Are mercury emissions from satellite electric propulsion an environmental concern?

A new generation of satellites for Earth observation and telecommunications are being designed and built with off the shelf components. This is driving down costs and permitting the launch of large satellite swarms with unprecedented spatial and temporal coverage. On-orbit maneuvers are commonly performed using ion thrusters. Mercury is one of the cheapest and easiest to store propellants for electric propulsion. While some mercury released in Low Earth Orbit may escape Earth’s gravitational field, mercury emissions originating from many common orbital maneuvers will return to Earth. The environmental and human health implications of such releases have not been evaluated. Using an atmospheric chemical transport model, we simulate global deposition of mercury released from satellite propulsion systems. We estimate that 75% of the mercury falling back to Earth will be deposited in the world’s oceans, with potentially negative implications for commercial fish and other marine life. Understanding the scale of this novel mercury source in a post-Minamata Convention world is necessary to limit ecosystem exposure to mercury contamination.

to increase the number of satellites in Low Earth Orbit (LEO, altitude <2000 km) by a factor of ten over the next 10-20 years (Ailor et al 2017, Le May et al 2018, Peterson et al 2018. For efficient orbit raising and orbital station-keeping (maneuvers required to maintain the desired altitude and path), many of these satellites will include electric propulsion systems, predominantly Hall thrusters (Lev et al 2017a(Lev et al , 2017b. Hall thrusters produce thrust by ionizing a propellant and accelerating it across an electric field (Gallimore 2008). There has been little discussion of the risks associated with electric propulsion systems because the most common propellants, for example xenon, are not an environmental concern. However, the first ion thruster and the first orbital test in 1970 used the neurotoxicant, mercury, as a propellant because of its ease of storage, high molecular mass, and low ionization energy (Kaufman 1961, Kerslake andIgnaczak 1993). Concerns related to toxicity led to the abandonment of mercury as a propellant in the 1980s in favor of noble gases, predominantly xenon (Rawlin 1982). However, the high price of xenon ($2000/kg) coupled with burgeoning commercial space development has recently renewed interest in alternative, low-cost propellants such as mercury for electric propulsion (Kieckhafer and King 2007, Holste et al 2015, Saevets et al 2017, Elgin 2018. The potential magnitude of mercury emissions from satellite electric propulsion depends on the propellant mass per satellite and the number of satellites. Satellites in the proposed constellations have masses in the range of 100-500 kg (Ailor et al 2017, Le May et al 2018, and propellant typically comprises a large fraction of the mass of satellites due to the large velocity changes required for orbital maneuvering (Gallimore 2008, Lev et al 2017b. As a representative magnitude, 2000 satellites (the average size of two approved constellations (Federal Communications Commission 2017) each containing 100 kg of mercury propellant would emit 20 Mg of mercury per year over a 10 year lifetime. For comparison, this rate is about 50% of current annual total North American emissions, and about 1% of global annual anthropogenic mercury emissions (Muntean et al 2018, Streets et al 2019, UNEP 2019. These relatively large emissions from a single source will represent an even larger percentage under future emission scenarios, as more nations adhere to treaties such as the Minamata Convention on Mercury (UNEP 2017).
The Minamata Convention entered into force in August 2017 and has been ratified by 113 countries as of August 2019. The convention draws attention to anthropogenic mercury releases to the atmosphere, water, and soil and associated impacts on human health and ecosystems. The objective of the convention is to ban additional mining of mercury, and to phase-out mercury use in a number of manufacturing processes and everyday products. The convention aims to limit mercury emissions to the atmosphere and releases to terrestrial ecosystems from major anthropogenic sources such as fossil-fuel fired power plants, metals manufacturing facilities, caustic soda production plants, ore processing facilities including artisanal and small scale gold mines, waste incinerators, cement plants, and chemical production facilities. During the past two Conference of Parties in 2017 and 2018, the issues related to the interim storage of mercury and its disposal once it becomes waste, mercury contaminated sites, and human exposure were also addressed. In addition, among these anthropogenic sources, the potential contribution of mercury use as propellant in satellite propulsion has never been considered in past assessments of global mercury emissions and deposition (Pirrone et al 2010, UNEP 2013, 2019. However, the use of mercury as propellant in the aerospace sector is not specifically controlled by this convention. Therefore, just as rocket emissions are currently under scrutiny for their environmental impact (Ross and Vedda 2018), the potential repercussions of using mercury as a propellant in electric propulsion should be seriously considered.
