic devices on board or to provide MHD
acceleration of the engine exhaust flow. The latter the so-called MHD bypass is a key element of AJAX
concept.1, 2
The MHD generator would reduce the total enthalpy of the flow at the engine inlet. If the MHDreduced total enthalpy becomes low enough, a conventional turbojet or ramjet engine, instead of a scramjet,
could be operated in hypersonic flight. Additionally, MHD devices are flexible and could be used for
electromagnetic, with no moving parts, control of engine inlets in off-design environments.
The practicality of the MHD power generation and bypass concepts depends on system issues,
such as cost, robustness, thrust-to-weight ratio, etc. Prior to the system studies, however, the physical
feasibility of operating hypersonic MHD channels and the extraction of significant power must be
considered, the performance envelope outlined and critical technical issues identified. In earlier papers,3-8
we suggested a concept of hypersonic MHD channels with ionization by electron beams. We developed
theoretical and computational models of these channels and concluded that substantial amounts of enthalpy
could be added to or extracted from the flow in an energy-efficient manner, although a number of issues
have to be resolved to fully assess the viability of the concept. In the present paper, we review the critical
aspects of MHD generators with external ionization and report the results of computations of performance
in a wide range of Mach numbers and altitudes.
______________________________________________________________________________________
* Research Scientist, Associate Fellow AIAA
** Research Staff Member, Member AIAA
***Professor, Fellow AIAA
Copyright 2001 by Princeton University. Published by the American Institute of Aeronautics and
Astronautics, Inc., with permission.

AIAA-2001-0795

2. Ionization in cold-air hypersonic MHD channels

2.1. Electron beams and high-voltage pulses
The first-order challenge in developing on-board hypersonic MHD generators is to create an
adequate electrical conductivity in air. With air entering the MHD channel at a static pressure of about 0.01
to 0.1 atm and temperature of a few hundred Kelvin, thermal ionization of even cesium-seeded gas is far
below that required for MHD operation. When the free-stream Mach number is very high, above Mach 12,
the flow could be put through a series of oblique shocks, increasing static temperature to 2,800 4,000 K,
at which point a conventional MHD generator could in principle be operated with 0.1-1% of potassium or
cesium seed. This configuration is suggested and analyzed in recent NASA Ames studies.9
At Mach numbers lower than 10, even after a few oblique shocks, static temperature would not
reach the 3,000 4,000 K level required for conventional seeded MHD operation. Cold hypersonic MHD
channels would, therefore, have to rely on a non-thermal ionization mechanism.
It would be a misconception to believe that a preionization at the entrance to MHD generator
would suffice. At electron number densities of about 1012 cm-3 minimally required for hypersonic MHD
operation, plasma recombination time is about 10-30 microseconds. Consequently, at flow velocity of
1,500-2,000 m/s, the conductivity would decay in just a few centimeters. Thus, the nonequilibrium
ionization has to be applied throughout the MHD channel, or, at least, every 1-2 centimeters.
The principal factor in selecting the ionization method and assessing MHD generator performance
is ionization power budget. Indeed, while in conventional MHD generators conductivity is free in a sense
that it occurs naturally at high temperature, in nonequilibrium cold-air MHD systems one must pay a
substantial energy price for maintaining the conductivity. For a meaningful operation, power cost of
ionization should be much lower than the extracted electric power. This dictates rejection of many methods
of ionization and puts a severe constraint on operating parameters of nonequilibrium MHD channels.
In Ref. 10, the power budget for plasma generation by conventional DC or RF discharges, highvoltage repetitive pulses, and high-energy electron beams has been analyzed. Electron beams were shown
to provide by far the most energy-efficient way of creating ionization in cold gas.
The power budget of sustaining nonequilibrium plasma with a prescribed electron number density
ne is determined by two factors. The first factor is the energy cost, Wi , of creating a new electron in the
plasma. The second factor is a number of ionization events in unit volume per second that has to match the
rate of electron removal in recombination, attachment, and diffusion or convection.
When a nonequilibrium plasma is sustained by a DC or oscillating electric field, the electron
energy balance is determined by electron energy gain from the field in collisions and energy losses in
various inelastic and elastic processes. Thus, electron temperature, Te, or, more exactly, electron energy
distribution function, EEDF (the latter is typically non-Maxwellian), is defined, for a given set of gas
parameters, by the ratio of electric field strength to the gas number density, E/N, or, at low densities, by the
ratio of electric field strength to the field oscillation frequency, E/.11 Since EEDF determines rates of all
electron-impact processes, including ionization, then the requirement that the ionization rate should balance
the rate of electron removal determines the value of electric field in the plasma. Typical values of E/N in
nonequilibrium discharges in molecular gases at moderate-to-high pressures lie in the range
(1 6) 1016 V cm 2 .11 EEDF is strongly non-Maxwellian at these E/N, and electron temperatures
defined by the EEDF slope at low energy are in the range 1-3 eV.11 Under these conditions, only a small
fraction of plasma electrons are capable of ionization that requires at least 10-15 eV energy. Almost all
collisions of plasma electrons with molecules result in inelastic or elastic energy losses. Especially effective
in the electron energy range of 1-3 eV is resonant excitation of molecular vibrational states, consuming in
many cases more than 90% of the discharge power. Thus, only a very small fraction, less than 0.1%, of the
power is actually spent on ionization,11 corresponding to the energy cost Wi of at least several tens of keV
per electron.10
From the point of view of ionization efficiency, it would be desirable to have electrons of very
high energy, hundreds and thousands of electronvolts, that would be immediately capable of ionization.
This would increase the ionization rate and its efficiency. In this regard, high-energy electron beams are a
natural choice. Indeed, electron acceleration in a vacuum, as opposed to a gas, environment is not
accompanied by large inelastic losses. Injection of the electrons accelerated to keV and higher energies into
the gas results in ionization cascades, as beam electrons propagate and lose their energy. The resulting

AIAA-2001-0795
energy cost of a plasma electron is Wi = 34 eV for air,12 which is only a few times greater than the
ionization energy of air molecules.
An alternative approach is to try to produce high-energy electrons in situ by applying a strong
electric field. This field would be substantially stronger not only than the field needed to sustain a steady
plasma with the given electron density, but also than the field required for initial electrical breakdown of
the gas. The strong field would increase ionization rate and reduce the energy cost per electron. The lowest
energy cost, corresponding to the so-called Stoletovs constant11 can be achieved at E/N about an order of
magnitude greater than the critical breakdown field. For air, the Stoletovs constant is 66 eV per electron11
about twice the electron cost with e-beam ionization. To maintain a prescribed level of ionization, the
strong electric field should be applied for only a short time, then the plasma should be allowed to somewhat
decay, after which a new ionizing pulse would be applied, etc. Thus, a desired average ne would be
sustained by separating electron generation and removal processes in time, with pulse repetition rate
matched to the rate of recombination or attachment. An important factor in assessing performance of the
repetitive-pulse approach is that the reduced energy cost due to strong electric field in the pulse is offset by
the need to produce more electrons in the pulse than the required average ne.
Analysis of plasma generation by both electron beams and high-voltage pulses presents a
challenge, since here ionization and electron removal kinetics is non-local (high-energy electrons at a given
location come mostly from a different location), time-dependent, and strongly coupled with transport of
charged particles, dynamics of the electric field, kinetics of chemical species and excited states, gas
dynamics, and heat transfer.
Time- and space-resolved models of plasmas and electric discharges produced by e-beams and
high-voltage pulses have been developed by us earlier.10 The simplest model consists in zero-dimensional
analytical or semianalytical calculations assuming a uniform plasma and uniform ionization profile (for ebeams), and an ideal rectangular pulse shape, ideal fast-response impedance match, and no cathode sheath
effects (for high-voltage pulses). A more sophisticated model does consider spatio-temporal evolution of
plasma and electric field, while treating motion of charged particles in drift-diffusion approximation. The
third, most detailed, level of modeling adds a full non-local kinetics of electrons, with EEDF evolution in
time and space treated in forward-back approximation.13
For illustration, Fig. 1 shows the energy cost of ionization, in eV per newly produced electron, as a
function of the reduced electric field, E/N. As seen in Fig. 1, the energy cost per electron decreases with
E/N, reaching 66 eV at E / N = 1.1 1014 V cm 2 , corresponding to the threshold field for electron
runaway effect. This minimum energy cost is about twice the ionization cost for e-beams that is also shown
in Fig. 1. However, computations of power budget for repetitive-pulse ionization show that this method,
with 10-nanosecond pulses, is at least an order of magnitude less efficient than e-beam ionization. Fig. 2
illustrates the computed power budget for sustaining a mean electron density of 1012 cm-3 in air at 10 Torr,
300 K (these parameters are close to those expected in hypersonic MHD channels). The figure clearly
shows that the power budget passes through a minimum and increases again at very high E/N. This is due
to the fact the ionization rate in strong fields becomes excessively high, and too many electrons are
produced during the pulse, wasting power. With shorter pulses, the excessive ionization is limited, allowing
to take advantage of the reduced energy cost per electron.
2.2. Alternative (exotic) ionization methods: neutrons and alpha-particles
Some authors have suggested that an adequate conductivity for MHD generators could be
produced with neutron sources. Unfortunately, no numbers were made available to support that proposal.
However, even if all technology issues related to neutron sources have been resolved, simple physics works
against neutrons as sources of ionization for MHD channels. Being much heavier than electrons and having
no electric charge, neutrons do not interact efficiently with bound electrons in atoms and molecules.
Instead, they interact quite well with nuclei, transferring energy to the motion of atoms and molecules, that
is, into heat. Thus, the efficiency of ionization versus heating is very low with neutron sources, much lower
than ionization efficiency of electron beams.
There are other products of radioactive decay that are much more efficient in ionization than
neutrons. For example, alpha particles (helium nuclei) are produced with energy in the range 4-10 MeV.
Due to their electric charge, alpha particles do interact with electrons in atoms and molecules, which results
in high efficiency of ionization energy cost per ionization event is about the same as for e-beams (34 eV).
In the following table, mean free path of alpha particle in air and aluminum foil are listed:14

