Azimuthal Current Author Submitted

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 29

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/354829641

Compositions and distributions of the azimuthal currents in the magnetic


nozzle

Article  in  Plasma Sources Science and Technology · September 2021


DOI: 10.1088/1361-6595/ac2a0a

CITATIONS READS
6 73

7 authors, including:

Zhiyuan Chen Yibai Wang


Beihang University (BUAA) Beihang University (BUAA)
14 PUBLICATIONS   80 CITATIONS    29 PUBLICATIONS   128 CITATIONS   

SEE PROFILE SEE PROFILE

Hai-Bin Tang Junxue Ren


Beihang University (BUAA) University of Southern California
140 PUBLICATIONS   682 CITATIONS    36 PUBLICATIONS   271 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Applied-Field Magnetoplasmadynamic Thruster Simulation View project

Ion-driven magnetic nozzle investigation via PIC simulation View project

All content following this page was uploaded by Zhiyuan Chen on 12 October 2022.

The user has requested enhancement of the downloaded file.


Compositions and Distributions of the Azimuthal
Currents in the Magnetic Nozzle
Zhiyuan Chen1 , Yibai Wang1 , Haibin Tang2,3,4,† , Junxue Ren1 ,
Min Li5 , Peng Wu1 and Jinbin Cao2,4
1
School of Astronautics, Beihang University, Beijing 102206, China
2
School of Space and Environment, Beihang University, Beijing 102206, China
3
Key Laboratory of Spacecraft Design Optimization and Dynamic Simulation
Technologies of the Ministry of Education, Beijing 102206, China
4
Laboratory of Space Environment Monitoring and Information Processing, Ministry
of Industry and Information Technology, Beijing 102206, China
5
Shanghai Engineering Centre for Microsatellites, Chinese Academy of Sciences,
Shanghai, 201203, China
E-mail: thb@buaa.edu.cn†

Revised in July and resubmitted in September 2021

Abstract. The azimuthal currents in the divergent magnetic nozzle of the electrode-
less plasma thrusters are investigated using a two-dimensional axisymmetric particle-
in-cell (PIC) code. In this paper, the azimuthal currents are decomposed into four
components – the diamagnetic drift current, the E × B drift current, the viscous-
stress-induced current and the inertia-induced current. The azimuthal current and its
four components are analyzed under three magnetization levels of 0.25 T, 0.75 T and
2.15 T. In the absence of inlet ion temperature, the azimuthal currents mainly consist
of electron currents resulted from ∇pe , E × B drift and electron viscosity, while the
azimuthal currents caused by the inertia can generally be neglected. The azimuthal ion
currents considered negligible in previous studies are shown to be non-negligible in the
highly-magnetized and collisionless magnetic nozzle, where the dominant mechanism
for its formation is E × B drift and drift from the centrifugal force. In the upstream
and midstream, the compositions and distributions of the azimuthal currents can vary
at different radial positions and magnetization conditions because of the influence of
the electric potential barrier and the high density conic. In the downstream, a vast
range of paramagnetic currents are produced by the E × B drift due to ion inward
detachment. And the paramagnetic stress-induced current begins to prevail due to
the FELR effect. Superimposed with the paramagnetic part of the diamagnetic drift
current under a high magnetic field, they can undermine the thrust gain coefficient of
the magnetic nozzle.

Keywords: magnetic nozzle, plasma currents, thrust augmentation, finite electron Lar-
mor radius(FELR) effect, helicon sources.

Submitted to: Plasma Sources Sci. Technol.


Compositions and Distributions of the Azimuthal Currents in the MN 2

1. Introduction

The magnetic nozzle (MN) is usually a convergent-divergent magnetic field that transfers
the non-directed kinetic energy of the plasma into the directed one through complex
electromagnetic mechanisms[1]. Incorporating plasma-magnetic field interactions, the
magnetic nozzle constrains and accelerates the plasma flow while minimizing physical
contacts that may cause damage to the thrusters or spacecraft itself[2]. Nowadays,
magnetic nozzles are integrated as intrinsic parts of many electrodeless plasma thrusters
in the hope of improving propulsive performances, e.g., helicon plasma thruster[3], ECR
thruster[4] and VASIMR[5]. In the magnetic nozzle of electrodeless plasma thrusters,
multiple processes serve to accelerate the ions, such as the ambipolar acceleration[6, 7]
and, under specific circumstances, a current-free double layer[8]. While these two ion
acceleration mechanisms dominate in the electron-driven magnetic nozzles, Z. Chen et
al .[9] found that the ion gasdynamic acceleration may become the primary acceleration
mechanism in the MN where the ion temperature cannot be neglected.
Recently, many pieces of research focus on the induced magnetic fields and plasma
currents in the magnetic nozzles of electrodeless plasma thrusters, especially helicon
plasma thrusters. These studies often focus on the effects brought by the induced
magnetic field or the directions and compositions of the induced currents. The induced
field and currents are of research interests mainly because the diamagnetism of plasma
flow is the foundation of electromagnetic acceleration in the MN[10]. Moreover, a
magnetohydrodynamic (MHD) detachment model[11] also requires a paramagnetic
plasma to stretch the magnetic field lines to infinity to facilitate the particle detachment
for thrust generation.
Despite that Arefiev and Breizmann[11] and Hooper[12] have demonstrated that
the cold plasma flow in a MN is paramagnetic, Ahedo and Merino[13] pointed out that
a MN with a cold plasma inlet has no propulsive meaning. Furthermore, in 2016, they
investigated the induced magnetic field in the MN of a moderate to high β parameter
(0.01 ∼ 0.1) using a 2D two-fluid code[14], which claims that the induced field are all
against the applied one in their simulation domain of 20-inlet-radius long and predicts
the existence of a magnetic separatrix, despite that their magnetized electron model
breaks down near the separatrix. In terms of experimental measurements, Corr and
Boswell[15] investigated the induced magnetic field generated in the high-beta plasma
flow in the magnetic nozzle of a helicon plasma source, the plasma turned out to be
diamagnetic although the extent is rather small – only 2% decrease of the original field
was observed. In 2011, Roberson et al.[16] measured a significantly higher perturbed
field than previously reported, peaked in 15 G in strength. And the background B was
completed expelled in a large region along the axis downstream. And their estimated
induced azimuthal currents are all diamagnetic in their measurement region. While
these experimental results confirm the simulated ones of Merino et al., Takahashi et
al.[17] first observed a transition of the induced field from diamagnetic to paramagnetic
in the laboratory.
Compositions and Distributions of the Azimuthal Currents in the MN 3

In terms of thrust generation, Fruchtman[18] has argued that the spontaneous


potential drop does not contribute net axial momentum to the plasma flow, while its
actual function is to convert the electron pressure to the directed kinetic energy of ions.
This argument has been confirmed analytically and experimentally by Lafleur et al.[19].
Moreover, it has been supposed that the acceleration is caused by the azimuthal current
and the magnetic field[17]. Therefore, investigating the induced currents, especially the
azimuthal ones, can help us better understand the spatial distribution and mechanisms
behind the electromagnetic accelerations in the MN.
Based on a cold-plasma Lagrangian model, Hooper’s study[12] claims the azimuthal
currents are all paramagnetic, being consistent with the conclusions made by Arefiev
and Breizmann[11]. In contrast to their cold-plasma models, Little et al.[20] adapted
Hooper’s model to include electron pressure but followed the same calculation method
of azimuthal velocity uθ , excluding the effects of electron pressure on the azimuthal
currents. Therefore, they all ended up in the all-paramagnetic scenario. Whereas,
as mentioned above, Ahedo et al. and Roberson et al. showed that the azimuthal
currents in the MN of a helicon source are diamagnetic using two-fluid simulation
and experiments respectively. However, the situation seems more complicated
when considering the recently observed plasma-flow-state transition from diverging to
stretching a magnetic nozzle[21], as this indicates the direction of azimuthal currents is
not spatially uniform.
Recently, Takahashi et al.[17] measured the azimuthal currents jθ in the MN of
a 1 kW helicon source. They derived the expression for jθ from a simplified electron
momentum equation, in which the inertia term is omitted and the electron pressure
tensor is assumed to be diagonal and isotropic. And they found that both diamagnetic
and paramagnetic currents co-exist in the MN and that jθ mainly consists of two parts:
a) the diamagnetic drift (∇pe ) current and b) the E × B drift current of electrons. And
the diamagnetic drift current can be either diamagnetic or paramagnetic because of the
high density conic structure in the helicon MN[22], while the E × B drift current can
only be paramagnetic. The proportion between the two drift currents varies with the
magnetic field strength B0 such that the E × B drift current decreases as B0 intensifies,
for which they claim that the perpendicular electric field Er will vanish with weakened
charge separation as a result of the increasingly magnetized ions. As for the ion currents,
the measurement shows they are negligible compared to the electron currents. However,
as the authors admit in their paper, the ion magnetization was not enough and the
background pressure was not low enough to avoid ion-neutral collisions, it is unclear
which factor actually caused the ion azimuthal currents to be negligible. Thus the
effects of magnetic field strength require further studies. Moreover, Hu et al . [23]
investigated the thrust generation mechanism in the magnetic nozzle using a full-PIC
model. They found that the further downstream, the greater the error of the current
predicted by the simplified fluid model is. And the usually-neglected current arised from
the stress term will result in a more paramagnetic azimuthal current.
To summarize, some questions still remain unclear about the azimuthal currents:
Compositions and Distributions of the Azimuthal Currents in the MN 4

