Technique for Two-Frequency Optimization of an ECR

Magnetic Nozzle Thruster

IEPC-2019-219

Presented at the 36th International Electric Propulsion Conference

University of Vienna, Austria
September 15–20, 2019

Benjamin Wachs∗1 and Benjamin Jorns†1

1 Department of Aerospace Engineering, University of Michigan, Ann Arbor, MI 48109, United States of America

The experimental setup and optimization techniques are presented for

an upcoming test campaign to improve the performance of a low-power
Electron Cyclotron Resonance (ECR) thruster using multi-frequency heat-
ing techniques. Recent experiments at ONERA using ECR plasmas in a
magnetic nozzle thruster have demonstrated thrust efficiencies over 10%.
The goal of the work presented here is to improve upon these results by
using custom power input wave-forms, thus enabling rapid optimization
without physical alterations to the thruster design. Specifically, this exper-
iment is designed to implement two-frequency heating to increase thrust
while average power and mass flow rate are held constant. While this
technique has been utilized to improve the yield and stability of ECR ion
sources, it has not been employed in thruster design. The thruster design,
microwave signal generation and power input, including a new wireless
power coupler, and thrust stand are described in detail. The optimization
algorithm and techniques for an upcoming test campaign are presented.

I. Introduction
Low power magnetic nozzle thrusters promise several features that make them ideal for small
satellite applications. They offer simple operation, with only a single required power supply, and lack
the often life-limiting neutralizer cathode that is required by most mature Electric Propulsion (EP)
technologies. However, performance to date has typically been much lower than more established
EP thruster designs, with low power thrust efficiency typically on the order of 1% 1 . Magnetic nozzle
thruster designs using Electron Cyclotron Resonance (ECR) heating, in particular those designed
at Office National d’Etudes et de Recherches Aérospatiales (ONERA), have shown great promise in
overcoming the historically poor performance. Their recent experiments have demonstrated thrust
efficiency over 10% and 1000s during a 50 W test, while previous Helicon thruster experiments have
typically seen efficiencies under 2% at these power levels 2–6 .
Magnetic nozzle thrusters generate force by converting the random thermal energy of a plasma,
typically generated by externally applied radiofrequency (RF) or microwave fields, to directed kinetic
∗
Ph.D Candidate, [email protected]
†
Assistant Professor, [email protected]

1
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
energy. In the case of ECR, heating is a achieved when the frequency of the applied electromagnetic
wave matches that of the natural electron cyclotron motion that occurs when a DC magnetic field is
present within the plasma. This condition can be described mathematically as ωin = eB/me , where
e is the electron charge, B is magnetic field strength, and me is the electron mass. Because the DC
magnetic field is not constant is space, typical ECR discharges contain a single resonant surface over
which the plasma absorbs most of its energy 7 . The hot electrons generated through ECR are then
expelled through an expanding magnetic nozzle, pulling the ions with them in an ambipolar diffusion
process. Finally, the plasma must detach from the magnetic field lines in order to generate useful
thrust. This process has been characterized in many magnetic nozzle thrusters, but to this day is not
well understood and is the subject of ongoing research 8–10 .
ECR magnetic nozzle thrusters have a long history in the EP community, with many earliest
thruster concepts built around the technology 11 . Though there was some success operating these
thrusters at kilowatt power levels, the bulky microwave sources of the time prohibited their use
on satellites, and research on this topic declined sharply as gridded ion and hall effect thrusters
matured 12 . While not suited for spaceflight at the time, ECR technologies have seen extensive use
in both plasma processing reactors and as ion sources for particle accelerators 13,14 . Since the 1960’s,
the miniaturization of microwave sources has enabled ECR to once again become a viable technology
for both medium and small scale satellites, and it has been recently used as the ionization source for
gridded ion thrusters in deep space missions 15 .
The goal of the experiment detailed in this paper is to continue to improve ECR magnetic nozzle
thrusters using optimization techniques inspired by the ECR ion source technology, namely two-
frequency heating. As such, this paper is organized in the following way. In Sec. II, we outline
different design parameters that can be tuned to improve ECR performance and explain why two-
frequency heating was selected for the initial optimization experiment. In Sec. III, we describe the
optimization algorithm used in the experiments, and in Sec. IV we present the experimental setup
including the thruster, vacuum chamber, thrust stand, and trial point testing techniques used in the
experiments.

