Dawn at Ceres: The first exploration of the first dwarf planet discovered

Journal Pre-proof Dawn at Ceres: The First Exploration of the First Dwarf Planet Discovered Marc D. Rayman PII:

S0094-5765(19)31450-X

DOI:

https://doi.org/10.1016/j.actaastro.2019.12.017

Reference:

AA 7800

To appear in:

Acta Astronautica

Received Date: 11 March 2019 Revised Date:

3 November 2019

Accepted Date: 12 December 2019

Please cite this article as: M.D. Rayman, Dawn at Ceres: The First Exploration of the First Dwarf Planet Discovered, Acta Astronautica, https://doi.org/10.1016/j.actaastro.2019.12.017. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier Ltd on behalf of IAA.

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

Dawn at Ceres: The First Exploration of the First Dwarf Planet Discovered Marc D. Rayman Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., Pasadena, CA 91109, USA [email protected].

Dawn conducted an extensive exploration of dwarf planet Ceres, the largest object between the Sun and Pluto that had not previously been visited by a spacecraft. Following its arrival at Ceres in March 2015, Dawn acquired all the planned data from four circular polar orbits ranging in altitude from 13,600 km to 385 km. After the successful conclusion of the prime mission in June 2016, Dawn's mission was extended, and new investigations, not previously considered, were conducted from three new orbits. In 2017 the project was approved for a second and final extended mission at Ceres. Starting in April 2018, Dawn used its ion propulsion system to maneuver to two new orbits. These highly elliptical orbits enabled the acquisition of valuable new data, including significant improvements in the spatial resolution. Mission operations concluded in October 2018 upon depletion of hydrazine for attitude control. The mission provided a uniquely detailed view of the first dwarf planet discovered. This paper describes Ceres operations as well as some of the major findings there. Keywords: mission operations; Ceres; dwarf planet; asteroid; solar electric propulsion 1 Introduction The Dawn mission was designed to orbit and investigate dwarf planet Ceres and Vesta. With mean radii of 470 km and 261 km respectively, they are the two largest and most massive objects in the main asteroid belt. Based on a main asteroid belt mass of 2.7 · 1021 kg[1], Ceres is 35% of the total main belt and Vesta is 10%. Studies of these two protoplanetary remnants from the epoch of planet formation are expected to yield insights into physical and chemical conditions and processes then as well as during the subsequent evolution of the solar system. Dawn orbited Vesta from July 2011 to September 2012. The spacecraft acquired panchromatic (in stereo) and narrowband images in the visible and near infrared; neutron, infrared, visible, and gamma ray spectra; and gravimetry. The mission operated in six science orbits. The protoplanet is geologically more like a small terrestrial planet than like the smaller undifferentiated asteroids. Vesta operations and scientific findings have been described in detail[2-4]. Dawn's second destination was dwarf planet Ceres. Ceres was discovered by Giuseppe Piazzi in 1801, 129 years before Pluto's discovery. Dawn is not only the first spacecraft to explore a dwarf planet, it is the first spacecraft to explore the first dwarf planet discovered. Ceres is the largest body between the Sun and Pluto that had not previously been visited by a spacecraft. Dawn was the ninth project in NASA's Discovery Program and was managed by the Jet Propulsion Laboratory (JPL). The first principal investigator, Christopher Russell, was from the University of California, Los Angeles. Carol Raymond, from JPL, was the principal investigator during the 1

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

second extended mission. The spacecraft had strong inheritance from previous projects at Orbital Sciences Corporation (now Northrop Grumman Innovation Systems), which was responsible for the design, building, testing, and launch operations. (JPL delivered some subsystems and components to Orbital for integration into the spacecraft.) The scientific payload included visible and infrared mapping spectrometers integrated into one package known as VIR. VIR was contributed to NASA by the Agenzia Spaziale Italiana (Italian Space Agency). It was designed, built, and tested at Galileo Avionica (now Leonardo S.p.A.) and was operated by the Istituto di Astrofisica e Planetologia Spaziali (Institute for Space Astrophysics and Planetology). The nuclear spectroscopy was accomplished with the gamma ray and neutron detector (GRaND). GRaND was delivered by the Los Alamos National Laboratory and was operated by the Planetary Science Institute. Dawn's imaging (both for science and navigation) was accomplished with a prime and backup camera (framing camera 2, or FC2, and FC1, respectively). The instruments were contributed to NASA by the Max-Planck-Institut für Sonnensystemforschung (Max Planck Institute for Solar System Research) with cooperation by the Institut für Planetenforschung (Institute for Planetary Research) of the Deutsches Zentrum für Luft- und Raumfahrt (German Aerospace Center) and the Institut für Datentechnik und Kommunikationsnetze (Institute for Computer and Communication Network Engineering) of the Technischen Universität Braunschweig (Technical University of Braunschweig). The spacecraft and payload design, as well as the mission design and scientific objectives, have been described in detail elsewhere[5,6]. Dawn is the only spacecraft to have orbited an object in the main asteroid belt, and it is the only spacecraft to have orbited any two extraterrestrial destinations. The mission was enabled by solar electric propulsion, implemented as an ion propulsion system (IPS). Without electric propulsion, a mission to orbit either Vesta or Ceres alone would have been unaffordable within NASA's Discovery Program. A mission to orbit both would have been impossible. Dawn was launched on 27 September 2007. Mission operations at Vesta in 2011–2012 and in the entire interplanetary cruise through 8 September 2014 have been documented in detail [2,3,7-10]. 2 Cruise operations The interplanetary flight between escape from Vesta on 5 September 2012 and Ceres orbit insertion required about 2.5 years, with most of that time devoted to thrusting with the IPS. Indeed, Dawn (and the first interplanetary mission to use electric propulsion, Deep Space 1) spent more time in space thrusting than coasting. In important ways, the default state of the spacecraft was thrusting. Together with solar electric propulsion's tight coupling of mass, power, and thrust time, this had many significant implications for system engineering that are very different from missions with chemical propulsion[11,12]. The trajectory from launch through the end of the mission is shown in Fig. 1. The sensitivity to unexpected missed thrust varied significantly throughout the mission, and the 2