AIAA-2001-0795
Table 1. Mean free path of alpha particles in air and aluminum.
Particle
MeV
4.0
10.0

energy,

Mean free path in

air at STP, cm
2.37
10.2

Mean free path in

air at 10 Torr, cm
180.1
775.2

Mean free path in

Al, micron
16.5
61.6

An immediate problem, as can be seen in this table, is that at densities typical for hypersonic flight the
mean free path of alpha particles is very large, up to several meters. Therefore, in a channel of a reasonable
size, alpha particles could barely collide with a gas molecule and would dump almost all the energy into the
wall or in ambient air. (Being much heavier than electrons, alpha particles cannot be well guided with
magnetic field).
Other problems with alpha particles emerge when particle fluxes and mass requirements are
estimated. If the walls were coated with radioactive material, only a very thin surface layer (about 10
microns) would be active in generating high-energy alpha particles. Estimates of fluxes that the surface
coating could generate are shown here for three isotopes of radium. The fluxes depend mostly on half-life
of the isotope, so that the numbers would be about the same for other radioactive materials with similar
half-lives:
1
Ra-226 (T1/ 2 = 1, 599 years ):
3.6 108
cm 2 s
1
Ra-228 (T1/ 2 = 6.7 years ):
8.6 1010
cm 2 s
1
Ra-224 (T1/ 2 = 3.67 days ):
5.7 1013
cm 2 s
To balance dissociative recombination, while sustaining 1012 electrons/cm3 in the MHD channel,
1
. Thus, only highly radioactive isotope comes close to the flux
the required flux is 2 1013
cm 2 s
requirement. Coating MHD channel walls with a thin (10-micron) layer of highly radioactive, artificially
produced Ra-224 would probably be quite complex, expensive, and dangerous. Thermal and mechanical
ablation of the thin coating in hot hypersonic boundary layer, in addition to the natural radioactive decay,
would make this ionization method even more problematic.
If power required to sustain ionization is W, and the alpha particle energy is E, then the mass of
radioactive material should be:
mrWT1/ 2
M =
, where mr is the atomic mass of the material, and T1/ 2 is the half-life.
E ln 2
Assuming the power cost of ionization of 100 kW, the required mass of Ra-226 (T1/ 2 = 1, 599 years ) is
M=3700 kg, of Ra-228 (T1/ 2 = 6.7 years ) M=15.5 kg, and of Ra-224 (T1/ 2 = 3.67 days ) M=23.3 g.
Instead of coating the wall with radioactive material, the material might be injected as 10-micron particles.
In that case, the material would be constantly consumed and ejected into the atmosphere. With flow
residence time of the order of 1 ms, the consumption rate would be from 23 kilograms/s (Ra-224) to 3.7
million kilograms/s (Ra-226). These requirements are quite unrealistic.
To summarize this discussion, no exotic ionization source could meet the requirements of
nonequilibrium hypersonic MHD channels. Electron beams, on the other hand, are promising as the most
energy-efficient method of ionization. As to the repetitive high-voltage pulses, they require about an order
of magnitude more power to generate conductivity for hypersonic MHD operation than the electron beams
do. This comparable inefficiency would constrain performance of MHD systems with high-voltage pulse
ionization even further compared with those ionized by e-beams, although in some regimes high-voltage
pulse ionization could still be acceptable.
3. MHD Channels with Ionization by Electron Beams: Basic Configuration and Constraints
The basic configuration of an MHD channel with e-beam ionization3-5 is shown in Fig. 3. An array
of segmented electrodes allows a transverse electric current to flow through the gas. The magnetic field,

AIAA-2001-0795
orthogonal to both electric field and the flow direction, produces the accelerating or decelerating jB force
(in Fig. 3, the accelerator case is shown). Total flow enthalpy is affected not only by the jB force, but also
by the inevitable Joule heating and by the energy input from the ionizing beams. The beams, as shown in
Fig. 3, have to be injected along magnetic field lines, since the electrons tend to spiral around the field
lines, and calculations show that the Larmor radius is much less than a millimeter at magnetic fields of
several Tesla and higher.
To extract an enthalpy of a few MJ/kg from a high-speed (Mach 6 8) air flow at densities on the
order of 0.1 atmospheric density, over a reasonable length, 1-4 meters, an electron number density at least
on the order of 1012 cm-3 is required.3-5 Estimates performed in Refs. 3, 5, 7, and 8 indicate that this electron
number density could be sustained by an array of electron beams with energy of 5-50 keV and beam
current density of a few mA/cm2.
As a practical matter, it would be impossible to have electron beams fill out the entire sidewall of
the channel. Electron beams will be injected through foils or windows shown in Fig. 3 as a periodic array of
slots. The windows or foils would be heated by the beams passing through, and also would have to be of
sufficient mechanical strength. If the beam current density is on the order of 1 mA/cm2 or less, thin metallic
foils with passive cooling should suffice.15 For higher beam current densities, either active cooling or the
new technology of so-called plasma windows or ports 16, 17 would have to be utilized.
The rate of energy loss by the beam electrons is fairly independent of energy if the electron energy
is greater than a few hundred keV.12 However, in the 30-keV range, the energy loss per unit length
traversed increases strongly as the electron energy decreases. Thus, the energy deposition per unit length
by the beam would not be uniform across the MHD channel. As shown in Refs. 3-5 and 8, relatively low
ionization rate in the boundary layer is actually quite beneficial: the problem of short-circuiting through the
hot, highly conductive, boundary layer, limiting performance of conventional MHD devices, could be
minimized in devices with electron beam ionization. As to the core flow, ionization rate there could be
made quite uniform by injecting two beams from the opposite side walls.3, 4, 8
One important advantage of using and external ionizer is that plasmas with externally-sustained
ionization are inherently more stable than those of self-sustained nonequilibrium discharges.11, 15 Indeed,
ionization generated by electron beams is essentially decoupled from both temperature and electric fields.
A local positive fluctuation in temperature, which would have led to arcing instability in self-sustained
discharges, would result in reduced ionization by electron beams and reduced local heating, stabilizing the
plasma.
For better operation of an MHD channel, electrical conductivity, proportional to the ionization
fraction, should be high. However, when the conductivity is created by an external ionizer, such as an ebeam, higher conductivity means that more energy will have to be spent on ionization. As mentioned
earlier, the energy cost of ion-electron pair created by high-energy electron beam is Wi = 34 eV . Since
at steady state the ionization rate is balanced by the rate of recombination, the minimum power per unit
volume required to sustain electron density ne is kdr ne2Yi , where kdr 107 cm 3 / s is the rate constant of
dissociative recombination. This power has to be substantially lower than the power extracted from the
1
flow, which can be, per unit volume, estimated roughly as u 3 / L , where is the gas density, u is the
2
flow velocity, and L is a characteristic length on the order of the channel length. The criterion
u
ne < u
(1)
2LkdrWi
with the density of about 0.1 normal atmospheric density, velocity corresponding to Mach 6 8 flight, and
L of 3 meters (see below), requires that ne be not higher than about 1012 cm-3. At pressures of 0.01 0.1
atm, this results in electrical conductivity on the order of 0.1 1 mho/m, one or two orders of magnitude
lower than typical conductivity in conventional MHD channels.
With the conductivity that low, appreciable MHD effects can be obtained only in very strong
magnetic fields, at least several Tesla, and preferably 10-20 Tesla. An immediate consequence of having
very strong magnetic fields at low gas density is that electron Hall parameter, e, defined as the ratio of
electron cyclotron frequency and the electron-neutral collision frequency, becomes very large, 10 and
greater. A possibility of operating a hypersonic MHD channel with e-beam-sustained ionization at very
high Hall parameters is a very challenging issue. One effect that has to be taken into account is the so-