II. Optimization Design Parameters: Two-Frequency Heating

Selecting proper design variables for an optimization experiment can be a challenging endeavour,
particularly when the underlying physics is poorly understood. As such, we turn to previous research
on ECR plasmas to inform our design choices. Several techniques have been developed to enhance
the performance of ECR plasma sources over their multi-decade history. These techniques are largely
dependent on the end use, whether it be process uniformity and selectivity in processing plasmas, or
highly charged ion yield in the case of ECR ion sources. The methods developed include both physical
design changes, such as magnetic field topology, wall material selection, and waveguide coupling, as
well as changes to the input waveforms, which will be the focus of this study.
In the EP community, work on developing ECR powered gridded ion thrusters has yielded several
new magnetic field designs and microwave antenna configurations. These new features have generated
increased plasma density as well as reduced erosion rates in these thrusters, and have enabled their
use in deep space missions 16 . More recently, research performed by ONERA on an ECR magnetic
nozzle thruster has shown that small geometric changes to the inner antenna and thruster walls
can have profound impacts on performance, with thrust efficiency increasing over 400% 17 . These
performance changes are not easily captured in simple simulations, thus necessitating a large number
of experiments to find optimal design points 18 .
While geometric design studies have yielded great improvements in device performance, changing
these parameters often comes at a great cost, both in terms of testing time and fabrication expenses.

2
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
As such, it is highly desirable to find design variables that do not require physical changes be made
to the thruster geometry. The primary such parameter is the input RF or microwave waveform
that is supplied to the plasma. While this variable has remained largely untouched in EP studies,
there have been many techniques developed for plasma processing and ECR ion sources that rely on
tailoring the input waveform. These include pulsed power techniques used in plasma processing and
multi-frequency heating, a common practice in ECR ion sources since the 1990s 19,20 .
Given the virtually unlimited number of variables that can be tuned when creating a custom
input waveform (frequency, duty cycle, modulation type, bandwidth to name a few), we intentionally
choose to limit the scope of our initial optimization experiments to focus on two-frequency heating.
This technique was first successfully implemented in the Lawrence Berkeley Laboratories Advanced
ECR ion source in the 1990s, and is now standard practice in highly charged ion sources 21–24 . Simply
stated, two-frequency heating adds a second resonance zone to the discharge, as shown in Fig. 2 (b).
By increasing the volume over which the electrons are efficiently heated, it has been postulated that
power coupling efficiency can be enhanced. This effect increases the density of hot electrons, which
in turn leads to better ion source performance. Subsequent experiments showed that the addition of
a second frequency dampened the discharge oscillations by suppressing kinetic instabilities typically
present in ECR ion sources 25 . The underlying physics of these improvements are still not fully
understood 26 . These experiments did, however, demonstrate just how sensitive ECR plasmas are to
small changes in input waveform with changes of only a few MHz significantly modifying the output
ion beam of a 14 GHz experiment 24,27 . Although the operating regimes of the ECR ion sources for
highly charged ions are quite different than those of ECR thrusters (i.e. much higher frequencies and
magnetic field strengths), the experiments indicate that two-frequency heating may be a promising
starting point for ECR thruster optimization.
By using two independent frequencies and holding the total input power constant, we have opened
the design space to 3 independent parameters: f1 , f2 , and P1 /P2 . Here, f1 and f2 are the two input
frequencies, and P1 /P2 is the ratio of their powers. These input parameters can be tuned at different
total power and flow rate settings to find optimal operating conditions at alternate thrust levels and
specific impulses. By optimizing at several set points, we can generate a Pareto front of optimal
parameters such as that shown in Fig. 1. This information would allow the thruster to be operated
more efficiently across a wider envelope of mission-dependent operating conditions.

