March 31, 2023

In-orbit demonstration of an iodine electric propulsion system – Nature

Propellant cost comparison

High-purity iodine is not needed in our propulsion system, and the total propellant cost for a purity of 99.5% was approximately US$60, with an additional cost below US$200 for iodine-related hardware. The propulsion system qualification cost was just under US$4,000. A modified version of our propulsion system using xenon propellant has also been developed. For the same mass of propellant, the cost of xenon was US$1,275, and owing to the high-pressure titanium tank, flow control valves, pipe and sensors, the hardware cost was about 100 times higher than that for iodine. The qualification cost also increased to approximately US$9,000.

The high cost of xenon is one of the reasons that SpaceX has instead chosen krypton as an alternative propellant for their Starlink satellites40. However, krypton has a higher ionization threshold and lower atomic mass than both xenon and iodine, and the required propulsion system power increases by more than 25% to achieve the same thrust level. Furthermore, the storage density of krypton is approximately three times lower than that of xenon (and nine times lower than that of iodine)41, which increases the volume and mass of the propellant tank. Krypton is used in a number of competing industries, such as window insulation applications, which account for more than 50% of the market share, and which is expected to grow rapidly42 due to the demand for energy-efficient buildings. Considering a Starlink constellation size between 12,000 and 42,000 satellites, each requiring of the order of 10 kg of propellant40, a substantial amount of krypton will be needed in the coming years.

Plasma modelling

Owing to the similarity in physical properties between xenon and iodine, and the wider availability and accuracy of important physical data (such as reaction cross-sections), xenon has been used in numerical modelling to aid the design, development and testing of the propulsion system. The model (which is similar to that in ref. 43) is self-consistent and considers mass flow conservation, and volume-averaged ionization and power balance within the thruster summarized by the following steady-state conservation equations





$${P}_{{rm{RF}}}=frac{{n}_{{rm{p}}}{u}_{{rm{B}}}{A}_{{rm{eff}}}{varepsilon }_{{rm{T}}}}{{eta }_{{rm{rf}}}}$$


Here (dot{m}) is the total input propellant mass flow rate, ({dot{m}}_{{rm{i}}}) is the mass flow rate of ions extracted from the plasma source, ({dot{m}}_{{rm{n}}}) is the mass flow rate of any unionized neutral gas, Aeff is the effective surface area for plasma loss inside the source tube, ({u}_{{rm{B}}}=sqrt{q{T}_{{rm{e}}}/M}) is the Bohm velocity34q and M are the ion charge and mass, respectively, Te is the electron temperature, nn is the average neutral gas density in the source tube, Kiz is the ionization rate factor, V is the volume of the source tube, PRF is the RF power, np is the average plasma density inside the plasma source, ({varepsilon }_{{rm{T}}}) is the effective energy loss per ion-electron pair lost to the source tube walls34,43 (and which includes both collisional and kinetic energy losses) and ηrf is the antenna–plasma power transfer efficiency. The equations above implicitly include relevant electron-neutral reaction processes (such as elastic scattering, and inelastic excitation and ionization processes), RF antenna–plasma coupling and plasma-wall sheath physics34,43.

Ion optics

One of the key elements of a gridded ion thruster are the ion optics, which in our case consists of a two-grid assembly. Particle-in-cell (PIC) simulations (using the open-source code XOOPIC44) have been used to model ion extraction and acceleration by the grids. Extended Data Fig. 3d shows a PIC simulation of ion acceleration through a single set of grid holes. Ions are well focused through the holes with no direct impingement, and only low-energy ions generated by ion-neutral charge-exchange collisions and possible downstream electron-neutral ionization with unionized neutral gas in the plume, strike the second grid (called the accel grid), as is typical in gridded ion thrusters5.