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

missed-thrust margin was a carefully managed resource. For most of the travel from Vesta to Ceres, the spacecraft paused ion thrusting once every 28 days to turn its high gain antenna (HGA) to Earth. The long thrust periods were one of the measures implemented to conserve hydrazine following the second loss of a reaction wheel in August 2012[9]. In addition to these monthly high-rate telecommunications sessions, Deep Space Network (DSN) coverage was scheduled for occasional contact through a low gain antenna (LGA). Some of the DSN sessions were to return low-rate telemetry and to reset the command loss timer, and all of them were to help verify that the spacecraft was still thrusting. The frequency of these thrust verification (TV) sessions depended on the sensitivity to missed thrust. In many cases, the TV passes used the no-downlink thrust verification (NDTV) strategy[10]. The peak sensitivity occurred in mid 2014. At that time, the ratio of missed thrust duration to delay at Ceres was about 1:9. While a delay would not have been fatal for the mission, it would have been inconvenient for mission planning and, if long enough, could have affected the project's cost and the hydrazine available for Ceres operations. By 11 September 2014, Dawn had thrust 100% of the time planned since departing Vesta, which was 93.2% of the time. On that day, a single event upset occurred in the controller for the IPS. The same phenomenon occurred in 2011, 19 days before the spacecraft smoothly and gently entered orbit around Vesta[2]. (A similar event occurred on 15 May 2017, although Dawn was not thrusting at the time.) The consequence was that thrusting terminated, and the spacecraft responded by entering one of its safe modes, in which the HGA was pointed at Earth. In the transition to that safe mode, a previously unknown bug in the attitude control software, which had not manifested itself even with eight years of flight operations, caused the spacecraft to have trouble holding its attitude. When the attitude error grew large enough, a deeper safe mode was invoked, this one doing all pointing relative to the Sun rather than Earth and using an LGA. Because of the frequent DSN passes in that phase of the mission, the spacecraft's anomalous condition was discovered quickly. Attitude behavior was correct for the safe mode, but telemetry showed that the normal attitude control mode would still experience the pointing problem. The operations team made several attempts through 13 September to correct it, none of which succeeded. Because of the high sensitivity to missed thrust, the project wanted to return to thrusting quickly. However, at a geocentric range of 3.2 astronomical units (AU), with a corresponding round-trip light time of 53 minutes, and a downlink rate of 10 bits/s, operations in safe mode were not rapid. Moreover, although extra DSN coverage had been scheduled for TV/NDTV sessions, the passes were short (sometimes as little as one hour) and inadequate for recovery from safe mode. Although the flight team normally worked a single shift, as soon as the difficulty of resolving the problem became evident, they changed to two shifts. In addition, to improve the efficiency of operations, the DSN used Deep Space Station 35 (DSS-35), which was not yet operational. The Canberra complex was undergoing maintenance at that time, thus limiting the availability of other stations, so DSS-35's coverage was particularly valuable. Following the normal Dawn sequence process, a new thrust sequence covering 28 days was already scheduled for uplink on 15 September in preparation for a thrust period that would commence on 16 September. Using that mature sequence could be lower risk than devoting effort to developing a new sequence with a later start (or even truncating the beginning of the new sequence). That defined 3

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

a target for recovery and readiness to resume normal operations. Nevertheless, throughout the recovery efforts, the project investigated trajectories with a range of thrusting restart dates. On 13 September, with the root cause of the attitude anomaly still elusive, the team determined that the most likely way to meet the self-imposed deadline was to reset the main flight computer. That was commanded later that day, and it was confirmed the next day that the software was no longer exhibiting the problematic behavior. In subsequent months, the root cause was identified, and operational methods were devised to avoid triggering it. Thrusting resumed on schedule on 16 September without some of the subsequent coast periods that had been included in the mission for nonessential activities incompatible with optimal thrusting. With the 95 hours of missed thrust, orbit insertion at Ceres shifted by less than 24 hours, from 5 March 2015 to 6 March. However, the first targeted science orbit was delayed from 18 March to 20 April because of the subsequent thrusting required to reach it. Fig. 2 shows the approach trajectory both as it had been planned before the interruption in thrust and after. Operations continued normally through the rest of the cruise phase, which formally concluded on 26 December 2014. Dawn had lost one of its four reaction wheel assemblies (RWAs) in June 2010[2] and a second in August 2012[9]. As a result, a major focus of work in the cruise phase had been the conservation of hydrazine, both during cruise and in reformulating and refining plans for Ceres. In addition, following the loss of the first RWA, the team developed and uplinked software for an attitude control mode that used two RWAs in combination with the hydrazine-based reaction control system (RCS). This "hybrid mode"[13] was planned for use at Ceres. All RWAs were powered off during the travel from Vesta to Ceres. When Dawn thrust with its IPS, RCS controlled the attitude around the thrust vector, and the two orthogonal axes were controlled by gimballing the ion engine. Shortly after departure from Vesta, the predicted hydrazine expenditure through the end of the interplanetary cruise was 12.5 kg, assuming no anomalies. The flight team implemented some important changes, both on the spacecraft and in operational procedures, to reduce that, yielding a prediction of 4.4 kg for the cruise phase. Upon completion of the cruise phase, the actual hydrazine consumption turned out to be 4.4 kg. 3 Ceres operations The strategy for exploring Ceres during the prime mission was based strongly on the extremely successful 16 months of Vesta operations[14]. Nevertheless, there were two significant reasons for making changes. The first was the continuing need to manage hydrazine use very carefully. That translated principally (although not exclusively) into reducing the number of spacecraft turns[10,14]. Hybrid mode was more hydrazine efficient than pure RCS control but was planned to be used only in the fourth and final orbit[10]. In that orbit, at the lowest altitude, RCS control would be the most expensive, so the limited life of the remaining two RWAs would be the most valuable. For the rest of the mission at Ceres, RWAs would be powered off, as they were during the interplanetary cruise.

4

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

The second major reason for a difference from the plans at Vesta was that Dawn would not depart from Ceres. Therefore, there was no motivation to ascend from the lowest altitude orbit. It is interesting, however, that with Dawn's unexpected capability to operate beyond its prime mission at Ceres, the first part of the first extended mission (see sections 3.6 and 3.7) made the mission even more like that at Vesta. At Vesta and during the prime mission at Ceres, Dawn's science observations were conducted principally from circular, polar orbits. The IPS was used to transfer to each science orbit. Ceres command sequences were designed, built, and reviewed during the interplanetary cruise from Vesta to Ceres. Some portions were tested on the project's spacecraft simulator as well. Updating them at Ceres was significantly less work and lower risk than would have been necessary without this extensive preparation. All plans included substantial margin to ensure requirements would be met even in the presence of typical classes of anomalies. Operations followed the plan very closely, and Dawn exceeded all of its requirements, acquiring far more valuable data than anticipated. The highlights of each phase are described below. 3.1 Ceres approach The principal objective of the approach phase was to deliver the spacecraft to its first science orbit. Approach began on 26 December 2014 and concluded on 24 April 2015. About 88% of this phase consisted of ion thrusting targeted to the science orbit. (Indeed, thrusting even during the interplanetary cruise phase targeted that orbit.) Along the way, Dawn was captured into orbit. Most of the rest of the time was devoted to optical navigation imaging sessions (opnavs) and telecommunications. Based on experience at Vesta, Dawn conducted seven dedicated opnavs at Ceres (plus two other observations described below, both of which provided optical navigation data). This was a significant reduction from the 24 planned for Vesta and provided a valuable hydrazine savings by eliminating the extra turns as well as the additional time that would have been needed with the HGA pointed to Earth to downlink the image data. (Note that telecommunications were conducted in all-RCS control, which used more hydrazine than ion thrusting.) Opnav 1 was on 13 January 2015 at a range of 383,000 kilometers, when Ceres had a diameter of about 27 pixels (px) in FC2. (See Fig. 3.) Opnav 2, on 25 January at a range of 237,000 km, provided images with 22 km/px, the first time Dawn improved upon the best Hubble Space Telescope (HST) images of 30 km/px [15]. The first two opnavs acquired images for one hour each, and the rest imaged Ceres for two hours. Although the purpose of the observations was to improve the spacecraft-Ceres ephemeris, VIR acquired data as well. In addition, for all but the first two opnavs, additional images were included to search for satellites of Ceres. Periods immediately before and after opnav 3 were dedicated to acquiring an even larger set of images for the search. No satellites were detected[16]. In addition to the seven opnav sessions, Dawn observed Ceres on two other occasions during the approach phase. On 12 February and 19 February (between opnav 3 and 4), FC2 and VIR collected 5