3
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
Fig. 1 Hypothetical Pareto front showing the highest efficiency operating points at
various flow rates and power levels. By optimizing the input waveform at these flow
rates and power levels, we can potentially widen the operating envelope of the thruster.

III. Optimization Algorithm

As we will show in Sec. IV, measuring thrust at each new test point can take several seconds.
Furthermore, data from ECR ion source experiments indicates that the system may be very sensitive
to changes in input waveform. Thus, by introducing three free parameters, we have already expanded
the design space to an extent that a brute force approach cannot be executed in a realistic time frame.
We therefore require a more intelligent strategy for selecting new test points. Typical optimization
algorithms are likely insufficient for this task as there can exist several local optima and the data
produced by the experiment is often noisy. However, there exist several gradient-free non-convex opti-
mization techniques that are well-suited for this type of data. These types of optimization algorithms
have been successfully employed in fusion experiments in which tens to hundreds of parameters can
be tuned between each run 28 .
With this in mind, we have selected a Bayesian optimization solver for the initial ECR experiment.
This algorithm works by creating a surrogate model of the output function based on a predetermined
set of initial data points. It then uses an acquisition function incorporating the surrogate model to
select new test points. The surrogate model is then updated with the new data using a Bayesian
posterior probability, and the acquisition function is called again. This procedure is repeated until a
stopping criteria is met. The acquisition function can take several forms, but generally searches the
region where the most improvement of the surrogate model is expected. This algorithm is particularly
well suited for these experiments as it can optimize functions with noisy outputs. Furthermore, it
can easily be expanded to include more optimization parameters. A more complete explanation of
the full algorithm can be found in Ref. 29. Several Bayesian optimization packages are available for
open source and commercial platforms.

4
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
 IV. Experimental Setup

A. ECR Thruster
The thruster used in this experiment, shown firing in Fig. 2, is based on the ECR thruster
designed at ONERA 2 . This thruster utilizes a coaxial design in which microwave power is injected
from the back of the thruster between an inner antenna and an outer conductor which serves as both
a waveguide and as the walls of the plasma source. A DC block is placed between the input coaxial
cable and the thruster allowing the thruster body to float with respect to the chamber. The magnetic
field is generated by permanent NdFeB magnets with a peak magnetic field of 1100 gauss inside the
thruster. For the design used in this paper, gas is injected radially into the discharge region, however
both axial and radial gas injection schemes have been utilized in other experiments with varying
levels of success 30 .

(a) (b)

Fig. 2 (a) ECR thruster firing on 2 SCCM xenon at 20 watts input power, (b) Schematic
of the thruster showing the ECR resonance zones created by two-frequency heating

B. Vacuum Facility
Initial experiments were performed in a 0.9 meter diameter by 0.9 meter vacuum chamber at
PEPL, shown in Fig. 4. This chamber is equipped with a cryogenic pump capable of approximately
1,300 L/s pumping speed on xenon. Because high background pressure has been previously shown
to inhibit the performance of these devices, the experiment was moved to the Junior vacuum facility,
a 1 meter diameter by 3 meter chamber equipped with both turbomolecular and cryogenic pumps
capable of a combined pumping speed of roughly 32,000 L/s on xenon.