For given grid dimensions, if the total accelerating voltage is too low, the space charge of the ions between the grids can lead to under-focusing and direct impingement on the upstream surface of the second grid. This results in rapid sputter erosion, and possible shorting of the grids by sputtered material. The well known Child–Langmuir law34 can be used to estimate this space-charge-limited current, ICL, which in our case gives

$${I}_{{rm{CL}}}=frac{4{varepsilon }_{0}N{A}_{{rm{s}}}}{9}sqrt{frac{2q}{M}}frac{{{V}_{{rm{T}}}}^{3/2}}{{{L}_{{rm{eff}}}}^{2}}$$


where ε0 is the permittivity of free space, N is the number of grid apertures, As is the area of each aperture in the upstream grid (called the screen grid), VT is the total accelerating voltage across the grids and ({L}_{{rm{eff}}}=sqrt{{({t}_{{rm{s}}}+{L}_{{rm{g}}})}^{2}+{{r}_{{rm{s}}}}^{2}}) is the effective grid gap length with ts and rs the screen grid thickness and aperture radius, respectively, and Lg the physical grid gap length. A useful metric for quantifying the level of space charge between the grids is the perveance, p=Ibeam/VT3/2. When the ion beam current is equal to the space-charge-limited current, the maximum perveance, pmax, of the grids is reached. For the grids used in our propulsion system, pmax = 1.7 × 10−6 A V−3/2 for singly charged atomic iodine ions. If the total accelerating voltage is instead too high, the cross-over limit5 is reached and ions are over-focused, again leading to erosion. The space-charge and cross-over limits are indicated in Extended Data Fig. 3a.

Thermionic cathode neutralizer

Conventional electric propulsion systems typically make use of hollow cathode plasma bridge neutralizers5, which are capable of emitting a high electron current and are well suited to neutralizing large ion-beam currents. As our propulsion system operates at low power, and to further enable system miniaturization, two thermionic carburized thoriated tungsten filament neutralizers are used instead with a total estimated lifetime of 3,600 h.

Electrical system design

The electronics system is separated into modules as shown in Extended Data Fig. 2a. A main control unit coordinates the operation of the propulsion system, whereas each of the other modules controls a functional component by providing local regulation and monitoring of relevant parameters. The propulsion system is supplied by an unregulated voltage bus in the range 10–30 V and requires a power between 30 W and 70 W depending on the operating mode. A common mode filter on the power line reduces electromagnetic interference. The main communication channel with the satellite is a redundant Controller Area Network bus operating at data rates between 250 kbit and 1 Mbit. In addition, an inter-integrated circuit interface can also be used. Galvanic isolation is implemented on all communication channels.

The propulsion system uses five microcontrollers: one main processor and four second-level controllers managing local subsystems. The main microcontroller implements global control and safety algorithms, and also provides the interface with the satellite’s onboard computer (OBC). A real-time operating system with multiple tasks is used, where each task has a priority assigned and the scheduler decides which should be executed depending on the given priority.

After receiving a firing request from the OBC, the propulsion system switches on the subsystems, carries out built-in self-tests and proceeds with the plasma ignition sequence. Each microcontroller implements a bootloader allowing the OBC to reprogramme the user application in flight. This bootloader has several safety measures, such as redundancy or a triple voting algorithm, to avoid possible corruption caused by single-event upsets.

Thermal design

Heat is generated by ohmic losses in the power electronics and plasma losses to the source tube walls. The internal components of the gridded ion thruster reach the highest temperatures (up to 170 °C), whereas all other components and subassemblies are below 80 °C. The amount of heat needed for iodine sublimation is given by

$$Q={dot{m}}_{{rm{I}}2}triangle {H}_{{rm{s}}}$$


where ({dot{m}}_{{rm{I}}2}) is the mass flow rate and ∆Hs is the sublimation enthalpy of iodine (62.4 KJ mol−1). For a typical mass flow rate of 0.07 mg s−1, the sublimation power is less than 0.02 W. Owing to the reuse of waste heat, less than 1 W of additional power is needed by the flow management system to compensate for conductive and radiative losses and keep the propellant flow path sufficiently hot to prevent iodine deposition. Both the tank and flow path to the source tube have heaters maintaining the target temperature during start-up, ignition and steady-state operation. For a cold start, approximately 10 min is needed to heat the iodine to the required temperature.