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

data throughout a full Ceres rotation. (One Cerean sidereal day is 9.07 hours.) These observations were designated rotation characterizations, or RC1 and RC2. As Dawn thrust on 6 March, it was gravitationally captured by Ceres at about 12:50 UTC. Traveling at 45 m/s relative to Ceres, Dawn was at an altitude of 60,600 km. Like capture at Vesta, this was not a critical event, thanks to the nature of the low-thrust trajectory with the IPS. No special precautions were necessary for the sequence nor for managing fault protection, and there were no special operational procedures for the flight team. Thrusting at that time was essentially the same as it had been for the previous 45,000 hours of thrust in the mission (then 69% of the spacecraft's time in space). The scheduling of DSN coverage in that phase of the mission had no need to account for the time of capture. As it turned out, there happened to be a TV pass only about one hour later. The simple verification that the spacecraft was still thrusting proved that it was captured. Although there was special personal interest by members of the flight team (including this author) and by the news media, there were no special activities on that TV pass. The next time Dawn used its HGA for contact was more than 6.5 days later, following the normal schedule for thrusting in that phase of the mission. As seen in Fig. 2, Dawn's initial trajectory in orbit went far to the anti-Sun side of Ceres. From 2 March to 9 April, the Ceres phase angle was too high for useful optical navigation. On 19 March, Dawn reached its highest altitude in orbit, denominated apodemeter by the team, at 75,400 km. (Demeter is the Greek counterpart of Ceres, the Roman goddess of agriculture.) GRaND was activated on 12 March. Useful nuclear spectra of Ceres could not be acquired until the fourth and lowest altitude orbit the subsequent December (see section 3.5), but the early operation provided a long baseline for optimization of instrument parameters. As we will see in section 5, GRaND's early data also provided a serendipitous scientific result. The last two opnavs were conducted on 10 April and 15 April. Thrusting concluded on 23 April, with Dawn in the targeted orbit. Dawn's X-band transmitter was off for most ion thrusting at heliocentric ranges greater than 1.93 AU (reached in May 2010), so electrical power could be devoted to the IPS. During the science orbits, however, the transmitter was on. When the spacecraft was not using the HGA, command sequences selected whichever of the three LGAs provided the best signal at Earth. This allowed frequent radiometric measurements and hence increasingly accurate determination of the gravity field. 3.2 First science orbit: RC3 RC3 (rotation characterization 3) was a circular, polar orbit at an altitude of 13,600 km with a period of 15.2 days. (Note: we use Dawn project nomenclature for the science orbit phases, but all names are derived from outdated mission concepts. Their literal meanings should be disregarded.) At that altitude, Ceres was ~ 710 pixels in diameter in FC2 and so fit easily in the instrument field of view. (See Fig. 4.) Dawn had observation objectives on both the lit and dark sides. Although ion thrusting targeted this orbit, it did not target the orbital phase. (In conventional orbital 6

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

elements terms, this is equivalent to not targeting the true anomaly at epoch.) That is, each time the approach trajectory was updated, the location in orbit at which Dawn would arrive was not a constraint. The data acquisition plans had been separated into five groups of sequences to allow observations to start as early as possible. Their execution would occur in the order that was most efficient based on the actual orbit. At the beginning of each observation session, the spacecraft turned to the required attitude for observing Ceres, and after each observation session, the spacecraft pointed its HGA to Earth and downlinked the data, so the sequences were independent of each other. Dawn completed ion thrusting at about 60°S on the sunlit side of Ceres and continued orbiting to the south. Before the RC3 phase could begin, the orbit knowledge had to be refined and the spacecraft configured for science operations. The time required to prepare for data acquisition then dictated that it begin with the nightside observations in the southern hemisphere. A procedural error in configuring for science operations caused the spacecraft to enter a safe mode with the HGA pointed to Earth. The resulting delay reduced the number of observations in the first session. All subsequent data in RC3 were acquired as planned. While on the dark side, Dawn observed the limb and the space above it over the southern and then northern hemisphere with FC2 and VIR to search for evidence of dust. Prior observations from the Herschel Space Observatory indicated occasional presence of water vapor[17]. Designed to study the solid surfaces of bodies with no atmosphere, Dawn's instruments would not be able to detect water vapor at the very low density measured by Herschel. Even the detection of dust entrained with lofted water was considered to be very unlikely, but the observations were conducted. As expected, no evidence of dust was found. When over the lit hemisphere, Dawn observed Ceres as it orbited above 45°–35°N, 5°N–5°S, and 35°–45°S. Each session covered a full Cerean day. RC3 yielded more than 2,500 images in the clear and seven color filters, providing global coverage at 1.3 km/px. VIR acquired more than 254,000 spectra of Ceres (as well as others in the space above Ceres). 3.3 Second science orbit: survey Following the return of data, Dawn departed RC3 on 9 May. After 534 hours of ion thrusting and 57 m/s, thrusting concluded on 3 June in the targeted circular, polar orbit at an altitude of 4,400 km. (See Fig. 5.) The survey phase had been planned to last for seven 3.1-day orbits. The project maintained flexibility in the mission timeline to accommodate not only anomalies but also human factors. It turned out that if Dawn had left survey orbit after seven revolutions, much of the work for the subsequent transfer to the third science orbit would not have been well aligned with standard workdays. Therefore, survey was extended by one full orbit, making the alignment quite good. During each pass over the illuminated hemisphere, Dawn acquired reflectance spectra and images. Most observations were at nadir, but portions of the second and fourth orbits included observations of the limb. During each pass over the unilluminated hemisphere, Dawn pointed its HGA to Earth. 7

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

As is evident in Fig. 4, Ceres has some regions of significantly higher albedo than the typical surface. In particular, Cerealia Facula, at the center of Occator Crater, and the more diffuse but still prominent Vinalia Faculae to the east (see Fig. 6 as well as Figs. 8, 10, 11, 21, 22, and 24-27), are so bright that FC images of those areas before survey orbit were saturated. Therefore, in survey and subsequent science phases, additional images with shorter integration times were included in the observation plans. During the fourth and eighth orbits, VIR autonomously reset. FC2 also reset during the eighth orbit (although there was good reason to believe the reset was unrelated to VIR's). Both instruments had experienced multiple resets at Vesta. It was not a surprise that resets occurred again at Ceres, and margin in the survey plans was more than sufficient to cover the missed data. Survey yielded more than 1,600 images in the clear and color filters, providing global coverage at 0.4 km/px. VIR acquired 4.3 million visible spectra and 2.9 million infrared spectra. 3.4 Third science orbit: HAMO Before the survey orbit phase, the flight team had been monitoring, studying, and managing occasional slips in some of the gimbals used by attitude control to point the ion engines. Gimbal motion would need to increase as the orbital altitude decreased. As the influence of the higher order gravity terms increased, the required thrust vector would change more rapidly. Because control of two spacecraft axes was accomplished by pointing the ion engine (with RWAs powered off, the third axis used hydrazine), greater demands were placed on the gimbals. During the survey phase, the flight team tested and implemented a mitigation for the gimbal that slipped the most. At the beginning of the transfer to the third science orbit on 1 July, the gimbal pointing errors were large enough to trip an attitude fault monitor and trigger a safe mode that pointed the HGA to Earth. Hydrazine consumption was very low in survey orbit, and the mission timeline remained flexible, so recovery was not considered urgent. The flight team switched to one of the other ion thrusters with a pair of gimbals that had not shown any evidence of slipping. (Dawn had three ion thrusters, all of which had been used extensively for routine thrusting, but no more than one was ever used at a time.) As before, the descent schedule was adjusted to align with standard workdays, and the transfer resumed on 15 July, so activities would occur on the same days of the week as for the transfer that had been planned for 14 days earlier. After 645 hours of ion thrusting and 73 m/s, thrusting concluded on 13 August in the targeted circular, polar orbit at an altitude of 1,470 km. (See Fig. 7.) Dawn mapped Ceres in each science orbit, but for historical reasons, this second lowest orbit was designated the high altitude mapping orbit (HAMO). The HAMO period was 18.8 hours. Unlike HAMO at Vesta, Dawn could not afford to point its instruments at Ceres on every pass over the lit hemisphere and then point its HGA to Earth on every pass over the dark hemisphere. The hydrazine cost of so many turns at Ceres was deemed too high. It was more efficient to turn less often and spend longer in HAMO. Therefore, sometimes Dawn kept pointing to nadir even over the dark side, accumulating and storing more data before downlinking for multiple revolutions. 8