C. Microwave Power and Diagnostics

We present an overview of the microwave power setup including signal sources and diagnostics in
Fig. 3. Microwave power is first generated by two Mini-Circuits voltage controlled oscillators with
output frequency ranges from 1,300 to 2,700 MHz. These signals are then combined using a Mini-
Circuits ZX10-2-252-S+ combiner and amplified using a Comtech PST linear amplifier. The power
is sampled using a Mini-Circuits ZABDC20-252H-N+ directional coupler, and the forward power is
fed through a 3-dB splitter after which one output is connected to a Mini-Circuits PWR-6RMS-RC
true RMS power sensor and the other is fed to a HP 8563E spectrum analyzer. The reverse port of
the directional coupler is connected to a Mini-Circuits PWR-6GHZ power sensor to measure reflected
power.
For the purpose of this experiment, efficiency is measured with respect to the forward power
input to the thruster. The reflected power, therefore, is not taken into account, and efficiency could
hypothetically be increased through the use of a matching network. It is worth noting that mixed

5
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
signals, such as those generated by two-frequency heating, cannot be measured by standard Contin-
uous Wave (CW) power sensors. Instead "True Power" sensors are required, and even with these
devices, measurement bandwidth must be carefully taken into account.

Fig. 3 Schematic showing the microwave signal generation and diagnostic components

D. Thrust Stand
The thrust stand used in these experiments, shown in Fig. 4, employs a hanging pendulum design.
For our implementation, we use a counterweight on top of the pivots to increase the displacement
caused by the thruster. This design is similar to that used in Ref. 3. We measure thruster displace-
ment with a Philtec DMS-63 fiber-optic displacement sensor giving a ∼ 10 nm resolution. For the
initial tests, we do not employ active control to null the thruster displacement, however, this feature
may be added to future tests. The thrust stand is calibrated by placing a series of ∼ 0.5 gram weights
at a known moment arm with respect to the pendulum pivots. A typical calibration curve is shown
in Fig. 5(a).

Fig. 4 Thrust stand with the ECR thruster mounted

Initial tests were performed at a 20 watt, 2 SCCM-Xe operating condition. We show a raw data
trace from this test in Fig. 5 (b). Using the calibration data in Fig. 5 (a), the measured thrust was 535

6
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
giving a thrust efficiency of 3.4%. However, these tests were performed at a relatively high background
pressure (over 30µTorr), which is known to cause a decrease in thruster performance. Furthermore,
these tests were conducted prior to adding several features to the thrust stand to improve the accuracy
of measurements taken with microwave powered thrusters, as discussed below. The measurements
taken during this initial test, therefore, may not be reflective of true thrust numbers.

(a) (b)

Fig. 5 (a) Typical thruster calibration curve showing displacement as successive ∼ 0.5
gram weights are added, and (b) displacement as the thruster is turned on and off at a
20 watt, 2 SCCM operation condition

Thrust produced by RF and microwave powered thrusters can be particularly difficult to measure
due both thermal deformation and RF/microwave interference issues. Delivering microwave power
to the thruster requires the use of relatively stiff coaxial cables that both limit the sensitivity of the
thrust stand and expand during operation causing false readings. Initial tests of the ECR thruster
using RG-400 coaxial cable showed that cable heating could cause thrust readings on the order of
those produced by the thruster itself. Although no microwave interference issues were encountered
during initial tests, the thruster’s proximity to sensitive electronics makes it vital to consistently check
for false readings. The thrust stand therefore features several additions to make it suitable for testing
low-power ECR thrusters. These include wireless power coupling, a microwave power diverter, and
PID temperature control of the thruster and thrust stand.

1. Wireless Power Coupler

In order to avoid false readings associated with cable heating, power is coupled to the thruster
wirelessly through a custom designed wireless power coupler shown in Fig. 6 (b). These types
of wireless power coupler are becoming standard practice microwave thruster measurements 30,31 .
However, because of the large bandwidth required for our two-frequency ECR tests, a new coupler
design was needed.
For our design, we employ a coaxial geometry somewhat similar to an air dielectric coaxial cable.
This design enables a large bandwidth and is insensitive to small changes in the relative position of
the two halves. Several iterations of the design were simulated in COMSOL Mulitphysics, with a final
design achieving a -1 dB bandwidth from 800 to 2,500 MHz, as shown in Fig. 6(a). The design was
constructed using brass as the inner and outer conductor, and the resulting scattering parameters
were measured to be similar to the simulations. During initial thruster tests, we were able to operate
the thruster over the full bandwidth at powers up to 50 W.