In the 1970s, several studies were undertaken to examine environmental effects of cesium and mercury released from gridded ion thrusters as part of the Space Electric Rocket Test program (Lyon 1971). These studies were focused on implications for resultant ion belts (zones of energetic charged particles) rather than potential inputs of propellant to the troposphere. They found that mercury ions released at about 40-50 km s −1 at 1000 km altitude in a polar orbit are effectively constrained to spiraling back and forth between the polar magnetic mirror points (where charged particles trapped in the Earth's magnetic field reverse their direction). Roughly half of the ion trajectories have mirror points at altitudes below 100 km. These ions are rapidly removed from the magnetic field through neutralization and energy loss interactions with the atmosphere, potentially causing a faint artificial aurora. They are eventually removed by gravity into the lower atmosphere. The remaining trapped ions are most likely lost to space as they slowly form neutral atoms.
Simulations of xenon ion trajectories released at 44 km s −1 at altitudes from 15 000 to 115 000 km in near equatorial orbits suggest that roughly two-thirds of the ions under these conditions exit the magnetosphere and escape Earth's gravitation (Crofton and Hain 2007). About one-third were retained in a stable orbit over the 2-7 week simulation period. In a few cases, the ions reached the 30 km minimum altitude condition of the simulation, and therefore would be retained in the atmosphere. Ions released at lower altitudes were more likely to remain in stable orbits.
These studies serve as lower bounds on electric propulsion emissions retention, given that propellant from the proposed LEO mega constellations would be released at a much slower initial velocity, and from a much lower orbit. The main objective of the present study is to characterize the potential environmental impact of widespread mercury use as a satellite propellant in LEO by tracing the fate of mercury ions from the satellite to Earth.

Methodology
We first briefly describe the orbital dynamics of the ions as they leave the satellite and are subsequently captured within the atmosphere. Subsequent subsections describe the chemical transport model used to simulate the deposition patterns to the Earth of mercury emitted from Hall thrusters.

Mercury emissions from Hall thrusters
For Hall thrusters in LEO, the velocity of the accelerated propellant ions (the exhaust velocity, v x ) is close to the satellite orbital velocity (v o ). This means that the initial propellant ejection direction with respect to the satellite orbital direction is significant, as it determines the initial kinetic energy of the ions with respect to Earth When performing orbit raising or drag make-up maneuvers (where the energy of the satellite needs to be increased), the thruster is fired in the opposite direction to the satellite trajectory in order to accelerate it. For a typical Hall thruster with a mean exhaust velocity of about 15 km s −1 (Gallimore 2008), the resultant mean ion velocity with respect to Earth (v i ) will be slower than the Earth escape velocity, up to an orbital altitude of about 4000 km (supporting calculations can be found in the appendix). Below this altitude, the vast majority of ions released during these types of maneuvers will be captured within the atmosphere, lacking the necessary energy to escape.
For propulsion maneuvers such as de-orbiting (where the thruster is fired in a retrograde direction in order to reduce the altitude of the satellite) the ions will have resultant speeds greater than Earth's escape velocity. Determining the fate of these ions is not as straightforward. The collisional mean free path (the mean distance that a particle will travel before encountering a collisional interaction with another particle) of mercury ions at about 1000 km altitude is 10 4 -10 5 km (Johnson 1969), and their maximum gyroradius (radius of gyration of a charged particle in a magnetic field) is about 2 km (Thébault et al 2015), see appendix. This means that these ions will be trapped in a shell defined by the Earth's magnetic field lines with a thickness of the gyroradius, bouncing between the polar magnetic mirror points at most a few times before encountering an interaction with another particle. A few of the ions will be neutralized at a high altitude point in their path and will escape Earth's gravitational field. However, the ions are more likely to be neutralized in the lower, polar altitudes where the mean free path is shorter (<10 km at 200 km altitude) (Johnson 1969), and from where they will be less likely to escape before undergoing further decelerating interactions (Lyon 1971). Furthermore, ions ejected at small pitch angles to the magnetic field lines (e.g. along or against the direction of a satellite traveling in a near polar orbit, as most constellation satellites are planned) will have mirror points at low altitudes and therefore will most likely be decelerated and captured by the atmosphere rapidly (Lyon 1971). Despite these ions having resultant speeds greater than escape velocity, most will be retained in the atmosphere.