r Open Boundary

Open Boundary
𝜕𝜕ɸ/𝜕𝜕r=0

𝜕𝜕ɸ/𝜕𝜕z=0

Open Boundary
𝜕𝜕ɸ/𝜕𝜕z=0
Coils
Rcoil=4.2Rp0

Rp0

Electron
Ion
Emission Plane Symmetric Axis z

Figure 1. Schematics of the computational domain and boundary settings.

what are the distributions and driving mechanisms of azimuthal currents in the MN,
do they vary with axial locations? Can the ion currents be considered negligible under
any circumstances? Is the simplified electron momentum equation precious enough to
analyze the azimuthal currents?
This paper extends the previous planar full-PIC model[9] to a two-dimensional ax-
isymmetric one to study the azimuthal currents in the magnetic nozzle. And the aim
is to study the effects of magnetic field strength on the profiles and compositions of
the azimuthal currents in different axial regions. Two generally overlooked components,
the azimuthal currents resulted from viscous stress and inertia, have been added to the
analysis along with the diamagnetic drift current and E × B drift one. The simulations
are collisionless, and the magnetic field is extended to include a higher range where
the ions can be magnetized, so that the validity of the negligible ion current can be
confirmed. The rest of this paper will be organized as follows. Section 2 describes the
simulation model used in this paper. Section 3 presents and discusses the fully kinetic
results of numerical investigations of the azimuthal currents. A summary and conclu-
sions are presented in Section 4.

2. Methodology

2.1. Simulation setups and boundary conditions


To investigate the azimuthal currents in the magnetic nozzle, the fully-kinetic two-
dimensional planar PIC model reported in our previous work[9] has been extended to a
2D axisymmetric one to include the cylindrical geometric effects.
In this study, the plasma is assumed to be fully ionized and collisionless, as the
characteristic length of the system, the radius of magnetic nozzle throat, is orders of
magnitude smaller than the mean free paths of the related collisions. The computational
Compositions and Distributions of the Azimuthal Currents in the MN 5

setups and the boundary conditions shown in figure 1 are similar to what has been
described in [9], which will be described below.
The simulation model used in this paper is based on electrostatic full-PIC method.
In this model, the ions and electrons are modeled as macroparticles. The motions of the
charged particles are calculated by solving the Newton-Lorentz equations below using
the standard leapfrog Boris scheme[24, 25]. And the electric field are calculated self-
consistently from the Poisson equation below with the dynamic alternating direction
implicit (DADI) method. The macroscopic properties, e.g., the number density and the
macroscopic velocity, are calculated on the mesh points using the Cloud-in-Cell (CIC)
scheme.[26]
dv
ms = qs (E + v × B) (1)
dt
∇ · (ε0 ∇φ) = qs (ni − ne ) (2)
The boundary conditions of this study is similar to that in the previous study[9],
which are summarized below. As shown in figure 1, the computational domain is
initialized as a vacuum cylinder at the beginning of the simulation, which has a length of
400 cells and a width of 240 cells with a grid resolution of one Debye length λD0 which
allows for stable simulations. The lower-left boundary stands for the plasma inlet of
MN with a radius Rp0 = 10λD0 . The bottom boundary is the symmetric axis where all
particles will be specularly reflected. Other boundaries are all open boundaries (OBs),
representing the infinite vacuum environment. In a typical simulation case, the time
step resolution is ∆tωpe = 0.1.
As shown in figure 1, the simulation domain is shrunken to a few hundred Debye
lengths as Rao and Singh[27] and Li et al.[28] did to reduce the computational cost to
a practical level. In order to keep the main physical processes intact, a dimensionless
number, representing the relative magnetization of ions in a magnetic nozzle system
of a specific size, is kept unchanged before and after the scaling. The dimensionless
number is the normalized Larmor radius of ions R̂Li0 which is defined by the ratio of ion
Larmor radius and the plasma inlet radius mqiiB vti0
0
/Rp0 . In the expression, qi denotes ion
charge, B0 is the magnetic inductive strength at the inlet center, Rp0 is the inlet radius
of magnetic nozzle, mi is the mass of ions, and vti0 is the thermal velocity of the ions.
From the definition of R̂Li0 , it is clear that the increment of B0 is necessary for keeping
the physical processes intact after the scaling down (Rp0 decreases). This is the reason
why the studies are conducted under seemingly strong magnetic field like 0.25 T and
0.75 T in Section 3 while actually within the range of interests.
At each time step, macro-particles representing ions and electrons are emitted from
the emission plane (i.e., the plasma inlet) into the domain. The non-axial velocity
components of both ions and electrons are sampled from a stationary Maxwellian
distribution given by equation (3). Whereas, in the axial direction, the electrons are
sampled from a Maxwellian flux distribution due to the free molecular effusion effects[29]
Compositions and Distributions of the Azimuthal Currents in the MN 6

given by equation (4),


r  
me me 2
fe,i (vr,θ ) = exp − v , (3)
2πkB Te0 2kB Te0 r,θ
 
me me 2
fe (vz ) = vez exp − v , (4)
kB Te0 2kB Te0 ez
while the ions are assumed to be a cold beam. The initial temperature ratio is set to
Ti0 : Te0 = 0.01 and the macroscopic beam velocity vbeam is set to Bohm velocity of the
ions. Additionally, the real mass ratio of proton to electron mi /me = 1836 is used.
In terms of the inlet boundary conditions, for the simulation cases shown in
Section 3, the number density and electron temperature at the inlet center are set
to n0 = 1 × 1018 m−3 and Te0 = 5 eV respectively, referenced from the experimental
results of the helicon plasma thruster[30]. However, this study does not involve the
simulation of the plasma source, thus a uniform distribution of electron temperature is
assumed at the inlet, and the density profile is set as Gaussian profile.
r2
 
n0 (r) = n0 exp − 2 , (5)


where the parameter σ = Rp0 /(2 ln 2) if the plume radius is assumed as Half Width at
Half-Maximum (HWHM). Moreover, the electric potential at the inlet center is set to
0 V as a Dirichlet boundary condition. For particles crossing the inlet boundary from
inside the domain, the absorption condition for both ions and electrons is applied.
In terms of other boundaries, the symmetric axis and the open boundaries share
the same homogeneous Neumann boundary conditions for the electric potential as
shown in equation (6). Whereas, they differ in handling the crossing particles. At
the axisymmetric boundary, according to the symmetry, all the particles are specularly
reflected. But at the open boundaries (OBs), the ions are simply deleted, while the
electrons are selectively reflected based on an energy criterion according to the global
current-free condition. This will be explained in detail below.
 
