Wes Faler, Miles Space, Inc.
Brad Berkson, COO, Miles Space, Inc.
The patented Poseidon™ thrusters use pulsed electrostatic acceleration, a virtual cathode, and pressure drops to create thrust from water vapor using a novel propulsion method. No neutralizer is needed as the exhaust is inherently neutral. In-space testing shows market-leading performance:
•
37.49 mN thrust (average) (7.49 mN/thrust head)
o
Cold gas thrust is <= 1.2% of total thrust
•
1.49 W input low voltage electrical power (average, limited to <= 2A instantaneous by hardware fuse)
o
<= 1.20 W high voltage power delivered to plasma production and acceleration
•
>= 3,612 sec Isp (average, based on choked flow limit of maximum gas flow rate)
•
3.84 degrees off-boresight thrust angle
Right now, water propellant is a strong choice for Earth-orbiting missions. Water’s inherent safety makes it ideal for rideshare missions while its lack of bulky high pressure containment devices increases payload space and mission ROI.
Its compact form makes the Poseidon™ ideal for future deep space exploration missions. These missions are likely to encounter water whether it’s in the form of dirty ice or wastewater from in situ resource mining. The ability to use this water will be critical to continued deep space exploration. Poseidon™ thrusters have produced thrust using water vapor, Xenon, Argon, Krypton, Iodine, and air.
Poseidon™ Thruster Operating Principles
The physical structure of a Poseidon™ thruster has:
•
a planar plasma formation region containing spark electrodes
•
two exhaust ports, each ringed by acceleration electrodes
•
a single power supply providing spark and acceleration power
Vapor enters the plasma formation region, expanding and changing pressure on its path towards the exhaust ports. Paschen’s law ensures a spark occurs within the vapor at the point where the supply voltage meets the pressure on the Paschen curve. Each exhaust port is ringed with high voltage electrodes. One exhaust port’s voltage acts to focus and extract positive ions from the plasma. The other affects electrons.
Electrons, being far less massive than ions, leave the plasma before ions, generating thrust from their interaction with the acceleration electrodes. Once outside the thruster, the electrons form a virtual cathode that pulls upon the ions remaining within the thruster.
As the ions leave, thrust is obtained from acceleration electrodes. However, the ions also derive kinetic energy from the virtual cathode, slowing the exhaust electrons and even causing electrons to flow back toward the thruster. This gives an increased acceleration voltage upon the ions, expanding the classic Child-Langmuir limits for space-charge flow rate and thrust density.
The geometry of the thruster’s reaction chamber combined with the spark effect causes some ions to become trapped within the chamber while electrons and other ions exit as exhaust plume. The trapped ions are repelled by the exhaust ions. The repulsion moves them toward the front wall of the chamber, gaining momentum until they rebound from the wall. This “pressure-style” thrust is the primary method of thrust creation. Comparatively, only small amounts of thrust are generated by electrostatic acceleration of charged particles, by the Lorentz force acting on the spark along the linear electrodes (as is done with pulsed plasma thrusters), or by cold gas.
As the ions exit the thruster, they meet the returning electrons, neutralizing the plasma. With water vapor, the interface between exiting ions and returning electrons appears as a white-hot sphere 5-8mm outside the ion’s exhaust port. This phenomenon is believed to be due to the presence of multiple ion species with different velocity profiles.
A resonance occurs between the incoming gas pressure, spark push back, and plasma drain rate through the exhaust ports (as driven by the supply’s high voltage which can be varied to align with mission Isp). The Poseidon™ uses a very specific geometry to drive this resonance, minimize wear, reduce power supply complexity, and reduce flight computing demands.
Pressure-Based Thrust
The geometry of the thruster’s reaction chamber combined with the spark effect causes some ions to become trapped within the chamber while electrons and other ions exit as exhaust plume. The trapped ions are repelled by the exhaust ions. The repulsion moves them toward the front wall of the chamber, gaining momentum until they rebound from the wall.
Soon after spark formation, an energetic boundary is observed in simulation results. This boundary is parallel to the spark, itself perpendicular to each electrode. Ions created on the side nearest the exit nozzles will leave the reaction chamber. Of the ions created on the interior side of the boundary, only the hottest have enough energy to overcome the boundary and exit the reaction chamber. The coldest ions created on the interior side remain trapped for a relatively long period of time. Energetically, it is important to understand that the expelled ions were not pumped up a potential energy hill to exit.
From a mass flow standpoint, it is important to understand that sufficient mass must exist near the spark region for a useful number of ions to exit and others to be trapped. When mass flow is too low, observations show spark formation without appreciable thrust (or plume) due to the lack of ions to form the exit plume.