AIAA-2001-0795
called ion slip.18, 19 To estimate the magnitude of this effect, consider basic physics of MHD deceleration of
weakly ionized gas.
When a weakly ionized gas enters an MHD generator, Faraday e.m.f. causes electric charges to
move in the transverse direction, and the resulting Lorentz force decelerates the plasma. Since electrons are
very light, they experience the deceleration first, while ions, due to inertia, try to move at constant velocity.
However, even a slight separation of ions and electrons generates an ambipolar electric field. This field
glues electrons and ions together, so that they move as a whole, maintaining quasineutrality, and
experiencing deceleration together. (It can be demonstrated that the ambipolar field is exactly equal to the
axial Hall field). To decelerate the weakly ionized gas, its neutral molecules must be affected. The
molecules cannot directly experience electromagnetic forces, and they could be decelerated only in
collisions with ions and electrons. Those collisions manifest themselves macroscopically as a friction force.
When velocities of electrons and ions relative to the neutrals are equal, electrons, due to their very light
mass, carry only a very small momentum relative to the neutrals. Therefore, it is the slip, or velocity
difference, between ions and molecules that transfers the Lorentz force to the molecules.
The transverse electric current in the channel can be expressed as
jy = ui B
(2)
where is a numerical factor between 0 and 1 (for a Faraday generator, =1-k, where k is the loading
factor); is the electrical conductivity; ui is the ion (and electron) axial velocity, and B is the magnetic
field. At quasi-steady state, the Lorentz force jy B is equal to the molecule-ion friction force per unit
volume:
jy B = ui B 2 = kinninmn (u ui )

(3)

where kin is the rate coefficient of ion-molecule collisional momentum transfer; n , mn , and u are number
density, mass, and axial velocity of molecules, and ni is the ion number density. The conductivity can be
written as:
e 2ne
(4)
mkenn
where ne and m are the number density and mass of electrons, and ken is the electron-molecule collision
rate coefficient. It is also convenient to introduce electron and ion Hall parameters:
eB
eB
e =
; i =
(5).
mkenn
mnkinn
=

The ratio of the ion and electron Hall parameters is:

i
k m
m
= en
<< 1
kinmn
mn
e

(6)

From equation (3), using definitions (4) and (5), we obtain the relation between the ion-electron velocity
and the gas velocity:
u
ui =
(7)
1 + e i
When the electron Hall parameter is not very large:
1 mn
4
(8)
e <
m
the ion-electron and gas velocities are close to each other. However, as magnetic field strength increases,
the product in the denominator of equation (7) becomes larger than 1, and ions appreciably slip against
molecules. Qualitatively, ion slip reduces performance of MHD generators, since the velocity ui in the
expression for the Faraday e.m.f., ui B , is smaller than the gas velocity u. A simple recipe for including the
ion slip effect into a formal set of MHD equations
the factor (1 + e i ) (see the next section).

18

is to divide both conductivity and Hall parameter by

With performance reduction by the low ionization level and the ion slip effect, the channel length
necessary for a substantial enthalpy extraction could get very large. The characteristic length L needed to
substantially reduce the flow enthalpy can be estimated from energy balance:

AIAA-2001-0795
1 2
u jy B L = ui B 2L .
2
Taking into account formulae (4), (5), and (7), we obtain the estimate:
1 + e i
1
L u i
,
2
e i
1
1
where i =
=
kinni
kinne

(9)

(10)
(11)

is the mean time for a molecule to experience a collision with ion.

The physical meaning of the estimate (10) is that for a gas particle to experience MHD
deceleration, it needs to have at least 1 collision with an ion. Therefore, the minimum length of the channel
is roughly the flow velocity multiplied by the i (or Mach number of the flow multiplied by the molecule
mean free path with respect to collisions with ions). The required channel length is minimal at very strong
magnetic fields, when
1 + e i
(12)
1.
e i
As magnetic field strength decreases, when e i << 1 , the required channel length increases as
1

(e i )

Numerical estimates with formula (10) show that if the density of electrons and ions reaches 1014
cm , typical for conventional seeded MHD generators, the length L stays within a few meters even at
electron Hall parameters of about 3 and loading parameter k of about 0.9. In contrast, the low electron
number densities (1012 cm-3) in e-beam-sustained channels require electron Hall parameters above 10 and
loading parameters of about 0.5 to extract a substantial amount of enthalpy in 3-4 meters. The loading
parameter of 0.5 means that a substantial part of the power would be dissipated in the channel, increasing
entropy and decreasing total pressure.
Note that there are important reasons for selecting an expanding MHD channel shown in Fig. 3.
First, proper expansion would help to avoid adverse pressure gradients, shocks, too rapid growth of the
boundary layer, and flow separation. Second, since temperature in MHD channels at flight Mach numbers
up to about 8 is below 2,000 K, electron attachment to oxygen could be a problem. To avoid attachment,
gas density should be kept low, otherwise the power budget for maintaining the conductivity would be
prohibitively high.
-3

4.

Anode sheath problems in Faraday and Hall generators

4.1. Arcing in Faraday channels and advantages of Hall generators

The need to have very large Hall parameters makes operation of a Faraday channel problematic. In
a Faraday channel, allowing a longitudinal Hall current to flow would sharply reduce performance. The
Hall current could be prevented, in principle, by segmenting electrodes (an ideal Faraday channel).18
However, the problem would exist in the boundary layer, where due to the low gas velocity an unopposed
Hall current would flow along the anode. At high Hall parameters, electric current near the anode would
flow almost parallel to the surface, and electrons would enter the anode near its edge, where electric field is
the strongest. The current concentration near anode edges, separated by insulation, would lead to arcing and
could seriously impede performance of the Faraday channel.
An empirical constraint on operating conditions of conventional Faraday MHD channels with
thermal ionization of alkali seed, is that the longitudinal electric field should not exceed about 5,000 V/m.9
If the longitudinal voltage gradient is higher, arcing between two neighboring electrode segments separated
by insulation would start, reducing MHD performance and damaging the hardware.
Thus, it could be beneficial to use Hall configuration, where opposing pairs of segmented
electrodes are short-circuited, and the longitudinal Hall current is collected.18 In addition to being a natural
choice at very large Hall parameters and the simplicity of having only one electric load, the Hall generator
could help solve the electrode-edge current concentration problem. Indeed, because of the strong shortcircuit longitudinal current, electrons would enter anodes almost normally to the anode surface, and only a
small fraction of the current would concentrate at the electrode edge. One important disadvantage of Hall