The decelerated mercury atoms may undergo continued re-ionization and re-neutralization as they traverse the ionosphere, but, as heavy particles, will gradually settle earthwards due to gravity (Brasseur and Solomon 2005) in a process similar to the atmospheric transport of meteoric ablation products (Feng et al 2013, Plane et al 2015. As atmospheric density increases and particle energies decrease at lower altitudes, most mercury will remain neutralized. Aside from this, the mercury chemistry in this region of the atmosphere is uncertain. The mercury will eventually diffuse to the turbopause (roughly 80-100 km altitude), which marks the boundary between the heterosphere and homosphere (Brasseur and Solomon 2005). To continue to trace the pathway of mercury from the turbopause to deposition on Earth, we turn to atmospheric modeling.  et al 2017b). The model uses a spectral grid, with a horizontal resolution ranging from T21 to T159. A horizontal resolution of T42 was selected for this study. Vertically, the standard model configuration is discretized with a hybrid-sigma pressure system with 19 or 31 non-equidistant levels up to 10 hPa (≈30 km). However, for this study, the model was extended to include a total of 39 non-equidistant levels in order to simulate the mesosphere up to 0.01 hPa (≈80 km), the approximate altitude to which the mercury will be transported due to gravitational settling.
For all simulations, an annual potential mercury emission rate from a constellation of satellites of 20 Mg yr −1 was used to force the model, as described above. More certainty on emission magnitudes would require more precise knowledge of planned constellation mission parameters such as orbital altitude, deltav (the velocity change required for maneuvers), and lifetime, as well as simulation and statistical analysis of ion trajectories and interactions (as in Crofton and Hain 2007), which we recommend for a future study.
The simulations were initialized with real meteorological data (January 2001), and then run with standard forcing for a ten year spin up period. The subsequent five years of simulations were used to analyze the tropospheric concentration fields and terrestrial and marine deposition fluxes. In order to account for temporal uncertainty in the propellant emission rates, the results from these final five years of the simulations have been averaged to give annual values and then standardized ([X−μ]/σ) to highlight the regions of high and low mercury concentrations, and to show more clearly the mercury transport patterns through the atmosphere.
In order to account for uncertainty in the spatial distribution and Hg speciation (oxidation state) of these emissions, an ensemble simulation approach was chosen (Parker 2013). A number of simulations were run, each with a different combination of spatial distribution and speciation flux input to the top of the model. The spatial distributions used were (a) globally uniform, (b) above 60°N only, (c) Northern hemisphere only, (d) Tropics only, (e) Southern hemisphere only, and (f) below 60°S only. The mercury speciation used was: (a) all elemental mercury (Hg 0 ), (b) all divalent mercury (Hg II ), or (c) all particulate mercury (Hg P ), for a total of eighteen simulations. The latter assumption (all Hg P ), although unlikely, was nonetheless considered for comparison purposes. To compensate for the relatively limited number of scenarios included in the ensemble, the bootstrap method (Mudelsee 2014), was used to calculate the average deposition fields that would be produced if a more comprehensive set of parameters and initial conditions had been tested (Parker 2013). To avoid redundancy of information within the ensemble, a reduced or 'inspected' ensemble (Solazzo and Galmarini 2015) was identified to include only the results which differ the most (as in De Simone et al 2015, 2017b. The similarity of the deposition fields was therefore assessed by means of both horizontal pattern correlation (Santer et al 1995(Santer et al , 1996 and the non-parametric Kolmogorov-Smirnov two-sample test, to include only those simulated Hg deposition fields where the test indicates that it is improbable (at a 95% level of confidence) that they belong to the same distribution.

Results and discussion
Model results show that convective mixing transports the Hg from the turbopause into the stratosphere, and eventually into the troposphere, consistent with our original hypothesis. The latitude-altitude profiles of atmospheric Hg concentration are similar irrespective of the location of the Hg emissions. All the simulations show that Hg predominantly descends into the stratosphere and troposphere at high latitudes, and that the highest concentrations at ground level occur at around 30°N. Figure 1 shows the latitudinal profiles for six simulations in which the mercury from satellites enters the top of the model with different spatial distributions. In these cases, the mercury was introduced into the top model layer as Hg II . Figure 1(a) shows the results for a uniform distribution of emissions, while figures 1(b)-(f) show the results when the emissions are confined to a single hemisphere, the tropics, or the poles. Higher concentrations outside of the emission region are predicted close to the stratopause and the tropopause over the North pole, and almost throughout the stratosphere over the South pole. Figure 1 suggests that, wherever the Hg input occurs at the top of the model domain, the dominant atmospheric descent pathways are the polar vortices, which is consistent with middle atmosphere dynamics (Brasseur and Solomon 2005) and studies of meteorite ablation products (Feng et al 2013). These concentration visualizations illustrate the atmospheric transport patterns, but deposition fluxes describe the actual ecosystem inputs.