7
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
 (a) (b)

Fig. 6 (a) Simulated S parameters of the microwave wireless power coupler and (b)
the final design undergoing testing with the network analyzer

2. Power Diverter
Because false readings due to microwave interference and thermal expansion are often difficult to
identify, we have added a microwave power diverter to the thrust stand as an additional sanity check.
For our design, a JFW RF switch is used to route incoming microwave power to either the thruster
or a 50Ω dummy load. By comparing the thrust measured with power sourced to the dummy load
to that measured with no power input to the thrust stand, we can quickly determine if the input
microwave signal is causing false thrust readings.

3. PID Temperature Control

Finally, in order to provide stable thermal environment, thrust stand and thruster are held at
constant temperatures using PID control. We implement this feature using Kapton heating strips
attached to both the thruster and thrust stand. The temperature at these points is read using
two DSB1820 digital temperature sensors, and heater current is controlled using an on-board micro-
controller.

V. Experiment Operation and Initial Results

The optimization experiment is implemented as a LabView VI. This VI controls the microwave
signal sources, diagnostics, and displacement sensor, and is coupled with a MatLab Bayesian Opti-
mization code to select new test parameters. The initial experiment is shown in operation in Fig.
7.
Initial tests revealed several challenges associated with this type of optimization experiment. The
most obvious of these difficulties was the discovery of hysteresis in thruster performance i.e. thrust
varied significantly depending on whether the test point was approached from a higher or lower power
state. This is a well known phenomenon in ECR plasmas, but is not typically encountered in EP
thrusters 32 . Furthermore, conflicting data between power sensors revealed the need for "True Power"
sensors, as mentioned above. While the power measurement issue could be solved by using new
equipment, the hysteresis problem required new testing methods. An overview of the new trial point
designed to eliminate the observed hysteresis technique is shown in Fig. 8. Here, the thruster is
first operated at a known low-power (∼ 5 watt) condition. It is then quickly switched to a higher

8
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
Fig. 7 Initial optimization experiment during operation. The thruster can be seen
firing on the left while the LabView VI controlling the experiment is shown on the right

power (∼ 30 watt) operating point, and then slowly transitioned down to the new test point using
variable attenuators. Once steady state is reached at the trial point, which was found to take around
10 seconds, the thruster is quickly transitioned to the low-power operating point, and thrust is
determined using the change in thrust. Finally, a new trial point is calculated by the optimizer and
the process repeats.

Fig. 8 Diagram showing the routine used to measure thrust at each new test point.
Relative thrust is measured by taking the ∆ Thrust between the test point and the
known low-power set point.

VI. Conclusions
In this paper, we have presented the preparations for an upcoming optimization experiment using
two-frequency heating to improve the performance of a low-power ECR magnetic nozzle thruster. We
show that ECR thrusters are well suited for optimization using custom input waveforms, and present
our choice of optimization variables: f1 , f2 and P1 /P1 . The thrust measurement techniques and
experiment facilities were presented, including a newly developed thrust stand. Finally, we present a
new technique for quickly iterating through trial points while avoiding the effects of hysteresis. The
full results of the tests implementing these new techniques will be the subject of future publications.

References
[1] Mazouffre, S., “Electric propulsion for satellites and spacecraft: established technologies and novel ap-
proaches,” Plasma Sources Science and Technology, Vol. 25, No. 3, June 2016, pp. 033002.

[2] Cannat, F., Lafleur, T., Jarrige, J., Chabert, P., Elias, P.-Q., and Packan, D., “Optimization of a coaxial
electron cyclotron resonance plasma thruster with an analytical model,” Physics of Plasmas, Vol. 22,
No. 5, May 2015, pp. 053503.

9
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
 [3] Vialis, T., Jarrige, J., Aanesland, A., and Packan, D., “Direct Thrust Measurement of an Electron Cy-
clotron Resonance Plasma Thruster,” Journal of Propulsion and Power , Vol. 34, No. 5, 2018, pp. 1323–
1333.