AIAA-2001-0795
generators is that it is hard to collect the current flowing along the wide channel. A partial solution to the
second problem is provided by using the so-called diagonal connection18 that is close in performance to the
Hall configuration at large Hall parameters, but makes the task of current collection easier.
In the Hall generator cases we computed, Ex is within 5,000 V/m in the second half of the channel,
but can exceed the 5,000 V/m value in the first half of the channel by a factor of 2 8, which could be a
reason for concern. However, it is not clear whether the 5,000 V/m empirical limit can be applied to Hall
channels with e-beam ionization. First, gas temperature even in the boundary layer is much lower than the
3,000 4,000 K temperatures in conventional seeded channels. Therefore, thermal ionization is negligible,
and electrical conductivity is controlled by e-beams, as in the core flow. Additionally, melting and
sputtering of electrode and insulation material would certainly be less severe in our low-temperature cases.
One has to bear in mind that Hall parameters in the boundary layer would be quite high (up to 30-50) in our
cases, impeding electron mobility across magnetic field and thereby substantially increasing the voltage
required for breakdown and sustaining arcs. Finally, arcing could have less of an effect on performance of
Hall generators, where there already is a large longitudinal current, than on performance of Faraday
channels. In any case, the issue of maximum longitudinal field that can be tolerated in cold Hall channels
with e-beam ionization should be addressed in future experimental and theoretical work.
4.2. Coupling between electrode sheaths and boundary layers: anode sheath instability
Understanding near-electrode space-charge sheaths and fluid boundary layers in hypersonic MHD
channels is critical because of a strong coupling between charge particle kinetics, electromagnetics, and
fluid mechanics. Here, we only present a qualitative analysis, illustrated by a few sample calculations. Note
that turbulent boundary layer calculations in Refs. 3-6 and in the present paper used the turbulence model
described in Ref. 6. Turbulent transport was accounted for by an effective eddy viscosity, using the
algebraic Prandtl-Boussinesq model.20, 21
Modeling of plasma flows in MHD power generators and accelerators is usually done within the
framework of single-fluid magnetohydrodynamics. For processes inside the plasma, far from the walls and
electrodes, this approximation is quite justified. However, because a strong transverse electric current has
to be sustained in the channel in order to give rise to accelerating or decelerating ponderomotive (or
Ampere) forces, pronounced cathode and anode charge sheaths are necessarily formed through which the
current has to flow. In a strong transverse magnetic field, electron mobility is reduced, which can
dramatically alter characteristics of the near-electrode sheaths, especially those of the anode sheath. The
thickness of the sheaths can be comparable with that of the hypersonic fluid and thermal boundary layers.
This can result in a strong coupling between the charge sheaths and fluid boundary layers. For example,
characteristics of a charge sheath would be affected by the reduced density, high temperature, and
turbulence in the hypersonic boundary layer, while the ponderomotive force could significantly increase or
decrease the effective viscosity, thus changing the boundary layer.
2D modeling of ionized gas flow in an MHD accelerator depends substantially on the choice of the
coordinate plane. In the (x , z ) or (u, Bz ) plane, there is no electric current, and the plasma is quasineutral
everywhere except in near-wall regions with the thickness of the order of the Debye length. The Debye
length is negligibly small in comparison with the boundary layer thickness, , and the voltage fall across
the near-wall regions, Vz ~ Te , can be neglected. Therefore, the flow in this plane can be treated with
single-fluid magnetohydrodynamics.
Much more complex is the flow in the (x , y ) or (u, jy ) plane, where electric current jy gives rise to
the ponderomotive force directed along x. Near the walls that are also electrodes, charge separation occurs
forming the anode and cathode sheaths. In conventional glow discharges, the anode voltage fall Va and the
anode sheath thickness da are negligible compared with the respective cathode-sheath values, Vc and dc .11
The dramatic difference between the anode and cathode sheaths stems from the strong difference between
electron and ion mobilities under normal conditions, e / + ~ 102 . However, in our case, electrode sheaths
are immersed into hot and rarefied hypersonic boundary layers. Additionally, and very importantly,
electric current in the sheaths has to flow across a strong magnetic field that substantially reduces electron
mobility. Thus, one could expect boundary layers and electrode sheaths to be quite different from those in
conventional non-self-sustained transverse discharges.

AIAA-2001-0795
Let us find out how the presence of a strong magnetic field Bz and the non-uniform density profile in
the boundary layer, N (x , y ) , affects structure and parameters of cathode and anode sheaths in a
hypersonic-flow transverse discharge supported by electron beams.
A key problem in modeling electrode sheaths that are also boundary layers is to determine what
heating mechanism viscous dissipation or Joule heating is dominant. Viscous dissipation rate in the
boundary layer is:

u
( + T ) u

(13)
W (y ) =

y
y
2
An estimate of this dissipation rate for a turbulent boundary layer is W T uext
/ 2 , where T is the

effective turbulent dynamic viscosity coefficient, averaged over the boundary layer. According to the
algebraic Prandtl-Bousinesq model in Clauser approximation, modified by Cebeci and Smith,21

T = uext

(1 u / u

ext

)dy uext , where uext is the flow velocity at the edge of the boundary layer,

and 0.0168 .
When evaluating Joule heating in the boundary layer, one has to bear in mind that, although in
Faraday-type MHD devices there is no longitudinal current ( jx = 0 ) in the core flow, the longitudinal
current in the boundary layer can be quite large. Therefore, the Joule dissipation rate can be calculated as:
WJ = ( jx2 + jy2 )/ ,
(14)
where current density and electric field components in the case of a Faraday channel are:3, 4, 18
(E x* Ey* )
(Ey* + E x* )
=
;
,
jx =
j
y
1 + 2
1 + 2
E x* = (Ey uB )ext = const ; Ey* = Ey uB .

(15)
(16)

Here is the electrical conductivity, and is the Hall parameter. In the core flow of Faraday-type MHD
devices, where jx = 0 , Joule heating rate is WJ = jy2 / = jy Ey* = (k 1)2 u 2B 2 , where k is the loading
parameter. Estimates show that the Joule heating rate in the core flow is much lower than the viscous
dissipation rate per unit volume in the boundary layer.
However, in electrode sheaths immersed into heated and rarefied boundary layers, Joule heating
greatly increases as compared with that in the core flow. First, due to charge separation and reduced
mobility of charge carriers, the transverse electric field may greatly exceed that in the core flow. Second,
because of the low gas velocity, the longitudinal Hall current becomes essential:18 jx (y )jy . In the
general case, the term (14) should be added into the energy equation of boundary layer theory, as was done
by Kerrebrock.22 Development of a self-consistent theory of boundary layers and electrode sheaths is a very
complex problem. In the first approximation, we can uncouple electrode sheaths from the boundary layers
and calculate the sheath structure as an overlay on independently computed boundary layers. If then it turns
out that Joule heating in the sheath is small compared with the viscous dissipation rate, such an uncoupled
analysis would be complete. If, however, the Joule heating is stronger than the viscous dissipation, this
would indicate a need for a fully coupled computation, and would point to a possible instability. Indeed, to
sustain the current, the anode voltage fall becomes very large. This increases Joule dissipation in the anode
sheath and, when the Joule dissipation becomes comparable with the viscous energy dissipation,
WJ = ( jE ) ~ W ~ ( + T )u 2 / 2 , could lead to an instability:

WJ W T N e ~ 1/ N Ey ,VA WJ .

The instabilty chain can be broken by increasing the ionization rate in the anode sheath to
approximately that in the core flow and/or by cooling the gas in the sheath. In fact, both cooling resulting in
the density increase and increasing e-beam current density in the anode sheath would enhance ionization
and stabilize the sjeath.
Fig. 4 shows profiles of plasma parameters across the anode sheath for the Mach 6, 30 km altitude
case. Physical models and algorithms can be found in Refs. 3-6. The sheath was computed for the cross
section 1 meter downstream of the MHD channel entrance. Figs. 4a, b correspond to ionization rate
q(y ) = q(y max )N (y )/ N (y max ) , that is, with no special cooling of the sheath and with e-beam current

AIAA-2001-0795
density equal to that in the core flow. Fig. 4c and d correspond to ionization rate
0.1
q = q(y max ) N (y ) / N (y max ) const , which, as described earlier, could be achieved by increasing the
local e-beam current density and by cooling the anode sheath. As seen in the figures, the enhanced
ionization rate can dramatically decrease the anode voltage fall and thus stabilize the anode sheath.
4.3. Electron removal and sheath growth between e-beam injection slots
In Hall generators, opposing pairs of electrodes are short-circuited, so that the current across the
channel is quite strong. This, together with ionization by a periodic array of electron beams, creates a
peculiar problem of sheath growth and current self-limitation. Indeed, between the beam injection slots,
there is no ionization, and the plasma decays. In addition to recombination and attachment processes,
electrons are swept away by the strong current into the anode. Consequently, the space charge sheath
growth near the anode, which could limit or even completely stop the current (Fig. 5). We have made
analytical estimates of this effect for an ideal Hall channel with short-circuited electrodes (loading
parameter k=1) where ionization is due to periodic array of electron beams only, neglecting electron
attachment and formation of negative ions (that is, assuming ne = n+ = const in the plasma), assuming
constant flow velocity between beam injection slots, and neglecting ion slip and boundary layers.
With these assumptions, the problem becomes quasi-one-dimensional. In y direction, between the
regions of ionization charge sheath thickness, s , grows along the x direction (region I), and the
quasineutral plasma width, H s , decreases along x.
The profile of potential across the channel is determined by the Poisson equation:
2
e
= (n+ ne ),
2
y
0
(17)
( 0) = (H) = 0
Putting ne = 0 at y s and ne = n+ at s < y H s , the profile is:
e
y

n+sy(1
y s
),

s
2

0
(y ) =
e n+s 2
(H y ), s <y H

20 (H s )
Electric field in the region II produced by the charge sheath of the thickness s :
e n+s 2
e n+s 2

E II =
=

.
20 (H s ) 20 H
y II

(18)

(19)

As a result, the effective field in y-direction is Eeff = E II uB . The direction of the electrostatic field
E II is opposite to that induced by the Faraday e.m.f., uB . Therefore, the growth of the sheath thickness s

leads to decrease of Eeff and the electric current.

The drift charge fluxes of electrons and ions in the plasma in y-direction are:
[(E uB ) + E x e ]
,
e,y = ne v e,y = ne e II
1 + e2
[(E uB ) 2E x + ]
,
+,y = n+ v +,y = n+ + II
1 + 4+2

(20)
(21)

where E x = kuB e .
Electron current across the channel ceases when e,y = 0. For this, the electric field should be:
E II = (uB E x e ) = (1 + k e2 )uB .