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

It took 12 observation orbits to map Ceres fully with FC2. Because some orbits did not include observations, 14 orbits were needed to acquire and downlink the data. This 11-day period was designated one "cycle." By the time Dawn arrived in HAMO, the hydrazine consumption had been lower than predicted. Therefore, additional hydrazine was allocated to perform more turns than had been planned, increasing the amount of data the spacecraft could acquire and downlink. Dawn planned for six complete maps, each made with the camera held at a different combination of angle and direction relative to nadir. As at Vesta, with views of the ground from multiple perspectives, the topography could be measured. Planning for six cycles provided significant redundancy. The topographical mapping required using the clear filter in every cycle. In addition, different combinations of color filters were used in different cycles. Because VIR produces so much data, it was not possible to acquire data throughout every pass over the lit side, but the timing of VIR observations was adjusted to build up extensive coverage over the course of the 66-day HAMO. As the Sun was in Ceres' northern hemisphere, VIR observations were concentrated there, although the southern hemisphere was observed as well. All operations were executed as planned with only a subset of images in one cycle not obtained because of another FC2 reset. The loss was small compared to the margin in the plans, however, and HAMO yielded a rich set of data for all investigations. Dawn acquired more than 6,700 images in the clear and color filters, providing global coverage at 140 m/px. (One scene is in Fig. 8.) VIR returned 3.5 million visible spectra and 9.1 million infrared spectra. In addition, the gravity data were so good that Dawn met its gravity science requirement for Ceres in HAMO. It had been expected that the requirement could only be satisfied with data from the fourth and lowest science orbit. 3.5 Fourth science orbit: LAMO/XMO1 The first three science orbits were dominated by FC and VIR observations, but gravity and GRaND measurements were made in all phases as well. GRaND began to detect Ceres in HAMO, but the signal was too weak for useful nuclear spectroscopy. As at Vesta, the motivation for going to the lowest orbit (the low altitude mapping orbit, or LAMO, subsequently designated extended mission orbit 1, or XMO1, when Dawn was in its first extended mission) was to satisfy the GRaND and gravity requirements. (When the project formulated detailed plans for Ceres, it did not anticipate being able to meet the gravity requirement above LAMO. Of course, LAMO measurements remained highly desirable to refine the gravity field.) There had been no FC or VIR requirements. Indeed, in an early plan for LAMO at Vesta, those two instruments would not be operated at all so that all resources (including onboard data storage and operations team focus) could be devoted to the acquisition of the high priority data. Ultimately, however, once sufficient confidence in the capability to meet the requirements was established, FC and VIR observations were included in the plans. Earlier in Ceres operations, new VIR requirements and bonus objectives were established, and FC objectives were defined as well. Dawn began its descent to LAMO on 23 October. As expected, this was the most challenging orbit 9

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

transfer because of the deep penetration into the gravity field. Moreover, the orbit perturbations from increasing RCS activity became more significant. Extensive Monte Carlo analyses had supported development of a plan for how often to update the thrust vectors as Dawn descended. It turned out to be possible to maintain a cadence of updates every seven days, each taking about four days to design, build, review, and uplink. That allowed the team to keep most of the work on a standard, prime-shift schedule. Unlike the transfers to the higher orbits, analysis showed that a pair of final, purely statistical trajectory correction maneuvers (TCMs) would likely be needed. When Dawn began its transfer from HAMO to LAMO, it was at a heliocentric range of 2.97 AU, as Ceres approached its 6 January 2016 aphelion at 2.98 AU. Dawn's solar arrays provided about 1.35 kW. The IPS was throttled down to 0.7 kW, delivering 25 mN at a specific impulse of 1,800 s. (At lower heliocentric ranges, the specific impulse exceeded 3,000 s.) After 976 hours of ion thrusting and 110 m/s, thrusting concluded on 7 December. As expected, TCMs were needed. Thrusting for 20.3 hours on 11–12 December and again for 15.3 hours on 12– 13 December provided another 4.1 m/s to bring the orbit closer to the design of a mean altitude of about 385 km and period of 5.4 hours. (See Fig. 9.) Variations in orbital radius and Cerean topography (dominated by the difference between the polar and equatorial radii) together had caused only a relatively small variation in orbital altitude in the first three science orbits. The altitude in LAMO varied between about 355 and 410 km, with roughly equal contributions from the orbit and from Ceres' shape. Following the plan established in 2013, after arrival in LAMO, the two operable RWAs were activated on 15 December and Dawn operated in hybrid mode. Given the two prior failures, the RWAs were not considered reliable, so all LAMO plans were independent of the control mode. The team was always ready to continue LAMO in RCS control (with greater hydrazine consumption) if another RWA failed. Dawn pointed GRaND near nadir for up to 15 consecutive orbits, or 75% of every 4.5-day period on average. The remaining time was devoted to a five-orbit, 27-hour HGA telecommunications session. As in the higher orbits, when the spacecraft was not using the HGA, command sequences selected whichever LGA was closest to pointing at Earth to allow gravity measurements. (Roughly 70% of the nadir time had scheduled DSN coverage in order to meet the requirements of the gravity investigation.) In addition, FC2 and VIR data were acquired during some passes over the lit hemisphere. (See Fig. 10.) To provide additional perspectives, sometimes FC2 was commanded to take images even while Dawn was rotating from nadir to point the HGA to Earth. LAMO was organized in cycles of 23 days. Each cycle consisted of five 20-orbit segments plus one more orbit to ensure that adjacent cycles did not point the HGA to Earth when flying over the same terrain. These phasing orbits helped improve surface coverage. Two of the five segments per cycle had allocations for orbit maintenance maneuvers (OMMs) in case they were needed to provide the desired ground track. Each OMM window was three consecutive orbits of the 15-orbit nadir time. Although some of the OMMs were executed, most were not necessary, and in those cases, Dawn continued pointing to near nadir. Two cycles of data were needed to meet the project's science data acquisition requirements, so four 10