When a particle rebounds from a surface, momentum is imparted to the surface. The total momentum before and after the rebound are equal with the surface gaining the momentum needed to account for the direction change. Over a period of time with many strikes, the statistical average momentum per second transferred to an area is called “pressure”. Thus, the trapped ions apply a form of pressure to the front wall.
The rebounded ions travel back toward the exit, only to be slowed by the ongoing repulsion from the exhaust ions. At some point, the ions cannot reach the exhaust port and are turned back toward the front wall where they again apply pressure.
A trapped ion is one with sufficiently low kinetic energy to overcome the repulsive force of the exhaust ions. One of the exhaust ports is polarized to prefer focusing and accelerating electrons, rejecting ions. However, the exhaust port has a finite voltage level and so actually rejects low energy ions, allowing high energy ions to escape. In this way, the exhaust ports serve to sort the ions and foster an imbalance that traps some. This sorting also ensures that only cold plasma remains within the reaction chamber, minimizing wear and improving lifetime.
With no wall to strike, the exhaust ions gain speed as the repulsion from trapped ions accelerates them. The exhaust ions that exit first are not only the fastest ones within the reaction chamber but are at the leading edge of the pulse. These experience the most repulsion and become the fastest ions. The pulsed nature of the device serves to sort the ions, with the fastest ions getting faster.
A thrust head that ran through a large amount of water was torn down and inspected for wear. Deposits of water-borne minerals were seen on the face plate as rainbow-colored patches. Tests show that heat does not cause rainbow discoloration. Simulations show the plume briefly backflows during a pulse at these locations. Simulations also show a polarity inversion on the face plate consistent with the accelerator grid polarity inversion. The rainbow-colored patches were reversed in their color progression between the two exit ports, consistent with ion implantation driven by the field line polarization as simulated.
Further, deposits of minerals were seen on the internal front wall of the reaction chamber, where the trapped ions are simulated to strike the wall. Simulations predict the bulk of the pressure to be slightly off axis, biased toward the negative spark electrode. The tear down shows mineral deposits in this area and not in the area near the positive electrode where simulations show low pressure.
The thruster behavior does not comport with the underlying assumptions of the classic equations that relate beam power, thrust, and mass flow. These equations assume that a particle starts at an equilibrium point and only moves because energy is put into it by the accelerating effects of the electric field.
Use of these equations given observed results predicts nonphysical efficiency many times over 100%. In-space results show successful operation in the regime deemed impossible by these equations, thus necessitating a review of the underlying assumptions for the situations that apply to these equations. Furthermore, in-space results are consistent with ground test results obtained from in-house testing and from an independent third-party facility, belying the common belief that equipment misuse is the source of results inconsistent with the equations.
The classic equations make several algebraic leaps for convenience and omit efficiency terms for clarity. The equations assume that thrust is entirely created by accelerating propellant from zero velocity and that the propellant had zero initial potential energy. Pjet is often taken as the total energy imparted into the output beam and used for energy balance considerations. Some formulations of the equations consider an “effective exhaust velocity” that is a nonphysical term used for convenience, so take care not to use “effective exhaust velocity” as a real velocity in a kinetic energy equation.
Soon after spark formation, an energetic boundary is observed in simulation results. This boundary is parallel to the spark, itself perpendicular to each electrode. Ions created on the side nearest the exit nozzles will leave the reaction chamber. Of the ions created on the interior side of the boundary, only the hottest have enough energy to overcome the boundary and exit the reaction chamber. The coldest ions created on the interior side remain trapped for a relatively long period of time. Energetically, it is important to understand that the expelled ions were not pumped up a potential energy hill to exit, having been created near the peak of the potential energy hill.
In simulations that account for every particle and interaction, the energy sources are spark electrical energy and energy within the electric field. Energy uses include ionization and kinetic energy. At all points in time, the total energy present in the simulated systems is much less than the total energy input to that point. Much less in fact, in the 10% range. Yet, the accounting for momentum exchanges with thruster walls shows thrust on par with experimental values.
In-Space Test Results 2024-09
Setup and Results Summary
A Model M1.5 thruster is rigidly mounted onto a larger satellite owned by a third-party, with a total mass in the 100-200 kg range. The thruster was fueled with 250g of liquid water. The satellite was on the SpaceX Transporter 11 mission, launched Aug 24, 2024. The thruster was tested in space on Sep 11, 2024.