The deposition flux simulations, as illustrated by the profiles in figure 2, indicate generally similar patterns for the three different mercury species tested, with some variation due to differences in the depositional processes affecting each species. The high deposition flux from 22°S to 22°N°reflects the deep convective scavenging of Hg II and Hg P by precipitation and the higher concentrations in the troposphere at those latitudes (figure 1). As expected for a wet deposition process, the general deposition profiles for Hg II and Hg P are similar to the global rainfall distribution, except for the dip in the region of the equator. This dip is most likely due to most of the Hg II and Hg P emissions being scavenged before they can be transported to the equator. In contrast, the longer residence time of elemental mercury (Hg 0 ) allows for a much broader distribution of the deposition flux, with only a slight dip around the equator. Dry and wet deposition processes are less efficient at the poles than they are at mid-latitudes and in the tropics, hence the tailing off of the deposition flux above 75°N and°S (Sprovieri et al 2016, Travnikov et al 2017. The map in figure 3 highlights the geographical regions of highest and lowest potential impact from mercury contamination by satellite emissions. While the simulated deposition fields differ noticeably, it is evident from this ensemble deposition field that some regions are more likely to be impacted by mercury deposition from satellite emissions than others. Compared to deposition patterns from biomass burning At 20 Mg yr −1 , mercury emissions from satellites would represent about 1% of global anthropogenic mercury emissions. While this would already comprise a significant quantity from a single source (Pirrone et al 2010, Pacyna et al 2010, it could constitute an even larger fraction in future emissions scenarios as the signatories to the Minamata Convention continue to take steps to reduce their use of mercury and employ abatement measures. Table 1  Furthermore, about 75% of the mercury from satellite emissions is deposited in the oceans. This fraction is larger than the magnitude (63%) of land-based anthropogenic emissions that are eventually deposited to the oceans (De Simone et al 2017b). Marine fish are the predominant vector of methylmercury exposure to human populations in many countries (Sunderland 2007) and thus such deposition patterns could potentially affect future human exposure to mercury. For example, marine fish supply more than 85% of the methylmercury intake to the United States population and the largest contributing source region is the Pacific Ocean (Sunderland et al 2018). Both the tropical Pacific Ocean and Indian Ocean are important supply regions for global tuna fisheries, which account for the dominant fraction of methylmercury exposure for many individuals (Sunderland et al 2018). In figure 3, enhanced mercury deposition regions over the global oceans between 10°and 40°projected to result from satellite emissions correspond to some of the most important harvesting regions for global commercial fisheries (Zeller et al 2016).
Compounding this, mercury deposited to aquatic ecosystems atmospherically may be more readily converted to the bioacccumulative methylmercury species than legacy mercury that has resided in a watershed or strongly bound to organic carbon complexes for an extended period (Hintelmann et al 2002, Harris et al 2007. This means that newly deposited mercury from satellite propellant may have a proportionally larger impact on biotic accumulation than suggested by the increase in the magnitude of deposition alone. A recent study suggests that climate change will further enhance mercury bioaccumulation in fish in the future (Schartup et al 2019).

Conclusion
In this study, we have traced the fate of mercury released from satellites through the atmosphere to deposition at the Earth's surface. The relevant literature and our summary of the ion orbital dynamics indicate that the vast majority of mercury emissions from satellites in LEO with Hall thruster propulsion will reach the turbopause (about 80 km altitude). The results of our model simulations suggest that almost 75% of the mercury reaching the turbopause is deposited over the oceans, independent of the initial Hg release conditions considered. Mercury emissions from satellite constellations, if implemented at the scales proposed, would constitute a single addressable source representing about 1% of existing global anthropogenic mercury emissions. If launched, satellite electric propulsion using mercury propellant would be a major environmental concern. While beyond the scope of this work, even the possibility of a failure of a launch vehicle carrying a number of satellites would clearly represent a catastrophic risk to local ecosystems. The environmental impact of mercury propellant is not worth the satellite cost savings of moving away from existing non-toxic propellants. The use of mercury as a satellite propellant should be Table 1. Distribution of mercury deposition (Mg) by region resulting from satellite emissions of 20 Mg yr −1 compared to annual deposition rates (Mg yr −1 ) for the MFR scenario and 2010 anthropogenic emissions, including confidence intervals. The geographical regions are based on those used in De Simone et al (2017b) and AMAP/UNEP (2013). monitored, quantified, and regulated by environmental intergovernmental organizations.