∂φ
=0 (6)
∂n b
As stated in our previous work[28, 31], we think quasi-neutrality and current-free
condition should both be conserved globally within the electrodeless plasma sources.
However, considering the counter streaming of electrons[6], it is difficult to guarantee
these two conditions by the settings of the inlet boundary alone, thus they are guaranteed
by both the inlet and open boundary settings:

(i) Quasi-neutrally ni0 = ne0 and current-free condition ni0 ui0 = ne0 ue0 is set at the
inlet boundary, where the ui0 and ue0 indicate the initial macroscopic velocity of
ions and electrons.
(ii) At all three open boundaries, within each time step, if the number of crossing ions
and electrons are Ni and Ne respectively (generally Ni < Ne ). Reflect specularly
(Ne − Ni ) electrons with the lowest energy.
Compositions and Distributions of the Azimuthal Currents in the MN 7

Table 1. Simulation parameters in the verification case. (For verification case only)
Parameters(Unit) Value
B0 (G) 1000
Rl /R 5.4
n0 (m−3 ) 7 × 1018
Te0 (eV ) 20
Ω̂i0 ≡ RΩi0 /cs 0.1

Therefore, the flux of electrons and ions crossing each open boundaries will be equal,
thus the current-free condition will be forced at three OBs and the symmetric axis. At
last, due to current conservation, the remaining inlet boundary will be forced to obey
the current-free condition when the simulation stabilizes.
Another reason to set this selectively reflective electron OB is to prevent the plasma
pump instability[32, 33]. As previously reported, the electrons in the MN will bounce
axially back and forth under the effects of the ambipolar electric field. Before the leading
edge of the ions leaves the open boundary, the ambipolar electric field can be generated
self-consistently and confine the electron flux to be equal to that of ions by effectively
reflect the lower-energy electrons back. But once the leading edge leaves the open
boundary, the plasma system in the domain loses the ability to confine the electron flux
through the ambipolar field. Therefore we added this type of energy-based selectively
reflective open boundary to mimic the ability to confine the electron flux, without which
the simulations will never reach steady state. More details and verification of this type
of open boundary can be found in [9] and [28].
Note that, despite our best possible efforts to make the open boundary simulate
an infinite vacuum, we still cannot completely exclude the influence of this artificial
boundary. To be more rigorous, only the data in the lower left quarter region of the
computational domain (200λD0 × 120λD0 ) are used for the discussion and analysis in
this paper.

2.2. Benchmark Verification


This code was developed based on a home-developed PIC-MCC simulation package
which has been successfully applied to simulations of plasma thruster plume[28], Hall-
effect thrusters[34], hollow cathodes[35], and MPD thrusters[36, 37].
For verification, the results of the classical two-fluid code DIMAGNO by Merino et
al.[38] are still used as a benchmark. The non-uniform inlet case has been used and the
plasma parameters at the nozzle inlet and the magnetic field settings are presented in
Table 1.
However, as Singh et al.[39, 27] previously found out, due to the radially-outward
electric field, some of the ions are radially accelerated and overshoot the Most Divergent
Magnetic field Line(MDML, the magnetic field line that intersects (x, z) = (Rp0 , 0))
Compositions and Distributions of the Azimuthal Currents in the MN 8

near the inlet, which results in the radial flux loss of plasma that cannot be reproduced
by the two-fluid code due to the limitations of its numerical method. Therefore, near
the nozzle inlet, this radial flux loss causes steeper decreasing profiles of plasma density
and electric potential along the axis compared to that of the two-fluid code, resulting
in the inconsistency between them. Additionally, as the ambipolar electric field is one
of the primary ion acceleration processes, the steeper descend of potential also leads to
a faster acceleration of ions near the inlet, causing the profile of ion velocity different
from that of the two-fluid code. Although we believe that allowing ions to overshoot
the MDML is more consistent with the physical reality, we managed to develop a post-
processing method to make this comparison meaningful and consistent by putting the
overshot particles back to the mainstream (Please refer to Appendix A for more details).
Utilizing this method, the axial profiles of plasma density ln n̂i (ẑ) and electric potential
φ̂(ẑ) can be converted to ones that are forced to obey the restrictions of not overshooting
the MDML as is compulsive in the two-fluid code.
Figure 2 shows the comparisions of the profiles of normalized number density,
electric potential and ion axial velocity between our PIC code and the two-fluid
DIMAGNO code. The converted PIC results fit well with that of the DIMAGNO code,
thus it is concluded that, despite many fundamental differences, our PIC model can be
roughly converted to the two-fluid code in this verification case under the assumptions
of not overshooting the MDML. Therefore the PIC code can accurately describe the
evolution of plasma in the magnetic nozzle and can be used for further research work.
Compositions and Distributions of the Azimuthal Currents in the MN 9

Norm. Num. Density


0 DIMAGNO
PIC
−2 PIC_converted
(a)
−4

−6
0 5 10 15 20
Norm. Elec. Potential

0 DIMAGNO
−1 PIC
−2 PIC _converted
−3 (b)
−4
−5
−6
0 5 10 15 20
Norm. Ion Axial Vel.

3.5 DIMAGNO
3 PIC
PIC_converted
2.5
(c)
2
1.5
1
0 5 10 15 20
z

Figure 2. Comparisons of the simulation results between the PIC code and the
DIMAGNO code; (a)comparison of number density profile (in natural logarithm). (b)
comparison of electric potential profile. (c) Comparison of ion axial velocity profile.

3. Results and discussions

Figure 3 shows the contours of various plasma properties in a typical case, including the
electron number density, electron pressure, electric potential and the total azimuthal
current. Note that the values in the figure have been normalized, x, z are normalized
by the inlet radius Rp0 , velocities are normalized by electron thermal velocity vte0 , the
electric potential φ and Te , Ti are normalized by the inlet electron temperature Te0 , the
number density are normalized by the electron number density ne0 at the inlet. The ”0”
in the subscript indicates inlet parameters. In order to avoid the influence of numerical
noise, the region where the plasma number density are less than 3 × 10−4 are set to the
minimum value in the contour plots of figure 3(a). As is seen in figure 3(a), (b) and
(c), in the axial direction, the plasma properties propagate monotonically downstream
under the dominance of a self-consistently developed ambipolar electric field. However,
in the radial direction, some interesting structures have formed. As we can see in figure
3(b), a dense region appeared on the edge of the magnetic nozzle, which was identified
and named high denstiy conic by C. Charles[22]. Moreover, in figure 3(c), an electric
potential barrier appears in the nozzle edge near the inlet, which corresponds to the
electric potential side lobe discovered in the experiments of Saha et al .[40]. It will be
Compositions and Distributions of the Azimuthal Currents in the MN 10

shown later that these plasma structures can significantly affect the distributions of the
azimuthal currents. The formation mechanisms of them and their interactions with the
azimuthal currents will be described below where needed.
In this paper, the results indicate that the driving mechanisms and compositions
of the azimuthal currents are various among different axial regions. Therefore, the
magnetic nozzle has been artificially divided into three parts in order to analyze and
express the results more clearly:
(i) the upstream: the region locates immediately downstream of the inlet with no sign
of the high density conic structure[22];
(ii) the midstream: the region locates further from the inlet with prominent high density
conic structure, while the ions have not detached from the magnetic field;
(iii) the downstream: the region lies far downstream from the inlet with ions beginning
to detach evidently from the diverging magnetic field.
In this study, the axial positions where the plasma profiles satisfy the descriptions
above are determined based on a simulation case such that the B0 = 0.75T and Ti0 ∼ 0,
which corresponds to the experimental conditions of 710 G (or IB = 14 A) in the
experimental research of Takahashi et al.[17]. Therefore, in all the cases discussed in
this paper, the upstream, midstream and downstream refer to z = Rp0 , z = 2.5Rp0 and
z = 6Rp0 respectively.
The total azimuthal current can be directly computed by jθ = ene (ui,θ − ue,θ ),
where ui,θ and ue,θ stand for the azimuthal velocity of ions and electrons respectively. To
investigate the detailed driving mechanisms behind the azimuthal currents, the currents
caused by four contributors are calculated by the following equation with the plasma
macroscopic properties obtained from the PIC simulations.
Taking the electron azimuthal current as an example, the expression for the electron
azimuthal current can be calculated using the full electron momentum equation, i.e.,
equation (7). pe is the scalar electron
 pressure. τ e is the viscous stress tensor which can
be expressed by τ e = τe,jk = − ρ(ve,j − ue,j )(ve,k − ue,k ) − δjk pe , in which ve,j and ue,j
stand for the electron total velocity and macroscopic velocity respectively in j direction,

and δjk denotes the Kronecker delta[41]. When the simulation is stablized( ∂t = 0), the
electron azimuthal current can be obtained from the radial component of equation (7),
of which the detailed derivation can be found in Ref.[23]. As shown in equation (8),
the azimuthal current consists of four components, i.e., the E × B drift current, the
diamagnetic drift current caused by ∇p, the azimuthal currents caused by inertia and
stress effects.