Propellant valve

To enhance miniaturization and eliminate moving parts, the propulsion system does not use a conventional solenoid control valve. Instead, controlled iodine deposition and blocking of a submillimetre hole between the propellant tank and source tube is used. When the propulsion system is not operating, the temperature of the orifice causes deposition, which blocks the hole. At this deposition temperature, the resulting sublimation rate is very low. In addition, the effective gas flow conductance is substantially reduced owing to the design of the orifice, the gas distribution head, source tube and acceleration grid themselves, so that iodine leakage is low. Ground-based experiments with the propulsion system stored under vacuum for over two weeks show an upper limit leakage rate of less than 0.08 μg s−1.

Propellant loading

Iodine is filled into the porous matrix, which is placed inside the propellant storage tank before the filling process. To improve thermal conductivity, a polymeric thermal pad is placed between the matrix and the walls of the tank. Although iodine does not have a strong chemical affinity with oxygen under normal conditions, owing to its oxidizing nature the tanks are purged with argon before propellant filling to remove any residual gases that could contaminate the plasma during operation.

Iodine is melted at a temperature close to 120 °C in a separate reservoir and immediately poured into the matrix. This improves the packing factor over typical solid iodine crystals, and helps to minimize the formation of voids. The absolute pressure in the reservoir is just above atmospheric pressure with the argon partial pressure kept close to 100 kPa. A saturated state is maintained inside the tank as the vapour pressure of iodine is close to 14 kPa at 120 °C (ref. 45).

Diagnostics for ground testing

Vacuum chamber testing

Performance and plume characterization was performed by operating the propulsion system inside a cylindrical vacuum chamber with a length of 0.83 m and a diameter of 0.6 m. The chamber was pumped with a combination of rotary, turbo-molecular and cryogenic (operated at −75 °C) pumps. The pressure was measured with a MKS Baratron 627B absolute pressure transducer and a cold cathode Balzers IKR 050 gauge (with gas-specific correction factors applied). The chamber base pressure was better than 5 × 10−4 Pa, with a background pressure below 1.4 × 10−3 Pa maintained during operation. Although the neutral iodine gas dissociation fraction is not well known in the chamber, the effective pumping speed is estimated to be between 700 l s−1 and 1,400 l s−1.

Automated beam diagnostic system

Ion-beam current and divergence measurements are performed with a semi-circular, automated, beam diagnostic system46 that includes an array of 15 planar electrostatic probes. Motors at each end of the semi-circular arm precisely control the azimuthal arm position, which allows spatial measurements of the ion-beam current density over a two-dimensional hemispherical surface. The probes are biased at −40 V to reflect electrons and any possible negative iodine ions in the plume. The measured current is corrected to account for secondary electron emission due to ion bombardment of the probes and plasma sheath expansion around each probe due to the applied voltage46. The total ion-beam current, Ibeam, and effective beam divergence half-angle, θdiv, are obtained by integrating the measured current density profiles according to the following equations

$${I}_{{rm{beam}}}={R}^{2}{int }_{-frac{{rm{pi }}}{2}}^{frac{{rm{pi }}}{2}}{rm{d}}varPhi {int }_{-frac{{rm{pi }}}{2}}^{frac{{rm{pi }}}{2}}{rm{d}}theta {J}_{{rm{i}}}(varPhi ,theta )$$


$${theta }_{{rm{div}}}={{rm{cos }}}^{-1}left[frac{{int }_{-frac{{rm{pi }}}{2}}^{frac{{rm{pi }}}{2}}{rm{d}}varPhi {int }_{-frac{{rm{pi }}}{2}}^{frac{{rm{pi }}}{2}}{rm{d}}theta {J}_{{rm{i}}}(varPhi ,theta ){rm{cos }}varPhi {{rm{cos }}}^{2}theta }{{int }_{-frac{{rm{pi }}}{2}}^{frac{{rm{pi }}}{2}}{rm{d}}varPhi {int }_{-frac{{rm{pi }}}{2}}^{frac{{rm{pi }}}{2}}{rm{d}}theta {J}_{{rm{i}}}(varPhi ,theta )}right]$$


where R is the radius of the semi-circular probe arm, (varPhi ) and θ are the probe azimuthal and latitude angles, respectively, and Ji is the ion beam current density.