(22)

From (3) and (6), the critical sheath thickness corresponding to current termination is:
s ,c 20uB(1 + k e2 )H / ene 1.05 104 uB(1 + k e2 )H / ne , m.

(23)
18

-3

For example, at k = 1 , u = 1500 m/s; e 10 ; B = 7 T; H = 0.5 m; ne = 10 m , s ,c 7.65 mm.

10

AIAA-2001-0795
To get a better idea of the sheath growth between beam injection slots, we solved a set of 1D
equations consisting of the Poisson equation (17) and the balance equations:
n+ +,y
+
= n+ne ,
(24)
t
y
ne e,y
+
= n+ne ,
(25)
t
y
with initial and boundary conditions:
ne (t = 0, y ) = n+ (t = 0, y ) = qi / ; +,y = 0 at y = H and e,y = +,y at y = 0 .
(26)
Expressions for electron and ion fluxes are:
[( / y uB ) 2E x + ]
[( / y uB ) + E x e ]
, +,y = n+ +
.
e,y = ne e
2
1 + 42+
1 + e

(27)

The results of computations of the current and s versus time as an element of the gas moves between the
beam injection slots for the following set of conditions: n 0 = 5 1011 cm-3; u = 1500 m/s; p=25 Torr; B=7
T; and T=290 K are shown in Fig. 6. As follows from the analysis and as seen in the figure, the beam
injection slots should be placed as close to each other as practically possible in order to prevent the sheath
growth and current termination. Also, the undesirable sheath growth could be mitigated by increasing the
channel width H (this would decrease the surface-to-volume ratio, thus suppressing the charge removal
through the electrode surface as compared with volume ionization-recombination processes). We took this
into consideration when modeling and optimizing Hall MHD channels.
5. Quasi-1D Modeling of MHD Generators with E-beam Ionization
5.1. Basic Equations
In the case of an ideal Hall power-generation channel, the set of equations5 is similar to the set of
equations used in Refs. 3 and 4 for an ideal Faraday generator (Eqn.(30) in Ref. 3):
d(uA)/ dx = 0,
uA = const ,
where A=A(x) is the MHD channel cross-sectional area
udu / dx + dp / dx = jy Bz ,
ud( + u 2 / 2)/ dx = jx E x Qv + Ev Ev0 (T ) / VT (T ) + Qb,

dEv
E Ev0 (T )
E d
= Qv v
/u + v
; Qv = v ( jy2 + jx2 )/

dx
T
dx
(
)

VT

p = RT = ( 1)

(28)

In addition to continuity, momentum, and energy equations, this set includes the vibrational energy
equation.
Nonequilibrium and equilibrium vibrational energy can be expressed through the respective
temperatures (Tv is the vibrational temperature) by the Planck formula:
0
0
Ev = N v = N
; Ev0 = N v0 = N
(29)
exp( 0 /Tv ) 1
exp( 0 /T ) 1
where 0 is the lowest vibrational quantum of nitrogen molecules, and the temperatures are expressed in
energy units.
The vibrational excitation term Qv can be expressed as a fraction v of the Joule heating rate:
Qv = v ( jy2 + jx2 )/ .

(30)

The fraction v is determined on the basis of the solution of Boltzmann kinetic equation for plasma
electrons.23

11

AIAA-2001-0795

Total power extracted from the MHD channel,

Wext =

jx E x dV , and the efficiency,

= Wext / uA(u 2 / 2 + c pT ) x =0 , can be calculated after computing the flow.

According to Ref. 18, to take into account the ion slip effect, the conductivity and Hall parameter
in all equations for quasineutral plasma flow should be substituted by:
= /(1 + e i ) and e = e /(1 + e i ) .
(31)
For example, with the ion slip taken into account, Qv = v ( jy2 + jx2 )/ .
In this case equations for the currents jx and jy take the following form:18
jx = E x* jy

(32)

jy = Ey* + jx

where E x*,y are the electric field components in the reference frame moving with the gas.
Current density and electric field need to satisfy the following conditions in a steady state MHD
channel:
i j = 0
(33)
E = 0
In our case, ionization by e-beams occurs in narrow layers, and between the layers we have a
motion of decaying plasma (see Fig. 3). Therefore, electron density and conductivity vary along the flow.
In the 1D model, the transverse component of the current, jy , does not depend on y. Therefore, because of
continuity of the total current, the longitudinal current is I x = jx (x )S (x ) = jx (0)S (0) = const . Thus, the
longitudinal current density in an arbitrary selected cross section S (x ) is
jx (x ) = I x / S (x ) .
(34)
Longitudional current at the channel inlet is found from the standard 1D model :18
uB(1 )
,
<1.
(35)
jx (0) =
x =0
1 + 2
The transverse field Ey* in (19) is an effective transverse e.m.f. per unit width. In the present 1D
model, as in the case of uniform volume ionization in conventional Hall MHD generators,
Ey* = uB .

(36)

*
x

Taking into account (34) and (36), longitudinal field E and transverse current density jy can be easily
determined from the set of equations (32).
Parameter < 1 is one of optimization parameters in our model.
In the standard Hall MHD generator model,18 loading parameter k is defined as k = E x / uB .
In the present non-uniform case, we can define an effective loading parameter k :
Lx

k=

Lx

E x*dx

uBdx

(37)

The plasma in both accelerator and power generator channels is modeled as consisting of
electrons, positive and negative ions, whose number densities ne , n+ , n obey the quasineutrality:
n+ ne + n . The set of equations for kinetics of charge species, accounting for electron beam-induced

ionization rate (qi term), attachment of electrons to molecules with formation of negative ions (frequency

a), collisional detachment of electrons from negative ions (rate constant kd), and electron-ion and ion-ion
recombination (rate coefficients and ii, respectively), is:

12

AIAA-2001-0795
ne e
+
= qi + kd Nn ane n+ne ,
t
x
n+ +
(38)
+
= qi iinn+ n+ne ,
t
x
n
+
= kd Nn + ane iinn+
t
x
For 1D steady flow ( ne,+, / t = 0 ), the fluxes of charged species can be written simply as
e,+,(x ) = ne,+,u p . In general, plasma velocity ui differs from the gas velocity u due to the ion slip effect.

Because of quasineutrality, there is a single electron-ion, i.e. plasma, velocity in the laboratory reference
frame. This plasma velocity, as follows from Eqn. (3) (see also Ref. 19), is equal to:
u p = ui u jy B / n+mn in ,
(39)
where mn is the ion mass, and in is the ion-molecule collision frequency.
The approximate formula for the e-beam ionization rate is:7
2 j
qi = b b .
eWi Z

(40)

Rate coefficients of electron-ion and ion-ion recombination, and of electron attachment and detachment
processes, discussed in Refs. 3 and 4, were taken from Refs. 24 and 25. Since some of those rate
coefficients depend on electron temperature Te, it is important to calculate the electron temperature in the
modeling. In our computations, electron temperature was determined from the tabulated data on electron
diffusion and mobility coefficients of Ref. 23. In that paper, the diffusion and mobility coefficients are
listed as functions of E/N, determined from experimental data and extrapolation based upon solution of
Boltzmann kinetic equation for electrons in air.
5.2. Estimates of chemistry and vibrational relaxation
Nitrogen vibrational relaxation time is sensitive to both temperature and the concentration of atomic
oxygen:26

141 N
1
128
= N 7 1010 exp 1/ 3 + O 5 1012 exp 1/ 2 , s 1
(41)
T
T N

VT

where NO and N are the number density of oxygen atoms and the total number density of gas molecules,
respectively, expressed in cm 3 / s .
Oxygen atoms are produced by direct dissociation of oxygen molecules by high-energy beam
electrons. Additionally, beam electrons dissociate nitrogen, and, if the gas temperature reaches 600-700 K,
oxygen atoms are produced in the reaction N + O2 NO + O . Since the energy threshold for
dissociation is comparable with that for ionization, the beam-induced dissociation and oxygen atom
production rate (in cm 3 s 1 ) can be estimated as approximately equal to the ionization rate, which, for
Mach 6-8 cases at 30 km altitude, ranges from 4 1017 down to about 5 1016 cm 3 s 1 . An additional
channel of production of O atoms is dissociative attachment of low-energy plasma electrons:
e + O2 O + O . However, in all computed cases, attachment rate was at least an order of magnitude
lower than the ionization rate (so that the ionization is balanced by electron-ion recombination). Therefore,
the beam-induced ionization rate can serve as a good approximation for the O atom production rate.
O atoms could be destroyed in three-body processes such as O + O2 + M O3 + M ,
O + O + M O2 + M , and O + N + M NO + M . However, with the gas number density of
1018 cm 3 or lower, these recombination processes would take at least 10 milliseconds, while the flow
residence time in 3-meter long MHD channel is only 1-2 milliseconds. If nitrogen vibrational temperature
gets as high as 3,000-6,000 K, the reaction O + N 2 NO + N will become significant in removing O

atoms. However, the reaction N + O2 NO + O will replenish O atoms. Thus, a reasonable estimate of
O atom density can be done by putting their production rate equal to the beam-induced ionization rate and
neglecting all chemical removal processes, so that the produced atoms simply move with the flow. Since