The device results were:
•
37.49 mN thrust (average) (7.49 mN/thrust head)
o
Cold gas thrust is <= 1.2% of total thrust
•
1.49 W input low voltage electrical power (average, limited to <= 2A instantaneous by hardware fuse)
o
<= 1.20 W high voltage power delivered to plasma production and acceleration
•
>= 3,612 sec Isp (average, based on choked flow limit of maximum gas flow rate)
•
3.84 degrees off-boresight thrust angle
During the test, the thruster was fired for 5 minutes continuously then powered off. Before and after the 5-minute period, no electrical energy was flowing to the thruster. That is, the thruster was not merely in an “idle” or “standby” mode. The timing behavior was set by the satellite operator’s flight computer.
The satellite operator made measurements of:
•
Input voltage
•
Input current
•
Craft 3-axis angular velocities
Of these, the voltage and current values show quantization artifacts typical of ADC conversion without subsequent filtering or offset adjustment. The angular velocity data is smooth by comparison and likely the result of sensor fusion. No additional insight into the angular velocity determination method is available currently.
The satellite operator computed:
•
Craft Center of Gravity before firing in the craft body reference frame
•
Craft Moment of Inertia before firing
•
Thruster location and orientation in the craft body reference frame
Mass flow is passive in the tested M1.5 device. No propellant heater, mass flow controller, or mass flow rate sensor are present in the system.
The placement of the thruster onboard the craft is shown below. Dimensions are from the center of gravity to the center of the thruster face plate. The exhaust plume advances into the +Z direction, with ideal thrust being entirely into the -Z direction. There are both X and Y offsets.
Ground Test Results at Miles Space
Setup and Results Summary
In [Ref.8], Tsifakis et al teach the successful use of a strain gauge to measure small thrust values. Strain gauges are found to have superior zero-drift behavior compared to laser interferometer methods. Tsifakis also demonstrates success with a suspended thruster, rather than flexure-based inverted pendulum systems. The combination of these permits a small, economical thrust stand system. The thrust stand used at Miles Space is based upon Tsifakis’ success.
The thrust stand system used at Miles Space is shown schematically in the figure below. The thruster is attached to a level stage, itself suspended from a rigidly mounted frame by 4 wires. Lateral thrust is measured using a strain gauge. Microslips at the point of contact between the strain gauge’s set screw and the thruster body, as well as twisting by a vectored thrust or tests of a linear array of thrust heads, is reduced by introducing a 2D “linearization plate” that is attached to the set screw. The linearization plate ensures there is a line of contact with the thruster, not a simple point.
Calibration is done by adding lateral force upon the thrust measurement stage. Lateral force comes from an industry-typical arrangement of weights upon a fine line, allowed to flex over a thin, low-friction pivot point, and with the line being adjusted by an actuator motor. The “calibration pyramid” protocol is used in which weights are added in sequence and removed in sequence, such that each weight’s effect is measured in separate periods, permitting zero-drift determination.
An instrumentation amplifier is used to amplify the strain gauge signal. The amplifier output has an analog low pass filter attached. Thrust is determined from the digitized value of this voltage and calibration results expressed as mN/V.
The mass flow sensor’s output voltage is directly connected to the data acquisition system. Isp is calculated from this signal and the thrust (itself determined by a calibration process).
Input electrical voltage is measured by the data acquisition system. Input current flows through a Hall Effect current sensor that produces an output voltage linearly related to the current flowing through it. The current sensor’s output voltage is directly connected to the data acquisition system. Input electrical power is calculated as the product of DC voltage and current.
Signals are connected to a data acquisition system. Data is collected at 100Hz, though this can be varied per experiment. Analog data is post processed using a low pass filter from the Python scipy software package. The filter is a 4th order Butterworth low pass filter. For strain gauge movement, 2Hz is the critical frequency. For voltage, current, vacuum chamber pressure, and mass flow, 1Hz is the critical frequency. Filtered values are used for time-series analysis. Unfiltered values are used for frequency analysis.
Live filtering, including moving average, introduces phase delays in data and makes temporal correlations challenging (especially with boolean signals that do not merit filtering). Post-processing filtering is done using the Python function scipy.signal.filtfilt (and newer version sosfiltfilt). From [Ref.9], this function’s feature applies a linear digital filter forward and backward, with a zero phase and a filter order twice that of the original. Compared to moving average filtering (common in this author’s experience and used in [Ref.8]), the signal is cleaner and better suited for causal analysis due to the lack of phase delay. See [Ref.9] for further reading on the issues with use of a moving average filter.