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

had been planned in detail as part of the sequence development work conducted during the interplanetary cruise phase. The need for four cycles in LAMO, the most hydrazine-expensive phase of the mission, also set targets for hydrazine conservation. Enough hydrazine had to be available (with margin) to allow operations for four cycles even if a third RWA failed by the beginning of LAMO, precluding the use of hybrid control. When the Sun-Earth-probe angle (SEP) was less than about 20°, Doppler measurements tended to be too noisy for high-accuracy gravity science. SEP was less than 20° from 3 February to 3 April 2016. Normal operational navigation and telecommunications did not have such a wide, albeit soft, constraint. The minimum SEP was 7.9°, thanks to Ceres' inclination of 10.6°, and was large enough that it had no effect on operations other than reducing gravity science data quality. The first four cycles were very successful. No anomalies interfered with data acquisition and return nor consumed hydrazine. The RWAs operated perfectly, and hybrid control kept hydrazine expenditures low. With the successful acquisition of neutron and gamma-ray spectra for the first two cycles, by February, the project had met or exceeded every one of its prelaunch requirements for science data acquisition at Ceres[18]. (All requirements at Vesta had been met or exceeded as well.) Variations in the gravity field and the perturbing effects of RCS activity reduced the accuracy of ephemeris predictions as far in advance as needed for sequence development. That made the bonus objective of building up complete surface coverage with FC2 more difficult. Nevertheless, by the end of February, Dawn had fully imaged Ceres from LAMO. In January, when it became evident that the flight system likely would be capable of maintaining productive operations after the end of the fourth cycle on 19 March, planning for subsequent cycles began. They included acquisition of more GRaND and gravity data and VIR observations of selected regions. All FC2 imaging in the first four cycles had been with the clear filter at nadir. Subsequent cycles included color filter images of high priority targets. Moreover, instead of pointing to nadir, Dawn pointed slightly off-nadir for high-resolution topography. As the end of the prime mission approached, two options for an extended mission were considered. One was to continue to investigate Ceres. The other was to leave Ceres to fly by 145 Adeona, a Ch asteroid with a mean diameter of 150 km. It is noteworthy that after being the only spacecraft to orbit two extraterrestrial targets, and even maneuvering extensively in orbit around Vesta and Ceres, Dawn still had the capability to leave Ceres and reach a third destination. There was not enough xenon propellant remaining to orbit Adeona. Nevertheless, the encounter in the second half of 2019 would have been at the unusually low velocity of < 1 km/s, enabling a rich scientific return. Ultimately it was determined that the scientific value of a flyby was not as great as the continuing study of Ceres. Therefore, when the prime mission concluded on 30 June, NASA approved the first extended mission (XM1) for more Ceres science. Dawn continued in the same low altitude orbit, designated XMO1 starting on 1 July. The GRaND, gravity, VIR, and FC2 observation campaigns continued. Prior to completing the prime requirements, Dawn had never attempted targeted observations (other than acquiring multiple FC2 exposures for Occator Crater), and none had ever been planned. Dawn 11

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

was a mapping mission, designed to conduct the first investigation of the two largest unexplored bodies inside the orbit of Neptune. When targeted observations for VIR and FC2 color images began in LAMO, they were under the most challenging conditions of the mission because of the orbit perturbations. As LAMO/XMO1 progressed, strategies for making late adjustments to timing of data acquisition were developed, thus improving the probability of capturing selected targets. Despite the flawless operation of the RWAs in LAMO/XMO1, their reliability was still considered low. Although Dawn exceeded expectations by operating as long as it had, the remaining lifetime at low altitude was strongly limited by hydrazine. The project recognized that returning to a higher altitude, something not even contemplated prior to the middle of 2016, could provide a greater scientific return. There were several benefits. Rather than operate for a few more months at low altitude to increase the signal for the nuclear spectra, GRaND could operate for much longer at high altitude to improve the measurements of cosmic ray noise. Analysis showed that that strategy would yield a better improvement in the signal-to-noise ratio of the spectra. In addition, targeted observations in LAMO/XMO1 were inefficient, often requiring multiple attempts, but at higher altitude, areas of special interest could be observed with higher confidence in important geometries. The higher altitude also would provide opportunities to look for changes over time. By the end of XMO1 on 2 September 2016, Dawn had completed more than 1,100 orbits in LAMO/XMO1. During its residence at low altitude, the flight system had acquired more than 38,000 images with 30-40 m/px. They provided complete coverage of the surface with the clear filter and views from at least three angles for topography of 95% of the surface. (For context, Dawn's requirement was to acquire topographic data of 80% of the surface at 200 m/px.) The images also included color coverage of such high priority sites as Occator Crater; the cryovolcano Ahuna Mons (see section 5); a region near Ernutet Crater where VIR had found organics (see section 5); and Oxo Crater, which is the second brightest region on Ceres (after Occator) and site of the first clear detection of water on the surface (using infrared spectra from VIR). VIR had acquired 12.0 million visible spectra and 11.5 million infrared spectra. With a smaller field of view than FC2, targeted observations with VIR had been even more challenging, but the data included observations of a number of high priority targets, including Occator, Oxo, and Juling Craters. Ice was observed in Juling. As discussed in section 3.6, Oxo and Juling also were priorities for XMO2, and Juling became a priority for XMO6 (section 3.10). GRaND had integrated neutrons and gamma rays for a total of 183 days, exceeding the original LAMO objective by a factor of 5.2. There were 238,000 Doppler measurements (a total of about 165 days of Doppler data) with accuracy typically between 0.02 and 1.0 mm/s. The expectation upon arrival in LAMO was that Dawn would obtain up to 32,000 Doppler measurements. 3.6 Fifth science orbit: XMO2 The second extended mission orbit (XMO2) was to be similar in altitude to HAMO. The transfer from XMO1 to XMO2 differed from the transfer from HAMO to LAMO in several noteworthy ways. The most significant was that the gravity field was much better known in ascending to XMO2 than in descending to LAMO. The project elected to continue in hybrid control until another RWA fault made that impossible. 12

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

Therefore, during the transfer to higher altitude, the perturbing effects of RCS firings were negligible. The nature of the XMO2 investigations allowed for greater flexibility in the orbit parameters than for previous science orbits. Rather than targeting a particular orbit at the end of the transfer, Dawn ascended from XMO1 by thrusting along the (nominal) velocity vector. The end of the transfer was determined simply by when the altitude would be close to the HAMO altitude and within a range that would yield favorable ground tracks to allow full coverage in about a week. This strategy simplified the design of each thrust segment. After 753 hours of ion thrusting and 92 m/s, thrusting concluded on 6 October in a polar orbit at a mean altitude of 1,480 km. Because Dawn had not targeted a circular orbit, the altitude varied between about 1,445 and 1,510 km, with 40 km of that variation caused by orbital eccentricity. From an observing standpoint, XMO2 resembled HAMO. The principal difference was that the angle (ß) between the orbit plane and the vector to the Sun was very different. In HAMO, ß started at 24° and was allowed to increase naturally to 36°. It was held fixed near 45° in LAMO/XMO1 by controlling inclination, but it was allowed to increase again upon leaving XMO1. During XMO2, ß increased naturally from 54° to 56°. (The smaller change during XMO2 is simply because it lasted for 12 days rather than the 66 days of HAMO.) All the phases at Ceres (and Vesta) prior to XMO2 were designed principally for mapping with one (and usually more) sensor. Other sensors were included to gather as much data as possible. In XMO2, FC2 mapped the entire surface with all filters in order to look for changes in the 12 months since HAMO, but that was not the highest priority. The primary objective of XMO2 was to perform new VIR observations. VIR had detected ice in Oxo Crater in LAMO, but had not observed the crater at all in HAMO. In addition, there was reason to believe at the time that the amount of ice in Juling Crater might change over the course of a Cerean day. Dawn's plan included acquiring infrared spectra at multiple local solar times for each crater with dedicated pointing on different orbital passes. VIR also observed other regions in XMO2 that it had not covered during HAMO. In all cases, FC2 acquired data simultaneously. Dawn also imaged Occator Crater when it was on the limb (Fig. 11) to aid in searching for evidence of haze that might be associated with its extensive faculae. At high northern and southern latitudes, FC2 used long integration times to search for evidence of ice in persistently shadowed regions of craters. Data acquisition was designed to begin on the first orbit that allowed Oxo and Juling to be observed, which turned out to be on 17 October. Until then, hydrazine was conserved by keeping the HGA pointed to Earth. But with the successful completion of the prime mission, two RWAs continuing to operate, and only one extended mission objective after XMO2, the demands on the remaining hydrazine were reduced, so more frequent turns were included in XMO2 than in HAMO. In addition to turns for observing specific targets rather than maintaining a fixed angle relative to nadir, Dawn turned more often between pointing instruments at Ceres and pointing the HGA at Earth. The spacecraft acquired data during all 16 passes over the lit hemisphere, and the HGA was pointed at Earth during most of the passes over the dark hemisphere.