(me ne ue ) + ∇ · (me ne ue ue ) = −∇pe − ene (E + ue × B) + ∇ · τ e , (7)
∂t
Er 1 ∂pe
jθe = −ene ue,θ = ene +
Bz B ∂r
| {z } | z{z }
E×B drif t Diamagnetic drif t
Compositions and Distributions of the Azimuthal Currents in the MN 11

10.0607096 10.0603081
10 00..31143590 10 00..21509598
00..0063609942 00..0041370595
5 00..00104625894 5 00..000072194059
00..000031044057 04..0802151E8-54
63..409060EE--44 18..906050EE--45
00 5 10 15 20 00 5 10 15 20
(a ) e le c tro n d e n s ity (b ) e le c tro n p re s s u re

0-.00.060100 004...00000063E02-004
10 --11..282300 10 ---000...000000258420000
--23..404500 4 .0 0 E - 4
---000...000111036864
5 --34..626700 5 ---000...000122924208
4 .0 0 E - 4

--45..848900 ---000...000233703642
--00..00336808
0 .0 0 3 2 0

00 5 10 15 20 -6.100 00 5 10 15 20
- 0 .0 3 0 4
4 .0 0 E - 4 4 .0 0 E - 4

(c ) e le c tric p o te n tia l (d ) a z im u th a l c u rre n t

Figure 3. Contours of plasma properties under 0.75 T. Values of properties in the


figure have all been normalized.

 
1 ∂Me,rr ∂Me,rz Me,rr − Me,θθ
+ + + (8)
B ∂r ∂z r
|z {z }
Inertia term
 
1 ∂τe,rr ∂τe,rz τe,rr − τe,θθ
− + + .
Bz ∂r ∂z r
| {z }
Stress term

In equation (8), τe,jk denotes the (j, k) component of the electron stress tensor.
Me,jk stands for me ne ue,j ue,k , the subscripts j, k in the cylindrical coordinate system can
be r, θ or z. Although derived for electrons, since there is no assumption for electrons
specifically, equation (8) can still be used to calculate the ion azimuthal currents after
replacing the corresponding terms with the ion ones.
In the following sections, the magnetic inductive strength at the inlet center B0
are set to 0.25 T, 0.75 T and 2.15 T respectively. Keeping the dimensionless number
R̂Li0 unchanged, 0.25 T and 0.75 T corresponds to 250 G and 710 G in Takahashi’s
experimental research[17], while 2.15 T enters the highly magnetized regime such that
even the ions are magnetized at the inlet, with a Larmor radius of rL,i0 = 0.06Rp0  Rp0 .
Compositions and Distributions of the Azimuthal Currents in the MN 12

3.1. Azimuthal currents in the upstream and the ion component


3.1.1. Azimuthal currents in the upstream. Figure 4 illustrates the radial profiles of
azimuthal currents in the upstream (along ẑ = 1), in which the four contributors were
calculated using equation (8) with plasma parameters obtained from the corresponding
simulation cases. It should be noted that in all the plots of azimuthal currents, the
positive value means a paramagnetic current, which means that the current generates
a magnetic field in the same direction as the applied one. And vice versa. In the
meantime, the radial position of the MDML has been marked as a vertical line on most
line charts as a reference.
As the dominating component, the electron azimuthal currents should be
investigated first. As shown in figure 4, under all the magnetic strength levels, the
sums of all four contributors of electron currents ĵθe,calc fit fairly well with the directly-
calculated total currents ĵθ . As the magnetic field strengthens from 0.25 T to 2.15 T,
the magnitude of both E × B drift currents ĵE×Be and the diamagnetic drift current
ĵDe will decrease, so will the total azimuthal currents. The maximum of diamagnetic
drift currents decreases from -0.040 to -0.006, while the E × B drift current drops
dramatically from 0.020 to 0.002. The reason behind this trend is twofold: a) for
both ĵE×Be and ĵDe , the magnetic field B appears in the denominator of all four terms
of equation (8), resulting in smaller currents under stronger B0 ; b) for the E × B drift
currents only, as Takahashi et al.[17] proposed, the higher magnetic field magnetizes the
ions, undermines the charge separation tendency, resulting in a much smaller electric
field. Therefore, ĵE×Be decreases even further.
In terms of E × B drift currents, as mentioned in the literature[17], under all
three magnetic strengths, they are paramagnetic over the upstream due to the outward
radial electric field, as shown in figure 5(a). As for the diamagnetic drift currents,
the diamagnetic drift currents ĵDe are diamagnetic throughout the upstream, different
from Takahashi et al.’s conclusion[17] indicating that the ĵDe flows in the paramagnetic
direction in the inner radial region while flows in the diamagnetic direction in the outer
radial region. The reason for this difference lies in the profiles of electron pressure: as
shown in figure 5(b) and figure 3(b), there is no electron pressure peak off axis formed
in the upstream region regardless of the magnetic field strength. Instead, the electron
pressure decreases monotonically in the radial direction resulting in an all-diamagnetic
profile of ĵDe . This radial monotonic trend of density before the appearance of a
prominent high density conic at the upstream reflects the experimental results of Saha
et al .[42].
About the electron azimuthal current caused by the stress effect ĵstress,e , under
0.25T and 0.75T, it is of the same order with the currents resulted from E × B drift
and the diamagnetic drift near the MDML. And it is mostly paramagnetic, which means
this component affects negatively on the propulsive peformance of the magnetic nozzle.
When B0 strengthens to 2.15T, ĵstress,e is almost negligible. On the component of inertia
induced current ĵinertia,e , the magnitude is rather negligible with a small increase outside
Compositions and Distributions of the Azimuthal Currents in the MN 13
0.03

0.02 MDML j
θ
0.01 jθe,calc

0.00 j
 E×Be
j


-0.01  De
 -0.02 jinertia,e

jstress,e
-0.03 
jθe
-0.04
0.5 1.0 1.5 2.0 2.5 3.0
r
(a) 0.25T
0.03

0.02 MDML j
θ
0.01 jθe,calc

0.00 j
 E×Be
j

-0.01  De

-0.02 jinertia,e

jstress,e
-0.03 
jθe
-0.04
0.5 1.0 1.5 2.0 2.5 3.0
r
(b) 0.75T
0.02

j
θ
0.01 MDML jθe,calc

j
 E×Be
0.00 j

 De

jinertia,e

-0.01 jstress,e

jθe
-0.02
0.5 1.0 1.5 2.0 2.5 3.0
r
(c) 2.15T

Figure 4. Azimuthal currents at the upstream under different magnetic strength:


directly-calculated total current ĵθ , electron current ĵθe , the four contributors of
electron currents ĵE×Be , ĵDe , ĵinertia,e , ĵstress,e and their summation ĵθe,calc .

of the MDML (about r̂ = 1.3 ∼ 1.4). The negligible magnitude of ĵinertia,e is supposed
to due to the small mass of the electron. Therefore, the sum of ĵinertia,e and ĵstress,e
is mostly paramagnetic. Most fluid models of MN tend to omit the inertia term of
the electron momentum equation and assume an diagonal electron pressure without
considering the stress tensor. This is the reason why the simplified fluid models tend to
Compositions and Distributions of the Azimuthal Currents in the MN 14

-1

-2
ϕ
 0.25 T
0.75 T
-3 2.15 T

-4
0.0 0.5 1.0 1.5 2.0 2.5

r
(a) Radial profiles of φ̂

0.4

0.3

0.2
pe

0.25 T
0.75 T
0.1 2.15 T

0.0
0.0 0.5 1.0 1.5 2.0 2.5

r
(b) Radial profiles of p̂e

Figure 5. Radial profiles of electric potential φ̂ and electron pressure p̂e at the
upstream

predict azimuthal currents less paramagnetic than they actually are[17].