[4] Harle, T., Pottinger, S. J., and Lappas, V. J., “Helicon double layer thruster operation in a low magnetic
field mode,” Plasma Sources Science and Technology, Vol. 22, No. 1, 2013, pp. 015015.

[5] Lafleur, T., Takahashi, K., Charles, C., and Boswell, R. W., “Direct thrust measurements and modelling of
a radio-frequency expanding plasma thruster,” Physics of Plasmas, Vol. 18, No. 8, Aug. 2011, pp. 080701.

[6] Williams, L. T. and Walker, M. L. R., “Thrust Measurements of a Radio Frequency Plasma Source,”
Journal of Propulsion and Power , Vol. 29, No. 3, May 2013, pp. 520–527.

[7] Kinder, R. L. and Kushner, M. J., “Consequences of mode structure on plasma properties in electron
cyclotron resonance sources,” Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films,
Vol. 17, No. 5, 1999, pp. 2421–2430.

[8] Ebersohn, F., Girimaji, S., Staack, D., Shebalin, J., Longmier, B., and Olsen, C., “Magnetic Nozzle Plasma
Plume: Review of Crucial Physical Phenomena,” American Institute of Aeronautics and Astronautics,
July 2012.

[9] Olsen, C. S., Ballenger, M. G., Carter, M. D., Chang Diaz, F. R., Giambusso, M., Glover, T. W., Ilin,
A. V., Squire, J. P., Longmier, B. W., Bering, E. A., and Cloutier, P. A., “Investigation of Plasma
Detachment From a Magnetic Nozzle in the Plume of the VX-200 Magnetoplasma Thruster,” IEEE
Transactions on Plasma Science, Vol. 43, No. 1, Jan. 2015, pp. 252–268.

[10] Hepner, S. T., Collard, T., and Jorns, B. A., “Low Frequency Wave Detection in the Plume of a Low
Temperature Magnetic Nozzle,” 2018 Joint Propulsion Conference, American Institute of Aeronautics
and Astronautics.

[11] Miller, D., Gibbons, E., and Gloersen, P., “CYCLOTRON RESONANCE PROPULSION SYS-
TEM,” Electric Propulsion Conference, American Institute of Aeronautics and Astronautics, Colorado
Springs,CO,U.S.A., March 1963.

[12] Jahn, R. G., Physics of electric propulsion, Dover books on physics, Dover Publ, Mineola, NY, dover ed.,
unabridged republication ed., 2006, OCLC: 836917110.

[13] Lieberman, M. A. and Gottscho, R. A., “Design of High-Density Plasma Sources for Materials Processing,”
Physics of Thin Films, Vol. 18, Elsevier, 1994, pp. 1–119.

[14] Geller, R., Electron Cyclotron Resonance Ion Sources and ECR Plasmas, CRC Press, Jan. 1996, Google-
Books-ID: DtcVZnquQCAC.

[15] Kuninaka, H., Nishiyama, K., Funaki, I., Yamada, T., Shimizu, Y., and Kawaguchi, J., “Powered Flight
of Electron Cyclotron Resonance Ion Engines on Hayabusa Explorer,” Journal of Propulsion and Power ,
Vol. 23, No. 3, May 2007, pp. 544–551.

[16] Yamamoto, N., Masui, H., Kataharada, H., Nakashima, H., and Takao, Y., “Antenna Configuration
Effects on Thrust Performance of Miniature Microwave Discharge Ion Engine,” Journal of Propulsion
and Power , Vol. 22, No. 4, July 2006, pp. 925–928.

[17] Vialis, T., Jarrige, J., and Packan, D., “Geometry optimization and effect of gas propellant in an electron
cyclotron resonance plasma thruster,” Atlanta, GA, Oct. 2017.

[18] Alton, G. D., “Future prospects for ECR ion sources with improved charge state distributions,” Tech.
rep., Oak Ridge National Lab., TN (United States), 1995.