Typical calibration and experiment results are shown in the following sections, with values of:
•
29 mN average thrust, 0.8 W input low voltage power, 2,750 sec Isp
•
37 mN average thrust, 1 W input low voltage power, 3,250 sec Isp
The lab mass flow system is passive, relying upon water vaporization in vacuum. There is experimental evidence that the current mass flow system is delivering too much mass, showing signs of increasing thrust with decreasing mass flow. As such, the Isp values measured herein may be too low.
Ground Test Results at Miles Space
Setup and Results Summary
In [Ref.8], Tsifakis et al teach the successful use of a strain gauge to measure small thrust values. Strain gauges are found to have superior zero-drift behavior compared to laser interferometer methods. Tsifakis also demonstrates success with a suspended thruster, rather than flexure-based inverted pendulum systems. The combination of these permits a small, economical thrust stand system. The thrust stand used at Miles Space is based upon Tsifakis’ success.
The thrust stand system used at Miles Space has the thruster attached to a level stage, itself suspended from a rigidly mounted frame by 4 wires. Lateral thrust is measured using a strain gauge. Microslips at the point of contact between the strain gauge’s set screw and the thruster body, as well as twisting by a vectored thrust or tests of a linear array of thrust heads, is reduced by introducing a 2D “linearization plate” that is attached to the set screw. The linearization plate ensures there is a line of contact with the thruster, not a simple point.
Calibration is done by adding lateral force upon the thrust measurement stage. Lateral force comes from an industry-typical arrangement of weights upon a fine line, allowed to flex over a thin, low-friction pivot point, and with the line being adjusted by an actuator motor. The “calibration pyramid” protocol is used in which weights are added in sequence and removed in sequence, such that each weight’s effect is measured in separate periods, permitting zero-drift determination.
An instrumentation amplifier is used to amplify the strain gauge signal. The amplifier output has an analog low pass filter attached. Thrust is determined from the digitized value of this voltage and calibration results expressed as mN/V.
The mass flow sensor’s output voltage is directly connected to the data acquisition system. Isp is calculated from this signal and the thrust (itself determined by a calibration process).
Input electrical voltage is measured by the data acquisition system. Input current flows through a Hall Effect current sensor that produces an output voltage linearly related to the current flowing through it. The current sensor’s output voltage is directly connected to the data acquisition system. Input electrical power is calculated as the product of DC voltage and current.
Signals are connected to a data acquisition system. Data is collected at 100Hz, though this can be varied per experiment. Analog data is post processed using a low pass filter from the Python scipy software package. The filter is a 4th order Butterworth low pass filter. For strain gauge movement, 2Hz is the critical frequency. For voltage, current, vacuum chamber pressure, and mass flow, 1Hz is the critical frequency. Filtered values are used for time-series analysis. Unfiltered values are used for frequency analysis.
Live filtering, including moving average, introduces phase delays in data and makes temporal correlations challenging (especially with boolean signals that do not merit filtering). Post-processing filtering is done using the Python function scipy.signal.filtfilt (and newer version sosfiltfilt). From [Ref.9], this function’s feature applies a linear digital filter forward and backward, with a zero phase and a filter order twice that of the original. Compared to moving average filtering (common in this author’s experience and used in [Ref.8]), the signal is cleaner and better suited for causal analysis due to the lack of phase delay. See [Ref.9] for further reading on the issues with use of a moving average filter.
Typical calibration and experiment results are shown in the following sections, with values of:
•
29 mN average thrust, 0.8 W input low voltage power, 2,750 sec Isp
•
37 mN average thrust, 1 W input low voltage power, 3,250 sec Isp
The lab mass flow system is passive, relying upon water vaporization in vacuum. There is experimental evidence that the current mass flow system is delivering too much mass, showing signs of increasing thrust with decreasing mass flow. As such, the Isp values measured herein may be too low.
Industry Comparisons
NASA compiles data on In-Space Propulsion products. Data is taken from the published tables as well as vendor websites (when NASA-published data is incomplete regarding Isp, power, thrust, and propellant mass), to make a database of electric propulsion devices for spacecraft.
The Poseidon thrusters, both Ground and In-Space test results, compared to industry norms for thrust and Isp. The Poseidon thruster appear in the top-right quadrant, lagging other higher thrust devices and leading most others in Isp. As the Poseidon thrusters are themselves scaled up versions with multiple thrust heads, scaling will change the horizontal axis placement.
The Aerospace Corporation compiles data from industry as well. The Poseidon™ thruster is not yet present on these graphs. Graphs from this publication are shown below. The Poseidon thruster would appear near the Hall Effect thruster group and within the cluster noted for “Orbit Raising”, albeit with much hirher Isp and 400x less power.