3.1.2. The ion azimuthal currents and the non-negligible inertia effect. Comparing the
total azimuthal currents ĵθ and electron azimuthal currents ĵθe in figure 4(a) and figure
4(b), it can be found that under low or medium magnetic field strength (≤0.75T), the
two plots still coincide with each other. However, as figure 4(c) shows, when B0 reaches
2.15 T, the profile of electron azimuthal current ĵθe distinctly detaches from that of the
total current ĵθ , indicating the ion component is too great to be neglected.
To study the driving mechanisms of the ion azimuthal current, similar to the above
analysis of electron currents, the components of ion current resulted from the four
mechanisms, ĵE×Bi , ĵDi , ĵinertia,i , ĵstress,i and their summation ĵθi,calc , are calculated using
the ion parameters obtained from the 2.15 T case. Together with directly simulated ion
and electron azimuthal currents ĵθi , ĵθe , their radial profiles are illustrated in figure 6(a).
As the figure shows, the summed currents of the four components ĵθi,calc matches well
Compositions and Distributions of the Azimuthal Currents in the MN 15

0.002 MDML 
j
0.001  θe
j
 θi
0.000 j
 E×Bi
j


 Di

-0.001
jinertia,i
-0.002 
jstress,i

-0.003 jθi,calc

-0.004
0.5 1.0 1.5 2.0 2.5 3.0
r
(a) Ion azimuthal current and its components under 2.15 T: directly
simulated ion total current ĵθi , the four contributors ĵE×Bi , ĵDi , ĵinertia,i ,
ĵstress,i and their summation ĵθi,calc . The directly-simulated electron total
current ĵθe is also plotted for comparison.
0.004
0.003 MDML

0.002 
jinertia,i
0.001 1 ∂Mi,rr
Bz ∂r

0.000 1 ∂Mi,rz

Bz ∂z
-0.001 1 Mi,rr -Mi,θθ
Bz r
-0.002
-0.003
0.5 1.0 1.5 2.0 2.5 3.0
r
(b) The respective contributions of the three parts of inertia term

Figure 6. Ion azimuthal currents at the upstream under high magnetic field and the
decomposition of the inertia term.

with the directly-simulated ion currents ĵθi . Within the most divergent magnetic field
line (MDML), the absolute value of directly simulated ion total current is comparable
to that of electron current. Among the four components of ion azimuthal currents, only
E × B drift current ĵE×Bi and inertia term ĵinertia,i have evident non-zero values. The
E × B drift current is diamagnetic, while the inertia term is paramagnetic, but the total
ion current is diamagnetic thanks to the larger magnitude of ĵE×Bi . The absence of the
diamagnetic drift current and the stress-induced currents can be attributed to relatively
low ion temperature.
Since the inertia term cannot be neglected for the ion current, the specific
mechanism of inertia term should be determined. Refer to the inertia term of equation
(8), the ion version of this term is the same after replacing the Me components with
corresponding Mi ones. Plotting each of the three parts of the inertia term on figure
6(b), it can be seen that the first two parts cancel each other out and total inertia term
M −M u2 −u2
overlaps with the third part B1z i,rr r i,θθ = B1z ρ r r θ which is, as expected, the drift
current generated by the centrifugal force.
Compositions and Distributions of the Azimuthal Currents in the MN 16

Therefore, it can be concluded that the ion azimuthal current cannot be neglected
under a high magnetic field such that the ions start to be magnetized. The ion azimuthal
current is mainly composed of diamagnetic E × B drift current and a paramagnetic drift
current caused by the centrifugal force.

3.2. Azimuthal currents in the midstream


Figure 7 shows the radial distribution of azimuthal currents in the midstream
under different magnetic field strength. Under 0.25 T, the total azimuthal current
ĵθ shows an all-diamagnetic profile, while under 0.75 T and 2.15 T, the diamagnetic
and paramagnetic parts of ĵθ coexist, like the profile that Takahashi et al. measured,
the paramagnetic parts emerge inside the MDML, while the diamagnetic parts mainly
locate outside the MDML. As can be seen, the trends of ĵθ and electron diamagnetic
drift current ĵDe are almost the same, which indicates that the dominant contributing
mechanism at the midstream is diamagnetic drift. Moreover, as is clearly shown in
figure 7(c), the total azimuthal current and the electron one still separate from each
other conspicuously, which manifests, at least until midstream, the ion azimuthal current
cannot be neglected under the high magnetization scenario.
Specifically, as for the electron diamagnetic drift current ĵDe , the reason for the
coexistence of diamagnetic and paramagnetic parts under 0.75 T and above is the
formation of high density conic – plasma density peaks at the edge of the plume,
which can be spotted on the radial profile of electron pressure in figure 8(b) and the
pressure contour in figure 3(b). This phenomenon has also been reported by Charles[22],
Takahashi et al.[17] experimentally and Saha et al.[42] numerically. There are two
different explanations of this phenomenon. The research conducted by Takahashi et al
[43, 44, 45] indicates that the high density conic is resulted by the enhanced ionization
caused by the energetic electrons tranported along the MDML. While according to the
research of Singh et al.[39] and Saha et al [42], the ions accelerated by the radially-
outward electric field overshoot the MDML but are bound by the strongly-magnetized
electrons which move almost along the magnetic field lines. This tendency of charge
separation creates a radially-inward electric field that reflects the overshooting ions back
inside the MDML, where they will be pushed outward by Er again. In the end, those
ions are trapped and accumulated along the MDML, which leads to the formation of
the high density conic. It seems the two mechanisms mentioned above play a role in
the generation of the high density conic seperately or simultaneously. And it should be
noted that since this study does not include the simulation of upstream sources, thus
it reflects the scenario of Saha’s experiments. As shown in figure 8(b), thanks to the
conic structure, the gradient of p̂e points to the edge of the plume, which results in the
opposite directions between ĵθ at different sides of the MDML, with ĵDe > 0 inside and
ĵDe < 0 outside. This explanation can also be used to explain the vacancy of the high
density conic under 0.25 T, as shown in figure 8(b), as the relatively weak magnetic field
is incapable of trapping enough ions along the plume edge. Also because of this high
Compositions and Distributions of the Azimuthal Currents in the MN 17
0.01
MDML 
j
θ
0.00 jθe,calc

j
 E×Be
-0.01 j


 De
 jinertia,e

-0.02 jstress,e

jθe
-0.03
0.5 1.0 1.5 2.0 2.5 3.0
r
(a) 0.25 T
0.01
MDML 
j
θ
0.00 jθe,calc

j
 E×Be
-0.01 j

 De

jinertia,e

-0.02 jstress,e

jθe
-0.03
0.5 1.0 1.5 2.0 2.5 3.0
r
(b) 0.75 T
0.010
MDML 
j
0.005 θ
jθe,calc

0.000 j
 E×Be
j

 De

-0.005 jinertia,e

jstress,e
-0.010 
jθe
-0.015
0.5 1.0 1.5 2.0 2.5 3.0
r
(c) 2.15 T

Figure 7. Azimuthal currents at the midstream under different magnetic strength:


directly-calculated total current ĵθ , electron current ĵθe , the four contributors of the
electron current ĵE×Be , ĵDe , ĵinertia,e , ĵstress,e and their summation ĵθe,calc .

denstiy conic, as the readers may have noticed, ĵθe,calc and ĵθe do not match very well
at ẑ = 1.5 ∼ 1.7. In the vicinity of the conic, the gradient of p̂e is so large that the grid
density there cannot resolve it preciously. Note that this is not a systematic error, as
the matching is very accurate in the upstream and downstream where the gradient of
p̂e is relatively more moderate.
Compositions and Distributions of the Azimuthal Currents in the MN 18