13

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

During XMO2, VIR acquired 569,000 infrared spectra. FC2 obtained more than 2,900 clear and color images. Despite all the pointing optimized for observing specific targets with VIR and FC2, images covered more than 98% of the surface. 3.7 Sixth science orbit: XMO3 By the conclusion of XMO2, Dawn had met all of the formal requirements established for XM1 except one: measuring the cosmic ray background to reduce the noise in the LAMO/XMO1 nuclear spectra. Above 7,200 km, the contribution from Ceres to GRaND's measurements would be less than 1%, low enough to allow accurate quantification of the background. The only requirement then on XMO3 was that the altitude be greater than 7,200 km. Therefore, the same simple strategy of thrusting along the velocity vector used for the transfer to XMO2 was applied for the transfer to XMO3. Ion thrusting began on 4 November. After 722 hours of thrusting and 95 m/s, thrusting concluded on 5 December in a polar orbit that ranged in altitude from 7,530 km to 9,350 km. Since Dawn had left XMO1, ß had continued to increase, so the lighting on Ceres from the spacecraft's perspective was different from what it had been throughout the mission until then. The only requirement in XMO3 was to measure cosmic rays, but because the two RWAs continued to operate, a small amount of hydrazine could be allocated to additional FC2 and VIR observations of Ceres under this new illumination. In January and February 2017, Dawn observed Ceres three times, each throughout a full Cerean day and each in a different part of the 7.7-day-long orbit. ß ranged from 73° to 79°. Images in the clear plus all color filters were generally acquired every 20 minutes as Ceres rotated (corresponding to intervals of about 13° of longitude), but when features of particular interest were on the limb, the frequency was increased to images every 3 minutes. In the third observation, on 19 February, Dawn used FC1 and FC2 simultaneously for the first time in flight. The reason for carrying redundant cameras was that at least one was required for mission success, both for optical navigation and for the defined minimum science requirements. Prudence always dictated that before one camera could be powered on, the other camera would be powered off, with the cover closed and the filter wheel in the position necessary for navigation and most of the science. FC1 was kept as a cold spare, generally operated twice each year to verify its health and calibrate it. The motivation for simultaneous operation was to verify flight and ground system functionality in preparation for a new, bonus measurement planned for XMO4. The risk of doing this for the XMO3 observations was assessed and found to be acceptable. FC1 and FC2 together obtained almost 1,400 images in XMO3, and VIR acquired 668,000 visible spectra. 3.8 Seventh science orbit: XMO4 The acquisition of cosmic ray background data placed few strong constraints on the spacecraft apart 14

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

from altitude. Therefore, with the mission going very smoothly, the project was able to adopt a new objective. All of the observations of Ceres had been at phase angles (defined as Sun-Ceres-spacecraft angle) of 7°-155°. The plane of Dawn's orbit relative to the vector to the Sun for the first six science orbits is illustrated in Fig. 12. With the capability of the IPS, the mission targeted acquisition of new images in XMO4 near 0°, requiring a plane change of ~ 90° by the time it would be executed in April. Low-phase observations had been ruled out for low altitude orbits, because Dawn was not qualified for repetitive eclipses. An anomaly that prevented normal operation could leave the spacecraft in an orbit that was in Ceres' shadow half a revolution after being at low phase. Moreover, if the recovery time from an anomaly was long compared to an orbit period, eclipses would repeat. In a long-period orbit, however, the natural ß drift of almost 0.2°/day would avoid that problem. Even if Dawn were at 0°, the orbit plane would rotate enough to avoid eclipse by the time the spacecraft completed half a revolution. Indeed, it would require a maneuver to bring the spacecraft back to 0°. A benefit of being at high altitude was that it would allow the entire visible hemisphere to be at very low phase, rather than only a small area. But the altitude needed to be low enough so that Cerealia Facula, the region of greatest interest, would subtend at least three FC pixels. The target was set at 20,000 km at the time that Cerealia Facula was at local solar noon. Although hydrazine consumption was lower during ion thrusting than coasting, so many turns would be involved in reaching the new orbit and acquiring the data that it would consume significantly more hydrazine than remaining with the HGA pointed to Earth. Therefore, when the plans were formulated, their execution was predicated on the continued operation of the RWAs. The transfer was conducted in three deterministic segments plus a purely statistical TCM, shown in Fig. 13. Ion thrusting began on 23 February. After 418 hours of thrusting and 60 m/s, the deterministic phase concluded on 13 April in an orbit with a period of 59 days. Navigational accuracy was more challenging than for other maneuvers at high altitude, so Dawn conducted three dedicated opnavs. The TCM was broken into two segments. The first was executed on 22 April. The five hours of ion thrusting provided 0.7 m/s. The second was planned for 24 April, with four hours of ion thrusting. Two hours before thrusting was scheduled to begin, an RWA failed. Dawn entered one of its safe modes and did not execute the maneuver. As with the two previous RWA failures, telemetry gave no indication of a problem until very close to the time of the failure, much too late for the operations team to intervene or even know. The project never knew the failure mechanism for the RWAs. The RWA lifetimes are often expressed in revolutions, although it was not at all clear whether that had any bearing on the lifetime. The RWA that failed on 24 April had 1.2 billion revolutions (Grev). The final operable RWA was within 3% of that. (The two previous RWA failures occurred at 0.5 Grev and 1.1 Grev.) The flight team quickly returned Dawn to its normal operational configuration in all-RCS control. Studies following the first RWA failure in 2010 concluded that a one-RWA mode was not 15