13

AIAA-2001-0795
the flow residence time in e-beam ionized regions is about 1 millisecond, the relative concentration of
oxygen atoms reaches NO N 104 103 .
With gas temperatures in the range 400 900 K and O atom density quite low, nitrogen
vibrational relaxation time will be longer than 10 milliseconds. Therefore, vibrational temperature will keep
growing along the channel and could be quite high at the exit.
5.3. Results of 1D modeling
1D modeling was performed for many cases, corresponding to a range of flight Mach numbers and
altitudes. A representative case corresponds to Mach 6 flight at the altitude of 30 km. The flow was
assumed to pass through an oblique shock formed by a 10-degree wedge upstream of the MHD channel
entrance. The inlet cross section of the channel was taken as 25x25 cm2, and the exit cross section as
75x75 cm2. The channel length was fixed at 3 meters. The magnetic field was assumed to be uniform and
equal to B=7 Tesla, which is close to the current practical limit of superconducting magnets. Electron
beams were assumed to be injected through periodically spaced slots from both sides of the channel, as
shown in Fig. 3. The slot width was taken to be 0.5 cm, and the spacing between the slots 2 cm. The
beam energy was adjusted according to the channel width and the gas density at the location. The current
density of the beams was varied so as to give the best power extraction and the lowest stagnation
temperature at the exit, while keeping the static pressure from increasing.
The set of equations (28) and (38) at steady state was solved by the modified 2nd order Euler
method. The number of equal spatial steps was n 1000 . Further increasing the number of steps was
found to produce negligible variation of results.
The results are shown in Figs. 7-9. The main result is that, for the chosen channel geometry, more
than 3 MW of electric power could be extracted from the flow, while spending only about 180 kW on
ionization by e-beams. The extracted power is about 35% of the enthalpy flux into the channel.
As seen in Fig. 7, the gas is expected to be in a state of vibrational disequilibrium at the MHD
channel exit. This is an effect of low-energy plasma electrons vibrationally exciting molecules in collisions,
thus converting a part of flow energy into the energy of molecular excitation. Because of the low density
and temperature, vibrational relaxation in the channel is slow, and excited molecules exit the channel
virtually without quenching. The relaxation could occur either in the diffuser or in the engine downstream
of the MHD channel, where density and temperature increase.
Fig. 10 summarizes performance predictions for various Mach numbers and altitudes. The main
result is that a large percentage of flow enthalpy, corresponding to several megawatts of power, could be
extracted as electricity, while spending a reasonable small fraction of the extracted electric power on ebeam ionization.
5.4. Boundary layer heat losses
To assess effects of viscous and thermal boundary layers, we used methods and equations of Refs.
3, 4 and 6. 2D structure of quasi-steady turbulent boundary layers was computed with the method of Ref.
21, taking into account the decelerating ponderomotive force j B , as in Refs. 3, 4 and 6. The transverse
pressure gradient in the boundary layer p / y = 0 . In Ref. 6, boundary layers were computed assuming
constant pressure along the channel. In this paper, however, more realistic calculations were performed: the
longitudinal pressure gradient p / x for the boundary layer computations was determined by
interpolating the results of 1D calculations described in this section. Only x-component of the
ponderomotive force, jy Bz , was included in the calculations.
Since some energy is extracted from the gas and converted to electricity, gas temperature in the
boundary layer is lower than that computed disregarding the energy extraction. To approximately take this
effect into account, transverse current density, jy (x ) , was reduced in the computations by about a factor of
1.5 2 from its value in the 1D model, until x-profiles of gas velocity and temperature at the outer edge of
the boundary layer closely matched those in the 1D model. In thermal boundary layer computations, wall
temperature was assumed constant and equal to 600 K.
Computations for flight Mach number of 6 (30 kilometer altitude) showed that boundary layers
grow to about 1 cm at the channel exit, which is acceptable for the 75-cm channel width.

14

AIAA-2001-0795
To find out what fraction of the total flow enthalpy is lost to the cooled walls, heat flux averaged
L
1
over the channel length L was computed: q = (T / y ) y =0 dx . The wall heat flux in Mach 6 and
L 0
Mach 8 cases (30 km altitude) turned out to be 54 kW/m2 and 206 kW/m2, respectively. Full energy loss
into the wall per second is: Wq 4qS , where S is the total wall area of the generator. Simple
computations give the following numbers: Wq 324 kW (Mach 6) and 1.23 MW (Mach 8), which
constitutes about 3.7% 5 %, respectively, of the total enthalpy flux entering the channel.
6.

Post-MHD diffusers

The gas exiting MHD generator described in the previous section is quite rarefied, and the static
pressure is too low for engine operation. The pressure can be increased in a diffuser. We performed 2D
modeling of the gas flow in diffusers.
Gas flow in a converging nozzle with two parallel walls, with the other two walls symmetrically
converging from the inlet to the exit, was computed with a standard set of Navier-Stokes equations with
additional vibrational energy equation:
E v
E Ev0 (T )
,
(42)
+ div(Evu ) = v
t
VT (T )
where vibrational energy is defined by Eq. (29), and the vibrational relaxation time VT (T ) was
determined by Eq. (41).
The atomic oxygen fraction was put at 0.01%. Gas heating by vibrational relaxation was taken into
E Ev0 (T )
account by including source term Qv = v
in the energy equation of the Navier-Stokes set.
VT (T )
Computations were done by 2nd order MacCormack method
physical coordinates (x , z ) were transformed into (x , z ) :
x =x
,
z = 12 [1 + z / zb (x )], z [0,1]

20

on rectangular grid. For this,

(43)

where zb (x ) are the coordinates of the walls. The results presented her correspond to a grid size of 250x50.
Changing the grid size by a factor of 2 changes the results by about 1 %.
In all computed cases the diffuser geometry was the same: inlet width of 75 cm, matching the
MHD channel exit, and the diffuser exit width of 15 cm. Figs. 11 (a-d) show the results of modeling of
post-MHD diffusers for Mach 6 at 30 kilometers altitude. The flow passes through shocks and increases the
static pressure, becoming eventually subsonic. Gas temperature of the subsonic flow is about 1,150 K in
this case. As seen in Fig. 11, vibrational relaxation was incomplete at the diffuser exit. Table 2 lists
entrance and exit parameters in computed Mach 6 and Mach 8 cases (Mach 8 case is not shown in the
figures). As seen in the table, if the atomic oxygen fraction were two orders of magnitude higher, that is,
1%, the substantial concentration of atoms would have helped to relax the stored vibrational energy,
bringing the gas closer to thermal equilibrium. Fully coupled computations of vibrational relaxation,
chemical kinetics, and gas dynamics should be addressed in future work. Effects of vibrational
nonequilibrium on engine performance would also have to be assessed.
7.

Concluding Remarks

In this paper, we analyzed physical aspects of a novel concept cold-air hypersonic MHD power
generators with ionization by electron beams. The analysis demonstrated that, with careful choice of
parameters, electron beams could form stable, well-controlled plasmas, potentially enabling a good level of
generator performance. At Mach 6 8, about 1/3 of the total flow enthalpy could be extracted as electricity,
while combined heat and friction losses and power cost of ionizing beams could be kept on the level of
several percent. In Mach 6 flight, after the MHD channel and a diffuser, stagnation temperature could be
reduced to about 1,000 K, potentially enabling a turbojet operation.