It is also quite interesting to note that under 0.25 T, the profiles of electron E × B
drift current differ from those measured experimentally in the literature[17]. As figure
7 (a) illustrates, the E × B drift current ĵE×Be also has co-existing opposite-direction
parts in the radial profile with paramagnetic ĵE×Be inside and diamagnetic parts on or
outside the MDML, respectively. This type of distributions is the natural outcome of
the formation of an electric potential barrier outside of the MDML. Return to figure
8(a), this figure shows the radial profiles of the normalized electric potential, an electric
potential barrier clearly exists near the MDML, which has also been reported in the
latest literature[46, 40, 27, 39]. This electric potential barrier can also be explained
by the ion-overshooting theory mentioned above. Some of the ions accelerated by the
radially outward Er overshoot the MDML and creates an accumulation of positive space
charges outside of the MDML, which results in the potential barrier[39]. Unlike the
formation of the high density conic, the generation of electric potential barriers requires
a weaker magnetic field so that ions can have a greater probability of penetrating the
MDML to produce the positive charge accumulation. This explains why the electric
potential barrier vanishes when B0 rises to 0.75T and 2.15 T in figure 8(a), in which
case the ions are so magnetized that they are almost impossible to transport across the
magnetic field lines. As a result, the E × B drift current is paramagnetic throughout
all radial positions in the midstream above 0.75T, as shown in figure 7(b) and (c).
As for the stress term ĵstress,e , this term is caused by the electron transport due to
the stress effects, and it mainly exists near the MDML. Under weaker magnetic field, as
shown in figure 7(a), the stress-induced current ĵstress,e is mostly paramagnetic. When
the magnetic field is raised, from 0.25T to 0.75 even 2.15T, the stress-induced current
gradually tends diamagnetic, leading to a positive contribution to the magnetic thrust.
Due to the small mass of the electron, the inertia term ĵinertia,e is still relatively
small compared to the other terms. And the direction of the inertia term is generally
in the diamagnetic direction despite the magnitude of the magnetic field.

3.3. Azimuthal currents in the downstream and the global propulsive performance
3.3.1. The downstream azimuthal currents. The downstream azimuthal current is the
least-investigated one in the magnetic nozzle. In this region, on the one hand, from a
simulation point of view, the fully-magnetized-electron assumption usually employed in
the fluid simulations fails, the plasma starts to get detached from the magnetic field lines
and the finite electron Larmor radius (FELR) effect starts to play a role. If the FELR
effect were to be treated properly, the computational economy of the fluid method will
probably be ruined[47]. On the other hand, from a experimental point of view, the
vanishing induced magnetic field makes it even harder to measure the induced currents
indirectly[17]. In our simulation, the electron model is consistent regardless of magnetic
field strength, making the fully-kinetic PIC code a more suitable tool for investigating
downstream currents.
Figure 9 shows the radial distributions of azimuthal currents downstream under
Compositions and Distributions of the Azimuthal Currents in the MN 19

-2.0
-2.2
-2.4
-2.6

ϕ

-2.8
-3.0 0.25 T
0.75 T
-3.2 2.15 T
-3.4
0.5 1.0 1.5 2.0 2.5 3.0

r
(a) Radial profiles of φ̂

0.10
0.25 T
0.08 0.75 T
2.15 T
0.06
pe

0.04
0.02
0.00
0.5 1.0 1.5 2.0 2.5 3.0

r
(b) Radial profiles of p̂e

Figure 8. Radial profiles of electric potential φ̂ and electron pressure p̂e at the
midstream

0.25T and 0.75T. In the figure, the directly-simulated ĵθe matches perfectly with summed
currents ĵθe,calc of the four contributing mechanisms, demonstrating the correctness and
applicability of our model.
In terms of amplitude, the downstream ĵθ is an order lower than the one of upstream
: the maximum of ĵθ in figure 9 is 0.004 while the corresponding upstream one is 0.04.
Despite the small amplitude, the volume that downstream ĵθ occupies can be much
larger than the one in the upstream due to the diverging topology of the magnetic
nozzle. Thus the acceleration or deceleration effects caused by the downstream currents
cannot be overlooked.
About the diamagnetic current ĵDe , because of the vacancy of high density conic
structure under 0.25 T in the midstream, as shown in figure 9(a), ĵDe is all-diamagnetic
in the downstream too, and reaches a maximum near the MDML. However, under 0.75
T, due to the confinement of ions near the MDML in the midstream, the high density
conic still has some influence in the downstream, resulting in an minor paramagnetic
peak of ĵDe at around ẑ = 2.3. And the diamagnetic peak is moved more outward at
Compositions and Distributions of the Azimuthal Currents in the MN 20

0.004  MDML
j
θ
0.002 jθe,calc

j
 E×Be
0.000 j


 De

jinertia,e
-0.002 
jstress,e

jθe
-0.004
1 2 3 4 5 6
r
(a) 0.25 T

0.004
MDML
0.002

0.000

-0.002

-0.004
1 2 3 4 5 6
r
(b) 0.75 T

Figure 9. Azimuthal currents at the downstream under low and medium magnetic
field strength: directly-simulated total current ĵθ , electron current ĵθe , four
contributors of electron current ĵE×Be , ĵDe , ĵstress,e , ĵinertia,e and their summation
ĵθe,calc .

ẑ = 3.5 compared to the 0.25T case.


On the E × B drift current, ĵE×Be remains fully paramagnetic regardless of the
magnetic field strength, and it reaches a maximum in the inner region beside the MDML.
The reason for this distribution is the inward detachment of ions. As shown in figure 10,
the solid red line shows a typical ion streamline started at the inlet, and the filled colors
of the contour indicate the distribution of radial electric field Êr . After being accelerated
by the radially-outward electric field near the inlet, the ion motion is steered inward by
the blockage of the potential barrier outside the MDML. Later on, the ion shoots near
ballistically inside the MDML again and separates inward due to its large inertia and
dramatically-reduced magnetic field strength downstream. Whereas, the electrons are
still magnetized to some extent with a Larmor radius of 0.6Rp0 on the MDML at ẑ = 6
under 0.25 T. Therefore, the electrons follow roughly along the magnetic field while the
ions detach near-ballistically, resulting in a local charge separation near the MDML. The
result of this charge separation can be seen in the right line chart of figure 10, where
a large outward radial electric field is generated, leading to the paramagnetic peak of
E × B drift currents near the MDML. In other words, the peak can be used as a symbol
Compositions and Distributions of the Azimuthal Currents in the MN 21

0.2000 12 M D M L
120.000.010.020.030.04
0.1597 10 10
0.1193 8 8
0.07900 6 6
0.03867 4 4
-0.001667 2 2
-0.04200 00 5 10 15 20 00.4 0.6 0.8 1.0
Figure 10. Left: contour of radial electric field Êr under 0.25 T. The solid black curve
represents the MDML, and the red curve with arrows represents the ion streamline
starting from r̂ = 0.5. Right: profiles of Êr and the relative electron Larmor radius
rL,e /Rp0 along ẑ = 6.

1.35
1.30
1.25
Ct =Ttotal /T0

1.20
1.15
1.10
1.05 0.25 T
0.75 T
1.00
0 2 4 6 8 10 12
z

Figure 11. The thrust gain coefficient Ct along axial direction under different
magnetic field strength.

for ion inward detachment.


In terms of the electron current arised from viscous stress effect ĵstress,e , its
amplitude is of the same order of ĵE×Be and ĵDe , thus is non-negligible in the
downstream. In the 0.25T case, under a lower magnetic field, ĵDe and ĵE×Be cancel
out with each other, leaving the ĵstress,e dominating the downstream of the MN. As have
been mentioned above, the gyro-radius of electrons downstream is of the order of the
plasma inlet radius Rp0 . Therefore, although the plasma can be seen as collisionless,
Compositions and Distributions of the Azimuthal Currents in the MN 22

electron layers with different macroscopic velocities can exchange momentum by the
lateral mixing of electrons gyrating with a system-scale Larmor radius. As can be seen
in figure 9(a), obtained from the viscous stress tensor τ e , ĵstress,e peaked at around
r̂ = 2.9. At the same time, if we look at the right subfigure in figure 10, the electron
Larmor radius has a maxima at r̂ = 3.0. The overlap of the two peaks clearly shows
that the lateral transport of electrons caused by the FELR effect is one of the main
mechanism for the downstream ĵθ . And the profile of ĵstress,e does not change much
when B0 rises to 0.75T.