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

worthwhile. Therefore, the one operable RWA was left powered off, with no plans to operate it ever again. Even without the second segment of the TCM, Dawn was on a trajectory to fly through opposition, although the phase at Cerealia Facula would not be as low as if the TCM had executed. Nevertheless, the value of the data was still recognized to be very high. Dawn observed Ceres at (and near) 0° on 29 April from an altitude of 19,800 km. Although the measurements were a bonus for XM1, and not a requirement, both FCs were used, reducing vulnerabilities to rare instances in which FC2's computer reset and prevented acquisition of data. No resets occurred in XMO4, and all planned observations executed as planned. Dense phase coverage was obtained below 1.5°, and some observations were as high as 6°, filling the gap in the phase curve below 7°. The minimum phase observed at Cerealia Facula was 0.3°. The two FCs obtained a total of 382 images, analysis of which was reported by Schröder et al[19]. In addition, VIR acquired 131,000 visible spectra and almost 7,000 infrared spectra. The operations team had developed a plan for a backup opposition observation in case the first was not fully successful. It would have occurred one revolution later, on 28 June, and required 1.5 days of ion thrusting on 27–28 May in order to counter the natural increase of ß. Because the data for the first observation were so good, the backup was canceled. 3.9 Eighth science orbit: XMO5 As XM1 drew to a close, once again two options were considered for the second extended mission (XM2): remaining at Ceres and departing for a third target. Thanks to the flexibility afforded by the IPS, Adeona remained a viable option and was the most scientifically attractive target within reach. Both options were feasible even with the remaining hydrazine and with no reaction wheel control. By the time Dawn had completed its XM1 objectives at high altitude, there was not enough hydrazine available to return to LAMO/XMO1 and operate long enough to acquire any new, valuable data. Instead, if the spacecraft stayed at Ceres, it would be maneuvered to a highly elliptical orbit with a peridemeter below 200 km, well below the 385-km altitude of LAMO/XMO1. XMO4 had been established for the unique opposition observation. About 14,000 × 53,000 km, the orbit had a period of 60 days. As NASA conducted a thorough review of the two mission options, the project recognized that both XM2 options would be improved by reducing the period. For staying at Ceres, that would put Dawn energetically closer to the low-peridemeter orbit. Departure opportunities for Adeona occurred at intervals of the XMO4 period. Reducing the period would provide more frequent opportunities. Although Dawn had the capability to reach Adeona with departures throughout 2016 and most of 2017, later departures yielded slightly higher encounter velocities (although still about 1 km/s). Enabling a slightly earlier departure, at the expense of a small amount of Xe to change the orbit, yielded a lower encounter velocity at Adeona and hence a better scientific return. With three periods of ion thrusting from 3 June to 23 June 2017, totaling 257 hours, Dawn maneuvered from XMO4 to XMO5, which had a period of 30 days. The 5400 × 38,000 km orbit was aligned to provide optimal performance for a mission to Adeona. (The alignment was unimportant for staying at Ceres.) 16

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

Dawn was in solar conjunction during much of the thrusting. When thrusting commenced, SEP was less than 2°, and it reached as low as 0.7°. As Dawn had already thrust for 59% of its time in space, there were no concerns about thrusting during conjunction, when communications were not possible. In October 2017, NASA approved XM2 for further orbital exploration of Ceres. Dawn had already acquired extensive data there, well beyond what was originally planned, but the dwarf planet was still too interesting to leave even in exchange for a slow flyby of Adeona. Even as NASA was assessing the XM2 proposals, the flight team had already begun detailed trades and investigations of how to maneuver to and operate in highly elliptical orbits with low peridemeters to accomplish new measurements. There were several significant challenges. One of the constraints on the orbit was a planetary protection requirement. Dawn's data showed that Ceres has water, organics, and other astrobiologically important chemicals and has retained some internal heat from early radiogenic heating. To provide for a period of "biological exploration," during which a subsequent mission could study Ceres without risk of contamination by the terrestrial materials on the Dawn spacecraft, the orbital lifetime was required to be at least 20 years. (A common misconception is that the lifetime was based on the time required for the spacecraft to be sterilized by the space environment. That was not a consideration.) In addition to ensuring compliance with planetary protection, analyses were needed to determine how to maneuver to the new elliptical orbits, how to achieve adequate orbit knowledge and control, and what environmental effects might be caused on the spacecraft by going so much closer to the dwarf planet. Moreover, new concepts for science data acquisition and spacecraft operations were needed. For all studies, hydrazine consumption was a crucial metric to ensure operations would last long enough to be productive. As planned, after approval for XM2, six more months of work was needed before Dawn was ready to go to a lower orbit. While Dawn was in XMO5, it continued measurements of the cosmic ray background with GRaND that it had begun in XMO3. These data proved valuable in reducing the noise in the nuclear spectra acquired in LAMO/XMO1. In addition, Dawn undertook a campaign to compare GRaND observations of solar energetic particles with measurements by other instruments nearer Earth. Telescopic observations of Ceres were planned as well to test the finding that the energetic particle impingement on Ceres might create a transient water vapor atmosphere. With an unusually quiet Sun, no significant solar events were observed. 3.10 Ninth science orbit: XMO6 The prime objective of XM2 was to reach an orbit with a peridemeter below 200 km. However, there were other valuable objectives for altitudes above that. Dawn had fully mapped Ceres with its camera, in color and in stereo, from HAMO and acquired another global map in XMO2. It mapped Ceres at higher resolution in LAMO/XMO1, with color and stereo in selected areas. Nevertheless, even with additional VIR observations in XMO2, at an altitude close to HAMO's, VIR had not observed as much of Ceres because of its smaller field of view, higher data volume per unit surface area, and tighter constraints on surface illumination. The Sun had been in Ceres' northern hemisphere for all of the observations from the beginning of the mission through the end of XMO2, 17

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

so the mission had obtained more extensive coverage in the northern hemisphere than the southern. From XM2's planned, low-peridemeter orbit (XMO7), extensive new coverage would not be possible. Therefore, an intermediate orbit was designed to accomplish VIR objectives. XMO6 initially was designed as an extended forced-coast period in the optimal transfer from XMO5 to XMO7. Subsequently, small adjustments to the transfer design allowed XMO6 to be improved to meet specific VIR observation requirements without adding significant time to the transfer. Southern summer solstice was on 23 December 2017. Even with an obliquity of only 4°, illumination in the southern hemisphere during XM2 was better than it had been during the prime mission and XM1. The VIR observations of greatest interest were between 50°S and 75°S, so XMO6 was designed to optimize acquisition of infrared spectra in that range. Ceres' perihelion of 2.56 AU occurred on 28 April 2018. XMO6 also presented an opportunity to compare the surface with what had been observed at other times, including at the 2.98 AU aphelion on 6 January 2016. XMO6 provided the opportunity for other observations as well. A target of special interest was Juling Crater (Fig. 14). With infrared observations spanning six months in 2016, including targeted observations in XMO2 at three different local solar times, the area of ice on Juling's north wall was seen to increase from 3.6 km2 to 5.5 km2[20]. The change is attributed to seasonal changes in insolation. Juling is at 36°S. As the Sun moved south during that time, the floor of the crater experienced more heating and thus released more water vapor. The colder, shadowed north wall acted as a cold trap. Acquiring spectra of Juling again in XMO6 was thus of interest, and that objective constrained the ground track and timing of the orbit. Targeting of XMO6 accounted for the predicted perturbative effects of RCS activity during the orbits that preceded the planned Juling observation. The transfer from XMO5 to XMO6 (shown in Fig. 15) started on 16 April 2018. After 631 hours of ion thrusting and 121 m/s, ion thrusting concluded on 14 May. The 450 km × 4730 km orbit had a period of 37 hours. Science observations were conducted in the 10 orbits from 16 May to 30 May, generally for ~ 6 hours per orbit. In addition to VIR mapping in the southern hemisphere and the targeted observation of Juling, FC and VIR observed other areas, including in the northern hemisphere. Most observations were at nadir, but selected off-nadir attitudes were used to observe targets of particular interest. XMO6 was designed to provide an off-nadir angle for the Juling observation to optimize the view of the crater's north wall. It was placed near the middle of the XMO6 phase to allow time to refine the orbit knowledge after the transfer was complete. The hydrazine cost of turning between pointing the instruments at Ceres and pointing the HGA at Earth was managed carefully. The spacecraft turned to downlink data on only six of the orbits. The long orbit period with much of the time over the nightside of Ceres provided sufficient time to downlink the stored data. On the other orbits, to save hydrazine, Dawn continued pointing at Ceres 18