Time-of-flight probe

Time-of-flight (TOF) measurements were performed using a molybdenum disk with a diameter of 7 cm placed in the plume, and located 54 cm downstream of the accel grid. The probe was biased at −100 V to reflect electrons and any possible negative iodine ions29 in the plume, and the current collected by the probe was measured with a digital oscilloscope using short, low-impedance, connections. The time constant of the probe is much less than the ion transit time and is of the order of 1 μs. During measurements, both grids of the propulsion system are initially grounded before a rectangular voltage pulse with an amplitude and width of 1,000 V and 4.5 μs, respectively, is applied (with rising and falling times of approximately 0.5 μs). This causes an instantaneous extraction and acceleration of positive ions from the plasma source, and the appearance of distinct peaks in the measured TOF probe current due to the different ion transit times, τ, to the probe

$$tau =frac{L}{sqrt{2q{V}_{{rm{n}}}/M}}$$


where L is the distance between the exit of the propulsion system and the TOF probe, Vn is the net accelerating voltage and q/M is the ion charge-to-mass ratio. Owing to pulse-shape limitations, probe current peaks show a certain spread. Individual ion species contributions are determined by fitting exponential Gaussian functions and integrating to find the average current.

Retarding field energy analyser

A Semion 2500 Retarding Field Energy Analyzer (RFEA) from Impedans is used to measure the distribution function of beam ions. The RFEA has a diameter of 50 mm and includes a single grounded front grid, two internal grids with a controlled bias voltage and a biased collector plate. The RFEA is located 30 cm downstream of the propulsion system and is connected to an automated Semion Electronics Unit scanning system. The first derivative of the collector current as a function of the swept bias voltage47, Vbias, on the second grid then gives the ion flux distribution function, h(Vbias), defined such that

$$h({V}_{{rm{bias}}})propto frac{{rm{d}}{I}_{{rm{RFEA}}}}{{rm{d}}{V}_{{rm{bias}}}}$$


where IRFEA is the collector current measured by the RFEA.

Indirect thrust measurements

The integrated electronics in the propulsion system includes current and voltage measurement sensors that continually measure the applied accelerating voltage, and the current to both grids. For gridded ion thrusters, the ion-beam current that is extracted from the plasma source is balanced by an electron current to the first grid to maintain charge balance (Extended Data Fig. 2b). This current, after subtracting off the small current from the accel grid, then matches the net electron current emitted by the thermionic cathode neutralizer. During ground testing, and operation in space, the grid current and voltage measurements allow estimates of the extracted beam current and thrust to be made in real time.

Direct thrust measurements

Direct thrust measurements were performed with the propulsion system attached to a thrust balance placed inside the vacuum chamber. We developed a single pendulum thrust balance with a sensitivity of 0.03 mN that uses a force sensor to measure the thrust applied at the end of a moving arm. The force sensor and thrust vector location are shifted, which changes the respective pendulum lever arms and allows the measured force on the sensor to be magnified. The force sensor is an S256 load cell with an integral overload stop, which produces an analogue voltage output with a sensitivity of 1 mV V−1 at full-scale load (100 mN). To remove electrical interference, the low-level output voltage from the load cell is converted to a digital signal and sent to the measuring unit located outside of the chamber. The raw data are digitally smoothed with a second-order Savitzky–Golay filter. The thrust balance is calibrated with a set of known masses placed on a horizontal arm that produces a moment about the pendulum pivot balanced by the moment due to the force on the sensor.