3.3.2. The global propulsive performance. To assess the total propulsive effects of the
azimuthal currents, a criterion based on the thrust gain coefficient is derived first.
Since normalization is not related to the thrust gain coefficient, for the sake of reading
convenience, the normalization symbols (∧) will be omitted from the physical quantities
in the derivation process below. The total momentum flux per unit cross section τm can
be given by
τm (z, r) = mi ni u2i,z + pe (9)
where the thrust can be given by the integration of τm in a given cross section,
Z rp (z)
Ttotal (z) = τm (z, r)2πrdr (10)
0
where rp indicates the plume radius. Combining the momentum equations of ions and
electrons to eliminate the electric field[48], one can find that τm can also be expressed
as
∂τm 1 ∂
= −jθ Br − (rmi ni ui,r ui,z ) (11)
∂z r ∂r
thus Ttotal can also be expressed as
Ttotal (z) = T0 + TB (z) + Tr (z) (12)
where TB (z) and Tr (z) can, respectively, be written as
Z z Z rp (z)
TB (z) = −jθ Br 2πrdrdz (13)
0 0
Z z Z rp (z)

Tr (z) = − (rmi ni ui,r ui,z ) 2πdrdz
∂r
Z0 z 0
= −2π (rmi ni ui,r ui,z ) dz (14)

0 r=rp (z)

In terms of physical meanings, T0 = Ttotal (0) denotes the axial momentum brought
by the plasma flowing into the nozzle inlet. TB indicates the net contribution of the
volumetric Lorentz force resulted from the plasma azimuthal current, while Tr stands for
the force acting on the plasma-vacuum boundary r = rp (z). Apart from the magnetic
field, no mechanism can exert TB on the plasma-vacuum boundary, since Tr term is
not related to B whatsoever, it should vanish after the integration, i.e., Tr (z) = 0.
Compositions and Distributions of the Azimuthal Currents in the MN 23

Therefore, dividing both sides of equation (12) by T0 , we get the thrust gain coefficient
Ct
TB (z)
Ct (z) = Ttotal (z)/T0 = 1 + . (15)
T0
This ratio can be used to evaluate the global propulsive effects of ĵθ . If Ct > 1,
then the thrust is increased, the global character of ĵθ is diamagnetic, and the magnetic
nozzle are benificial to the thruster. The axial profiles of the thrust gain coefficient
under 0.25T and 0.75T are plotted in figure 11. Note that it also penalizes the thrust
gain coefficient when the plasma flows out of the computation domain, and the mass
flow will not be conserved in the axial direction. We know by integration that the axial
ion flux remains above 98% in the region of ẑ ≤ 12 for both cases. Therefore, it can be
considered that the variation of Ct (z) in this region is only brought by the azimuthal
current.
Comparing the two cases in figure 11, in the area before the downstream, the 0.75T
case has gained a larger thrust gain coefficient for larger diamagnetic ĵθ in the midstream
(Please refer to figure 7(a) and (b)). However, the thrust gain coefficient Ct in the 0.75T
case drops fast after reaching a peak at about ẑ = 8 and ends up even smaller than
that of 0.25T. This tendency can be well explained when we look at figure 9(b) where
the three paramagnetic peaks of ĵDe , ĵE×Be and ĵstress,e superimposed on each other at
around ẑ = 2.4, which induces great deceleration effects.
A similar result was also reported in our experimental research[49], in which the
thrust of an applied-field magneto-plasma-dynamic thruster (AF-MPDT) equipped with
a magnetic nozzle was measured using a target thrust stand. In the results, the
F − z curves increase to a local maximum and then decrease as z increases, which was
consistent with the trend presented in figure 11. Although this trend was attributed
to the insufficient size of the target then in the paper, it is more likely to result from
the paramagnetic ĵθ downstream of MN, as the results also show that the target is
large enough to block most of the ion flux. However, some earlier experiments[50, 51]
shows a continuous increase in the thrust with an increase in the magnetic field
strength and an increment saturation for the very strong magnetic field. The reasons
behind the discrepancy between the experiments and the simulations could be the
inconsistent working conditions, such as the different inlet density profiles or the neglect
of recombination processes in the simulations, which requires further studies.

4. Conclusions

This work filled some research gaps in studying the azimuthal currents in the magnetic
nozzle by employing a 2D axisymmetric PIC model to simulate the plasma flow inside a
MN with a quasi-neutral and current-free inlet. To the best of the authors’ knowledge,
this is one of the first works that incorporate the 2D-axisymmetric effect into a PIC
simulation of a magnetic nozzle / magnetized plume. Under the condition that the
electron temperature and the density profile at the nozzle entrance are set as uniform and
Compositions and Distributions of the Azimuthal Currents in the MN 24

Gaussian respectively, the profiles of the total azimuthal current and its four components
– ĵDe , ĵE×Be , ĵstress,e , and ĵinertia,e are investigated under different magnetic field
strengths. The main conclusions are summarized as follows:
On the electron azimuthal currents:
(i) The upstream and midstream azimuthal currents are dominated by diamagnetic
drift current ĵDe , whereas the downstream currents are influenced by ĵDe , ĵE×Be
and ĵstress,e together;
(ii) the diamagnetic drift current ĵDe in the upstream are all diamagnetic because of
the vacancy of the high density conic in the upstream;
(iii) the midstream has the most complicated distribution of different components.
Under low magnetic field strength, the ĵDe are all-diamagnetic while the
ĵE×Be displays an ”outer-diamagnetic, inner-paramagnetic” profile because of the
potential barrier appears outside the MDML. However, under a high magnetic field,
the ĵE×Be turns all-paramagnetic, while the ĵDe shows an ”outer-diamagnetic, inner-
paramagnetic” profile due to the formation of high density conic. Here, ”outer” and
”inner” refer to the direction relative to the MDML.
(iv) In the downstream, the FELR effect facilitates the electron viscosity, which results
in a conspicuous paramagnetic ĵstress,e . And the inward detachment of ions results
in an even stronger paramagnetic E × B drift current. Superimposed with the
paramagnetic peaks of ĵDe , they may reduce the thrust gain coefficient significantly
under a higher magnetic field.
Under the highly magnetized regime and collisionless condition, the volumetric ion
currents in the upstream and midstream are comparable to the electron ones. The ion
azimuthal currents are all-diamagnetic and mainly compose of drift currents caused by
E × B drift and centrifugal force. Different from the ion one, the electron azimuthal
currents caused by inertia can be safely neglected under most conditions discussed in
this paper.
Although the azimuthal currents have been investigated in detail here, there remain
some factors that may interest researchers in the field of electric propulsion, such as the
influence of inlet ion temperature and collisions on the azimuthal currents. These factors
will be subjects of future research.

Acknowledgments

This research was supported by the National Natural Science Foundation of China
(Grant No. 11872093) and the high performance computing (HPC) resources at Beihang
University. The authors also want to thank Prof. E. Ahedo and Prof. M. Merino from
University Carlos III of Madrid for their guidance in developing the initial version of
the PIC code used in this work. The authors also want to thank Prof. Yuan Hu from
Institute of Mechanics, Chinese Academy of Science for the useful talk on the numerical
diagnostics of pressure tensor.
Compositions and Distributions of the Azimuthal Currents in the MN 25

Appendix A. The conversion method to transform the PIC results to that


of the two-fluid code in the paraxial region

As explained in the main text, the primary reason for the inconsistency between
the axial profiles of parameters of the full-PIC code and the two-fluid code is found to
be such that the particle-based PIC code allows the plasma to cross the MDML self-
consistently, which would otherwise be prohibited in the two-fluid code because of the
numerical method. Therefore, the main idea of this conversion method is to put those
ions that have overshot the MDML back to the mainstream.
In order to do this, a post-processing program was developed and the main steps
of the algorithms are as follows:
(i) Check every ion in the computational domain and select those locate outside the
MDML, which are referred to as outside ions from here on;
(ii) Trace the outside ions back to the mainstream along the ion streamline. For each
outside ion, if the axial cell index of the intersection cell of this streamline and the
boundary of the mainstream(r = Rp0 ) is mi , then the traced-back-ion accumulator
ionTraceBack[mi ]+=1;
(iii) For every accumulator ionTrackBack[n], allocate the collected ions to all its
downstream cells according to the original density profile n̂i (z);
(iv) Re-calculate the number density in the paraxial region (r̂ < 1) and obtain the
converted axial number density profile n̂0i (z).
In this way, the density profile has been obtained, and the profiles of the electric
potential and ion axial number density can be calculated based on it.
It is assumed in the paraxial region, the magnetic field lines mostly align with z
direction, so the non-isothermal Boltzmann relation is evoked in the axial direction,
" γ−1 #
kB Te0 ne (z)
φ(z) = −1 . (A.1)
e(γ − 1) ne0
Eq.(A.1) can be normalized utilizing the same basis as mentioned in the main text,
1 
n̂e (z)γ−1 − 1 .