15

AIAA-2001-0795
Although e-beam generated nonequilibrium ionization in MHD channels may be technically
feasible, a number of problems exist. To keep the fraction of extracted power that has to be spent on
ionization at an acceptable level, electron density and conductivity would have to be at least an order of
magnitude lower than those in conventional channels with thermal ionization of alkali seed. Because of the
low conductivity, very strong magnetic fields are required for a substantial MHD effect. In strong magnetic
fields and low-density gases, Hall parameters become extremely large, which makes operation of Faraday
channels problematic due to possible arcing between anode segments. Hall or diagonal-connection
configurations appear to be the natural choices at high Hall parameters. However, there are anode sheath
problems in Hall generators. To avoid ionization instability in the anode sheath, careful tailoring of e-beam
ionization profile and cooling the sheath would be required.
Calculations of entropy generation, total pressure losses, and thermodynamic efficiency of
hypersonic MHD generators with external ionization are yet to be performed, but it is probably that,
because of Joule heating, e-beam supported generators produce more entropy than a series of oblique
shocks.
A number of technical issues should be addressed before a final assessment of the nonequilibrium
MHD concept. Among these issues are, for example, cooling channel walls, electrodes, and e-beam passthrough windows, configuring ionization profiles, and stabilizing anode sheaths at high Hall parameters.
In this paper, only MHD generators located at the engine inlet were considered. An alternative
location would be downstream of the combustor. In principle, MHD generator could even coincide with the
scramjet combustion chamber. There, if temperature does not rise above approximately 2,500 K,
nonequilibrium ionization by electron beams could also be required.
Acknowledgements
Support for this work by the National Science Foundation, Air Force Office of Scientific
Research, and by the DDR&E Air Plasma Ramparts Multidisciplinary University Research Initiative is
gratefully acknowledged. The authors also express their thanks to R.L. Chase, S.H. Lam, V.A. Bityurin, C.
Park, and T. Brogan for valuable discussions and critical remarks.
References
1

Fraishtadt, V.L., Kuranov, A.L., and Sheikin, E.G., Use of MHD Systems in Hypersonic Aircraft,
Technical Physics, Vol. 43, 1998, p.1309.
2
Gurijanov, E.P., and Harsha, P.T., AJAX: New Directions in Hypersonic Technology, Paper AIAA-964609, 1996.
3
Macheret, S.O., Shneider, M.N., and Miles, R.B., Electron Beam Generated Plasmas in Hypersonic
MHD Channels, Paper AIAA-99-3635, June 1999.
4
Macheret, S.O., Shneider, M.N., Miles, R.B., and Lipinski, R.J., Electron Beam Generated Plasmas in
Hypersonic MHD Channels, accepted for publication in AIAA Journal, 2001.
5
Macheret, S.O., Shneider, M.N., and Miles, R.B., MHD Power Extraction From Cold Hypersonic Air
Flows with External Ionizers, AIAA Paper 99-4800, November 1999.
6
Shneider, M.N., Macheret, S.O., and Miles, R.B. Electrode Sheaths and Boundary Layers in Hypersonic
MHD Channels, Paper AIAA-99-3532, June 1999.
7
Macheret, S.O., Miles, R.B., and Nelson, G.L., Feasibility Study of a Hybrid MHD/Radiatively Driven
Facility for Hypersonic Ground Testing, Paper AIAA 97-2429, June 1997.
8
Macheret, S.O., Shneider, M.N., Miles, R.B., Lipinski, R.J., and Nelson, G.L., MHD acceleration of
supersonic air flows using electron beam enhanced conductivity, Paper AIAA-98-2922, June 1998.
9
a) Chase, R.L., Mehta, U.B., Bogdanoff, D.W., Park, C., Lawrence, S., Aftosmis, M., Macheret, S.O., and
Shneider, M.N., Comments on an MHD Energy Bypass Engine Powered Spaceliner, Paper AIAA-994965, November 1999.
b) Park, C., Mehta, U.B., and Bogdanoff, D.W., Real Gas Calculation of MHD-Bypass Scramjet
Performance, Paper AIAA-2000-3702, 2000.
10
Macheret, S.O., Shneider, M.N., and Miles, R.B., Modeling of Air Plasma Generation by Electron
Beams and High-Voltage Pulses, AIAA Paper 2000-2569, June 2000.
11
Raizer, Yu.P., Gas Discharge Physics, Springer, Berlin, 1991.

16

AIAA-2001-0795
12

Berger, M.J. and Seltzer, S.M., Tables of Energy Losses and Ranges of Electrons and Positrons, NASA
SP-3012, 1964.
13
Raizer, Yu.P., and Shneider, M.N., Simplified Kinetic Equation for Electrons in Nonuniform Fields of
Arbitrary Strength in Connection with the Cathode Sheath of a Glow Discharge, Sov. J. Plasma Phys.,
Vol. 15, 1989, pp. 184-189.
14
Gusev, N.G., Klimanov, V.A., Mashkovich, V.P., and Suvorov, A.P., Fizicheskie Osnovy Zashchity ot
Izluchenii [Physical Principles of Protection from Radiation, in Russian], Zashchita ot
Ioniziriiushchikh Izluchenii [Protection from Ionizing Radiation] Vol. 1, Energoatomizdat, Moscow
1989.
15
Bychkov, Yu.I., Korolev, Yu.D., and Mesyats, G.A., Injection Gaseous Electronics [Inzhektsionnaia
Gazovaia Elektronika, in Russian], Nauka, Moscow, 1982.
16
Hershcovitch, A., A Plasma Window for Vacuum-Atmosphere Interface and Focusing Lens of Sources
for Nonvacuum Ion Material Modification, Review of Scientific Instruments, Vol. 69, 1998, p. 868.
17
Hershcovitch, A., A Plasma Window for Transmission of Particle Beams and Radiation From Vacuum
to Atmosphere for Various Applications, Physics of Plasmas, Vol. 5, 1998, p. 2130.
18
Rosa, R.J., Magnetohydrodynamic Energy Conversion, McGraw-Hill, 1968.
19
Cowling, T.G., Magnetohydrodynamics, Interscience Publishers, 1957.
20
Anderson, D.A., Tannehill, J.C., and Pletcher, R., Computational Fluid Mechanics and Heat Transfer,
Hemisphere Publishing Corp., 1984.
21
Cebeci, T., and Smith, A.M.O., Analysis of Turbulent Boundary Layers, Academic Press, 1974.
22
Kerrebrock, J.L., Electrode Boundary Layers in Direct-Current Plasma Accelerators, J. Aerosp. Sci.,
August 1961, pp.631-644.
23
Aleksandrov, N.L., Vysikailo, F.I., Islamov, R.Sh., Kochetov, I.V., Napartovich, A.P., and Pevgov, V.G.,
Electron distribution function in 4:1 N2-O2 mixture, High Temperature, Vol. 19, 1981, p. 17.
24
Kossyi, I.A., Kostinsky, A.Yu., Matveyev, A.A., and Silakov, V.P., Kinetic scheme of the nonequilibrium discharge in nitrogen-oxygen mixtures, Plasma Sources Sci. Technol., Vol. 1, 1992, p. 207.
25
Bazelyan, E.M., and Raizer, Yu.P., Spark Discharge, CRC Press, Boca Raton, Florida, 1997.
26
Macheret, S.O., Martinelli, L., and Miles, R.B., Shock Wave Propagation and Structure in Non-Uniform
Gases and Plasmas, AIAA Paper 99-0598.

Table 2. Computed parameters of MHD channels and diffusers for Mach 6 and Mach 8 flight at 30 km
altitude. The flow was assumed to pass through an oblique shock with 100 wedge angle prior to entering the
MHD channel. MHD channel geometry: Ain = 0.25 0.25 m2, Aexit = 0.75 0.75 m2, L=3 m. Post-MHD
diffuser geometry Ain = 0.75 0.75 m2, Aexit = 0.15 0.75 m2, L=2.5 m. The numbers in the cells
correspond to Mach 6 (before the slash symbol) and Mach 8 (after the slash symbol) cases.

M=6 / 8
M
u, m/s
T, K
T v, K
p, Atm
,%
k
Wout, MW
Wbeam, MW

MHD channel
Entrance
Exit
4.647/5.76
1.556/3.01
1743/2331
815/1636
349/407
660/734.1
349/407
5122/4699
0.0433/0.0613
0.0209/0.0185
35.2/25
0.447/0.465
3.143/6.04
0.178/0.0943

17

Diffuser exit
=0.0001
0.43/0.82
310/687.8
1297/1750
3580/3925
0.2/0.55

=0.01
0.36/0.67
280.1/586.1
1520/1919
2361/2943
0.231/0.57

AIAA-2001-0795

10

10

W i, eV

10

e-beam

10

(E/N) c

0.0

2.5x10

-15

5.0x10

-15

7.5x10
.

E/N, V cm

-15

1.0x10

-14

Fig. 1. Energy cost of ionization, in eV per newly produced electron, in air as a function of E/N (the ratio of
electric field in the plasma to the gas number density). Energy cost of ionization by high-energy electron
beams (34 eV) is also indicated. The critical value E / N = ( E / N )c 1.1 1014 V cm 2 corresponds to the
threshold of electron runaway.