Diagnostics for in-flight testing

The propulsion system includes eight thin-film platinum temperature sensors for measurement of the temperature (with an accuracy of 0.1 °C) at key locations, including all electronic subsystems, the propellant tank, and the interface flange between the tank and plasma source tube. The input current and voltage from the satellite, as well as output currents and voltages from different subsystems, such as the cathode neutralizer, grids and RF antenna, are continuously measured. The data acquisition frequency is set by the satellite onboard computer and is equal to 1 Hz.

Flight qualification

The propulsion system has undergone extensive vibration, radiation, thermal and flow testing for flight qualification. Vibration testing consisted of sinusoidal, random and sine-burst (quasi-static acceleration testing) at levels set by the spacecraft launch vehicle. Sinusoidal vibrations include low-frequency tests (5–100 Hz) with accelerations up to 4.5 g. Random vibration tests ranged between 20 Hz and 2,000 Hz with a total root-mean-square acceleration of 6.7 g and a duration of 120 s per axis. Quasi-static tests were also performed for each axis with a maximum acceleration of 8.75 g. Additional shock tests were conducted at frequencies up to 5,000 Hz, with a shock response spectrum acceleration up to 1,500 g. Electronic components and electromechanical assemblies underwent single-event radiation testing (high-energy proton bombardment) at energies up to 200 MeV, as well as gamma-ray testing for a total ionizing dose compatible with a qualification level of 15 krad for the unshielded assembly. The entire propulsion system underwent thermal exposure and thermal cycling campaigns in both ambient conditions and under vacuum conditions in a thermal vacuum chamber (with temperatures between −25 °C and 60 °C). Propulsion system operation in a vacuum chamber confirmed iodine sublimation and overall performance stability over extended firing times with multiple on–off cycles. Long-term propulsion system operation was tested with a qualification model for a total cumulative time of 120 h with 109 separate on–off ignition cycles.

Collection and analysis of in-flight data

The propulsion system electronics records approximately 50 telemetry parameters that are downloaded from the satellite after each in-orbit firing. The thrust and power depend on the operational mode selected, with different modes possible depending on the power, mass flow rate and applied grid voltage. Two modes have been tested during the in-orbit demonstration as shown in Extended Data Table 1, and denoted N1 and FS. The N1 mode has a thrust-locked feedback loop with a target thrust of 0.8 mN and an upper limit of 60 W, whereas the FS mode has a minimum thrust of 0.35 mN with an upper power limit of 50 W. In this last mode, the propulsion system has all feedback loops disabled, and data from secondary sensors are ignored. An automated self-test is performed before each firing.

Example system temperature measurements performed during in-orbit operation are compared with ground testing measurements in Extended Data Fig. 4b. The results are similar for all parameters, and again show that ground testing conditions replicate the space environment.

Orbit changes resulting from each firing were confirmed using both direct and indirect evidence. Direct evidence includes satellite tracking data from a GPS receiver onboard the satellite, and independent tracking data obtained from the SSN (see ref. 37 with satellite catalogue number 46838). Indirect evidence comes from a comparison of satellite orbital elements calculated from the GPS data with those predicted by numerical simulations using GMAT36, and a simplified theoretical model based on low-thrust trajectories around a spherical Earth48. The theoretical model uses the GPS mean semi-major axis just before manoeuvre 1A begins as an initial condition (backpropagating for earlier times). GMAT simulations use the JGM-3 geopotential model of degree and order 70 × 70, as well as point mass perturbations for the Moon and Sun. Atmospheric drag is included using the MSISE90 model49, as well as solar radiation pressure using a spherical spacecraft model36. Simulations are initiated using times and positions from the GPS data before each firing begins, and use approximate thrust profiles taken from the downloaded telemetry.

Owing to gravitational perturbations, the osculating semi-major axis of the satellite shows oscillations with an amplitude of the order of 10 km. For this reason, mean orbital elements based on Brouwer theory50 are used, which smooth out these high-frequency oscillations. The mean semi-major axis is deduced from the SSN data after converting from the Kozai to Brouwer mean motion convention51.

Source link