φ̂(z) = (A.2)
γ−1
The electric potential and electron number density before and after the conversion
is denoted as φ̂, n̂e and φ̂0 , n̂0e , which both satisfy Eq.(A.2) respectively. Therefore, after
some simple derivation, the converted electric potential φ̂0 is obtained,
n̂0 (z)
h   
0 1 i
φ̂ (z) = (γ − 1) φ̂ + 1 exp (γ − 1) ln −1 , (A.3)
γ−1 n̂(z)
where γ = 1.085 is the electron polytropic coefficient, which was fitted using the
simulated Te (z) and ne (z) in the paraxial region.
Assuming the ion acceleration is only caused by the electric field, and the
normalized ion axial velocity before and after the conversion are denoted as v̂z0 (z) and
Compositions and Distributions of the Azimuthal Currents in the MN 26

v̂z (z). Here the normalization basis for the axial velocity is taken as the corresponding
value at the plasma inlet.
1  2
mi vz (z) − v02 = e [φ(z) − φ0 ] ,

(A.4)
2
v̂z02 (z) − 1 φ̂0 (z)
= . (A.5)
v̂z2 (z) − 1 φ̂(z)
From Eq.(A.5), the converted ion axial velocity is obtained,
s
φ̂0 (z)
v̂z0 (z) = (v̂z (z)2 − 1) + 1. (A.6)
φ̂(z)
Utilizing the above mentioned method, the paraxial results of the PIC model can
roughly converted to fulfill the constraint of ”not overshooting the MDML” which is
compulsive in the two-fluid model.
Compositions and Distributions of the Azimuthal Currents in the MN 27

References
[1] Takahashi K 2019 Reviews of Modern Plasma Physics 3 3
[2] Andersen S A 1969 Physics of Fluids 12 557
[3] Charles C 2009 Journal of Physics D: Applied Physics 42 163001
[4] Correyero S, Jarrige J, Packan D and Ahedo E 2019 Plasma Sources Science and Technology 28
095004
[5] Longmier B W, Bering E A, Carter M D, Cassady L D, Chancery W J, Dı́az F R C, Glover T W,
Hershkowitz N, Ilin A V, McCaskill G E, Olsen C S and Squire J P 2011 Plasma Sources Science
and Technology 20 015007
[6] Arefiev A V and Breizman B N 2008 Physics of Plasmas 15 042109
[7] Sheehan J P, Longmier B W, Bering E A, Olsen C S, Squire J P, Ballenger M G, Carter M D,
Cassady L D, Chang Dı́az F R, Glover T W and Ilin A V 2014 Plasma Sources Science and
Technology 23 045014
[8] Charles C and Boswell R 2003 Applied Physics Letters 82 1356–1358
[9] Chen Z, Wang Y, Tang H, Ren J, Li M, Zhang Z, Cao S and Cao J 2020 Physical Review E 101
053208
[10] Takahashi K, Lafleur T, Charles C, Alexander P and Boswell R W 2011 Physical Review Letters
107 235001
[11] Arefiev A V and Breizman B N 2005 Physics of Plasmas 12 043504
[12] Hooper E B 1993 Journal of Propulsion and Power 9 757–763
[13] Ahedo E and Merino M 2011 Physics of Plasmas 18 053504
[14] Merino M and Ahedo E 2016 Plasma Sources Science and Technology 25 045012
[15] Corr C S and Boswell R W 2007 Physics of plasmas 14 122503
[16] Roberson B R, Winglee R and Prager J 2011 Physics of Plasmas 18 053505
[17] Takahashi K, Chiba A, Komuro A and Ando A 2016 Plasma Sources Science and Technology 25
055011
[18] Fruchtman A 2006 Physical Review Letters 96 065002
[19] Lafleur T, Takahashi K, Charles C and Boswell R W 2011 Phys. Plasmas 18 080701
[20] Little J M and Choueiri E Y 2010 AIAA paper 6615 2010
[21] Takahashi K and Ando A 2017 Physical Review Letters 118 225002
[22] Charles C 2010 Applied Physics Letters 96 051502
[23] Hu Y, Huang Z, Cao Y and Sun Q 2021 Plasma Sources Science and Technology 30 075006
[24] Boris J P 1970 Relativistic plasma simulation-optimization of a hybrid code Proc. Fourth Conf.
Num. Sim. Plasmas pp 3–67
[25] Eastwood J W and Hockney R W 1988 Computer Simulation Using Particles (Bristol and New
York: Adam Hilger)
[26] Birdsall C K and Langdon A B 1985 Plasma Physics via Computer Simulation (New York:
McGraw-Hill) ISBN 978-0-07-005371-7
[27] Rao S and Singh N 2012 Physics of Plasmas 19 093507
[28] Li M, Merino M, Ahedo E and Tang H 2019 Plasma Sources Science and Technology 28
[29] Hu Y and Wang J 2015 IEEE Transactions on Plasma Science 43 2832–2838
[30] Takahashi K, Charles C, Boswell R and Ando A 2014 Plasma Sources Science and Technology 23
044004
[31] Li M, Merino M, Ahedo E, Ren J and Tang H 2017 Full-PIC code validation and comparison
against fluidmodels on plasma plume expansions Proc. Int. Electric Propulsion Conf
[32] Brieda L and Wang J 2005 Modelling Ion Thruster Beam Neutralization 41st
AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit p 4045
[33] Brieda L, Pierru J, Kafafy R and Wang J 2004 Development of the DRACO code for modeling elec-
tric propulsion plume interactions 40th AIAA/ASME/SAE/ASEE Joint Propulsion Conference
and Exhibit p 3633
[34] Jiang Y, Tang H, Ren J, Li M and Cao J 2018 Journal of Physics D: Applied Physics 51 035201–
Compositions and Distributions of the Azimuthal Currents in the MN 28

[35] Cao S, Ren J, Tang H, Zhang Z, Wang Y, Cao J and Chen Z 2018 Physics of Plasmas 25 103512
[36] Tang H B, Cheng J, Liu C and York T 2012 Physics of Plasmas 19
[37] Tang H, Cheng J, Liu C and York T 2012 Physics of Plasmas 19
[38] Ahedo E and Merino M 2010 Physics of Plasmas 17 073501
[39] Singh N, Rao S and Ranganath P 2013 Physics of Plasmas 20 032111
[40] Saha S K, Raychaudhuri S, Chowdhury S, Janaki M S and Hui A K 2012 Physics of Plasmas 19
092502
[41] Bird G A Molecular Gas Dynamics and the Direct Simulation of Gas Flows (Oxford Science
Publications)
[42] Saha S K, Chowdhury S, Janaki M S, Ghosh A, Hui A K and Raychaudhuri S 2014 Physics of
Plasmas 21 043502
[43] Takahashi K, Akahoshi H, Charles C, Boswell R W and Ando A 2017 Physics of Plasmas 24
084503
[44] Takahashi K, Charles C, Boswell R, Cox W and Hatakeyama R 2009 Applied Physics Letters 94
191503
[45] Takahashi K, Charles C, Boswell R and Hatakeyama R 2008 Physics of plasmas 15 074505
[46] Little J M and Choueiri E Y 2016 Physical review letters 117 225003
[47] Ahedo E and Merino M 2012 Physics of Plasmas 19 083501
[48] Takahashi K, Lafleur T, Charles C, Alexander P and Boswell R W 2012 Physics of Plasmas 19
083509
[49] Wang B, Yang W, Tang H, Li Z, Kitaeva A, Chen Z, Cao J, Herdrich G and Zhang K 2018
Measurement Science and Technology 29 075302
[50] Takahashi K, Charles C and Boswell R W 2013 Physical Review Letters 110 195003
[51] Takahashi K, Komuro A and Ando A 2016 Physics of Plasmas 23 033505

View publication stats

You might also like