[19] Samukawa, S., Ohtake, H., and Mieno, T., “Pulse–time-modulated electron cyclotron resonance plasma
discharge for highly selective, highly anisotropic, and charge-free etching,” Journal of Vacuum Science &
Technology A, Vol. 14, No. 6, Nov. 1996, pp. 3049–3058.

10
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019
[20] Alton, G. D., “A New Method for Enhancing the Performances of Conventional B-Geometry ECR Ion
Sources,” Vol. 749, AIP, 2005, pp. 103–107.

[21] Xie, Z. Q. and Lyneis, C. M., “Improvements on the LBL AECR source,” Proceedings: INS International
Symposium, 24th, and 12th International Workshop on Electron Cyclotron Resonance (ECR) Ion Sources,
Saitama, Japan, Apr 25-27, 1995 , 1995, pp. 24–28.

[22] Maimone, F., Celona, L., Lang, R., Mäder, J., Roßbach, J., Spädtke, P., and Tinschert, K., “Influence of
frequency tuning and double-frequency heating on ions extracted from an electron cyclotron resonance
ion source,” Review of Scientific Instruments, Vol. 82, No. 12, Dec. 2011, pp. 123302.

[23] Vondrasek, R. C., Scott, R., and Pardo, R. C., “ECRIS operation with multiple frequencies,” Rev. Sci.
Instrum., pp. 5.

[24] Rácz, R., Biri, S., Perduk, Z., Pálinkás, J., Mascali, D., Mazzaglia, M., Naselli, E., Torrisi, G., Castro, G.,
Cellona, L., Gammino, S., and Galata, A., “Effect of the two-close-frequency heating to the extracted ion
beam and to the X-ray flux emitted by the ECR plasma,” Journal of Instrumentation, Vol. 13, No. 12,
Dec. 2018, pp. C12012–C12012.

[25] Skalyga, V., Izotov, I., Kalvas, T., Koivisto, H., Komppula, J., Kronholm, R., Laulainen, J., Mansfeld,
D., and Tarvainen, O., “Suppression of cyclotron instability in Electron Cyclotron Resonance ion sources
by two-frequency heating,” Physics of Plasmas, Vol. 22, No. 8, Aug. 2015, pp. 083509.

[26] Gammino, S., Ciavola, G., Celona, L. G., Mascali, D., and Maimone, F., “Numerical Simulations of the
ECR Heating With Waves of Different Frequency in Electron Cyclotron Resonance Ion Sources,” IEEE
Transactions on Plasma Science, Vol. 36, No. 4, Aug. 2008, pp. 1552–1568.

[27] Celona, L., “Microwave Discharge Ion Sources,” arXiv:1411.0538 [physics] , Nov. 2014, arXiv: 1411.0538.

[28] Baltz, E. A., Trask, E., Binderbauer, M., Dikovsky, M., Gota, H., Mendoza, R., Platt, J. C., and Riley,
P. F., “Achievement of Sustained Net Plasma Heating in a Fusion Experiment with the Optometrist
Algorithm,” Scientific Reports, Vol. 7, No. 1, Dec. 2017.

[29] Frazier, P. I., “A Tutorial on Bayesian Optimization,” arXiv:1807.02811 [cs, math, stat] , July 2018, arXiv:
1807.02811.

[30] Vialis, T., Développement d’un propulseur plasma à résonance cyclotron électronique pour les satellites,
Ph.D. thesis.

[31] McDonald, M., “Preparations for Thrust Measurement and Error Discussion of the IMPULSE Resonant
Microwave Cavity,” .

[32] Aydil, E. S., Gregus, J. A., and Gottscho, R. A., “Multiple steady states in electron cyclotron resonance
plasma reactors,” Journal of Vacuum Science & Technology A, Vol. 11, No. 6, Nov. 1993, pp. 2883–2892.

11
The 36th International Electric Propulsion Conference, University of Vienna, Austria
September 15–20, 2019