18

AIAA-2001-0795

Air; p=10 Torr; T=300 K; <n>=10

12

cm

-3

10000

1 =10 ns
1000

5 ns

DC discharge

P, W /cm

100

1 ns

10
e-beam

1
0.1
0

2x10

-15

4x10

-15

6x10

f, sec

-1

E/N, V*cm
10

10

10

10

10

10

-15

8x10

-15

1x10

-14

1 ns

10 ns
0

2x10

-15

4x10

-15

6x10

E/N, V*cm

-15

5 ns
8x10

-15

1x10

-14

Fig. 2. Power input (upper plot) and pulse repetition rate (lower plot) required to sustain a time-averaged
electron number density of 1012 cm-3 in air at 10 Torr, 300 K, by repetitive ideal rectangular voltage pulses
as functions of E/N (the ratio of the electric field in the pulse to the gas number density) at various pulse
lengths 1 . Power inputs required to sustain the same electron density at steady state by DC discharge and
high-energy electron beams are also indicated on the upper plot.

19

AIAA-2001-0795

jy

electrodes

Bz
windows

e-beams

flow

e-beams

Fig. 3. Schematic of an MHD channel with electron beam ionization.

20

AIAA-2001-0795

a
12

ne

10

ne

12

n +, n e, cm

-3

10

n+

n+
10

11

10

11

10

10

10

10

-0.09

3x10

-0.06
-0.03
y-y m ax , cm

0.00

-0.09

E y, V/cm

1x10

-0.06

-0.03

0.00

y-y m ax , cm
1.2x10

2x10

0.1

anode sheath: q~N

anode sheath: q~N~1/T

9.0x10

6.0x10

3.0x10

V a =430 V
0

-0.09

-0.06

-0.03

0.00

y-y m ax , cm

0.0

V a =38 V

-0.09

-0.06

-0.03

0.00

y-y m ax , cm

Fig. 4. Anode sheath structure at x=1 m from the entrance to the Hall MHD channel corresponding to Mach
6, 30-km altitude flight conditions; the transverse current density is jy=0.5 A/cm2. Figures a and b
profiles of electron and ion densities and electric field when ionization rate is proportional to gas density
(no cooling, and e-beam current density is the same as inside the channel); figures c and d profiles in the
case when ionization rate is close to that inside the channel (due to increased e-beam current density and/or
gas cooling).

21

AIAA-2001-0795

y
0
I

II

II

/ y

uB

1 - regions of external ionization

2 electrodes
3 - dielectrics
I - near-electrode charge sheath
II quasineutral plasma
Fig. 5. Schematic illustration of the growth of space charge sheath in Hall generators due to electron being
swept away into the anode between the regions of ionization.

22

AIAA-2001-0795

-2

5x10

H=10 cm
H=20 cm
H=40 cm

-2

je(H/2), A/cm

4x10

-2

3x10

-2

2x10

-2

1x10

0
0

-6

2x10

-6

4x10

-6

8x10

-6

8x10

6x10

-6

1x10

-5

-6

1x10

t, sec
3

y, cm

-6

2x10

-6

4x10

6x10

-5

t, sec
Fig. 6. Anode current (upper plot) and space charge sheath thickness (lower plot) versus time from entering
the region without e-beam ionization. The curves correspond to different width, H, of the channel.

23

AIAA-2001-0795

700

b
0.04

600
p, Atm

500

400

300
0.0

0.02
0.01

0.5

1.0

1.5

2.0

2.5

3.0

0.00
0.0

0.5

1.0

x, m

1.5

2.0

2.5

x, m

1800

0.05
d
0.04

1400

0.03

1600

, kg/m

u, m/s

1200
1000
800
0.0

3.0

4
M
3

0.02

0.01

0.5

1.0

1.5

2.0

2.5

3.0

x, m

0.00
0.0

0.5

1.0

1.5

2.0

2.5

M=u/c(T)

T, K

0.03

1
3.0

x, m

Fig. 7. Profiles of a temperature, b - static pressure, c velocity, and d Mach number and density in
Hall MHD channel. Free stream conditions: M=6; h=30 km; wedge angle =100. Inlet parameters:
T=349.26 K, p=4.39x103 N/m2, u=1742 m/s, M=4.647. Inlet and exit cross sections: Ain=0.25x0.25 m2,
Aout=0.75x0.75 m2. Magnetic field B=7 T. Electron beam current density jb(x)=5x(zin/z(x))1.65 mA/cm2.
Electron beam injection slot width 0.5 cm; spacing between the slots 0.5 cm. Loading parameter
k=0.447. Extracted power: Woutput=3.134 MW (=35.2% of inlet enthalpy flux). Power spent on ionization:
Wbeam input=178 kW.

24

AIAA-2001-0795

a
18

10

17

n e, n +, n -, m

-3

10

ne
n+
n-

10

16

10

15

0.0

0.5

1.0

1.5

2.0

2.5

3.0

x, m
10

0.5

b
Tv

0.4

7
0.2

b , keV

0.3

T v, eV

6
0.1
5
0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

x, m

Fig. 8. Plasma components densities (a) and e-beam energy and gas vibrational temperature (b) for the
same case as Fig. 7.

25

1.2x10

1.0x10

8.0x10

6.0x10

4.0x10

2.0x10

8x10

6x10

4x10

2x10

Ey

Ex

0
3.0

0.0

0.0

0.5

1.0

1.5

2.0

E x, V/m

E y=uB, V/m

AIAA-2001-0795

2.5

x, m

3x10

jx

3x10

jy

jy, A/m

4x10

2x10

2x10

1x10

1x10

0
0.0

0.5

1.0

1.5

2.0

2.5

x, m

Fig. 9. E-field components (a) and electric current densities (b) for the same case as Fig. 7.

26

0
3.0

jx, A/m

4x10

AIAA-2001-0795

40

Altitude (km)

35
30
25
20
15
10

33%, 350 kJ/kg

196 kW / 55 kW

33%, 983 kJ/kg

30%, 620 kJ/kg
0.85 MW / 35 kW 2.2 MW / 42 kW

31%, 1.7 MJ/kg

4.3 MW / 52 kW

Thermal
MHD

35%, 332 kJ/kg

35%, 659 kJ/kg

25%, 782 kJ/kg

800 / 142 kW

3.1 MW / 178 kW

6 MW / 97 kW

13%, 130 kJ/kg

1.5 MW / 173 kW
8%, 140 kJ/kg;
7 MW / 3.2 MW

16%, 282 kJ/kg

6.3 MW / 1 MW
=100

=50

23%, 1.1 MJ/kg

13 MW / 153 kW

12%, 360 kJ/kg

13 MW / 1.2 MW

w/seed

9.2%, 420 kJ/kg

22 MW / 1.3 MW

12%, 2.2 kJ/kg;

7.4 MW / 1.7 MW

Flight Mach number

10

12

Fig.6. Performance of e-beam supported MHD generators. Numbers in boxes are: generated power as a
percentage of total enthalpy flux into the channels, generated power per kilogram of air, absolute power
generated in the channel, and (after the slash symbol /) power spent on e-beam ionization. Channel
geometry: inlet 25x25 cm2, exit 50x50 cm2 (for the altitude h=40 km only) or 75x75 cm2 (all other
altitudes), length L=3 m. Magnetic field B=7 T. Upstream of the MHD channel the air passes through an
oblique shock on a wedge with the angle =100 (h>15 km), and =100 or 50 (h=15 km). The channel
configuration may be either Hall or Faraday (at altitudes h>20 km), but only Faraday one at h<20 km.

27

AIAA-2001-0795

x=0

1.4

0.2

2.7

2.0

9.2

4.6

z, m

p/p

0.4

1.0

3.7

2.4

1.7

9.6
5.8

4
3.

2.2

1.7

-0.2

-0.4
0

0.5

2.5

u, m/s

0.4

8.1

.4

.5

452

2
2.

647

1
2.
65

58

27 9

679.5

311.9

52

0 814.9

5.4

2.1

62

65

71

1.9

0.2

z, m

1.5

x, m

.0

-0.2

-0.4
0

0.5

x, m

1.5

2.5

Fig. 11a, b. Static pressure (a) and gas velocity (b) contours in post-MHD diffuser in the Mach 6, 30-km
altitude flight conditions. Atomic oxygen mole fraction is 10-4.

28

AIAA-2001-0795

T, K

.7

757.7

840.4

1 07

655.0

965.7
997

3.3
86

769

1279.0

1169.3

.0

.2

z, m

92

5.4

8.4

82
4

.1

7 46

7 15

0.2

.4

0.4

-0.2

-0.4
0

0.5

1.5

x, m

2.5

Tv , K

0.4

0.2
4806.7

4645.3

4968.0

z, m

4566.1
4902.5

4322.7 3838.7
4484.0
4161.4

4733.0

3516

4566.1

-0.2

-0.4
0

0.5

x, m

1.5

2.5

Fig. 11c, d. Translational (c) and vibrational (d) temperature contours in post-MHD diffuser in the Mach 6,
30-km altitude flight conditions. Atomic oxygen mole fraction is 10-4.

29