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

even when altitude and illumination precluded the acquisition of useful data. Both cameras were used for the majority of XMO6, again to maximize the probability that the desired data would be acquired. No FC2 resets occurred in XMO6, and together the two cameras provided more than 1,800 images. On three orbits when there was data volume to spare, following the completion of science observations and before turning to point the HGA at Earth, the spacecraft turned to point the cameras at Ceres' limb. The direction of the turn was chosen to minimize the hydrazine cost. Imaging continued from nadir to the limb, yielding not only new images for topography (which had been a significant part of HAMO and XMO1) but also appealing views of the limb. Examples are in Fig. 16. Despite significant uncertainties in predicting the ground track, principally from errors in predictions of RCS-induced perturbations during the transfer to and operation in XMO6, all measurement objectives were met. In addition to the images with FC1 and FC2, Dawn acquired 2.3 million infrared spectra with VIR. 3.11 Tenth science orbit: XMO7 XMO7 was the final orbit of the entire mission. Dawn was recognized to be so low on hydrazine, and the rate of use would be so high in XMO7, that the project could not count on further orbit changes. Moreover, the orbit was intended to have a sufficiently low peridemeter that as long as Dawn could operate there, valuable new data could be acquired. The primary objective of XMO7 was to reach a low enough altitude to make significant improvements in both spatial resolution and sensitivity of nuclear spectroscopy. The lowest feasible peridemeter would yield the best science. Improvements in spatial resolution of gravity measurements, imaging, and reflectance spectroscopy were of great interest as well. Trade studies considered orbits with peridemeters as high as 200 km and as low as 30 km. The lowest peridemeter was attractive for all investigations, but the peridemeter chosen was 35 km to increase margin for the planetary protection requirement of no impact within 20 years. See Fig. 17. The planetary protection analysis was based on Monte Carlo studies that included the best 18 × 18 gravity field with conservative uncertainties and RCS activity during operations. The 35-km peridemeter had no impacts in 1,150 samples propagated for 20 years. Additional studies provided > 99% confidence of an orbital lifetime of 50 years[21]. The selection of apodemeter involved many issues. A high apodemeter would reduce the time needed for the orbit transfer. In addition, for nuclear spectroscopy, it was desirable to have an apodemeter high enough that GRaND could measure the cosmic ray background on each orbit. However, a higher apodemeter translated into greater velocity at peridemeter, thus complicating science data acquisition, increasing the hydrazine cost of pointing the instruments, and increasing FC and VIR smear. Moreover, longer orbit periods would yield a lower duty cycle for science data acquisition, an important consideration given the limited time before the hydrazine would be depleted.

19

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

The team recognized that Ceres' gravity field would cause the latitude of peridemeter to shift south over the dayside with each orbit. That was valuable for the investigation of the depth of ice as a function of latitude. GRaND had previously demonstrated that ice is shallower at higher latitudes. An area of special interest for high resolution observations with all instruments was Occator Crater[22]. And within Occator, Cerealia Facula and Vinalia Faculae presented the most scientifically attractive targets. Uncertainty in the extensive RCS activity near peridemeter made orbit prediction uncertainties significant. With standard procedures, the probability of acquiring images and spectra of specific targets was low. Credible timing uncertainties of a few minutes at the time of sequence development (weeks before execution) meant Ceres' rotation could move features by distances large compared to the FC and VIR fields of view. Occator is at 20°N latitude. With an apodemeter chosen to provide a 3:1 synchronous orbit with Ceres' rotation (generally called a resonant orbit by the Dawn project), Dawn would have multiple opportunities to observe Occator at very low altitude before peridemeter shifted too far south. Therefore, an orbit period of 27 hours was chosen, dictating an apodemeter of 4,000 km given the peridemeter of 35 km. The longitude of peridemeter was at Occator's longitude. Fig. 18 illustrates XMO7. The transfer to XMO7 began on 31 May. After 132 hours of ion thrusting and 26 m/s, ion thrusting concluded on 6 June. The transfer is shown in Fig. 19. Science data acquisition with all instruments began on 9 June. As in XMO6, the cameras were used simultaneously. Each peridemeter had a unique ground track, and so near the end of the mission, the reduction in risk of a camera reset that prevented imaging outweighed the small increase in risk of simultaneous operation. Moreover, even at the fastest rate of one image every seven seconds, it was not guaranteed that images from one camera would overlap, given the 95-mrad field of view. Therefore, the timing of the two cameras was offset to minimize gaps. In most cases early in XMO7, Dawn downlinked data after every other peridemeter, conserving hydrazine in some orbits by not turning to point the HGA at Earth. The initial latitude of peridemeter was chosen so the latitude of peridemeter would be nearest Cerealia Facula on 23 June and again on 24 June. That provided sufficient time to characterize the behavior of the orbit to target observations. Windows had been built into the plan for a pair of TCMs in preparation for the Cerealia Facula observations. During ion thrusting, the planned attitude was determined by the time-dependent thrust vector, which would not be compatible with pointing instruments at Ceres. Each peridemeter was so scientifically valuable that the TCM windows were scheduled not to interfere. On 21 June, Dawn thrust for 127 minutes and then 15 hours later for 71 minutes. The combined 0.55 m/s adjusted the orbit to bring the flight path over Cerealia Facula on 23 and 24 June. The TCM was the last planned use of the IPS for the entire mission. The orbit remained compliant 20

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

with planetary protection requirements, and the ground track was sufficient to yield excellent science opportunities for as long as the hydrazine lasted. All post-launch trajectory control was accomplished with the IPS, apart from a gravity assist at Mars[8]. The IPS provided 11.5 km/s over its 51,385 hours of accumulated thrust time. Uninterrupted thrust periods ranged in duration from 11 minutes for a statistical maneuver in orbit around Vesta to 31.2 days during the interplanetary cruise from Vesta to Ceres. The TCM executed correctly. Nevertheless, uncertainty in orbit perturbations induced by RCS activity were too large to be accommodated by the normal schedule of updating the onboard data needed for pointing. The project developed a streamlined procedure that allowed for a rapid update. Data acquisition for the 22 June peridemeter began three hours after the thrusting completed. Following three hours of normal peridemeter science observations, Dawn turned to point its HGA to Earth. Downlink with the HGA provided three data types of value for the update to the pointing: radiometric tracking, telemetry on RCS activity (used in fitting the orbit), and images. Dawn was so much closer to the ground in XMO7 than it had been in LAMO/XMO1, even the uncertainty in the location of Cerealia Facula itself was not negligible. There was not enough time for an optical navigation solution to be incorporated into the orbit prediction, but rapid visual inspection of the images revealed part of Cerealia Facula (and showed exciting details, as seen in Fig. 25), thus helping to establish the ground track relative to the target. All prior peridemeter observations had pointed the instruments to nadir (with nadir determined by the onboard ephemeris). On 22 June, a new ephemeris was developed for uplink to the spacecraft. The operating sequence had been built to allow a late change in the pointing relative to nadir for the two subsequent peridemeters. When it was determined to be probable that Dawn would not fly directly over Cerealia Facula, new commands with the appropriate angular offsets were developed and uplinked. The quick update was successful, yielding high resolution data on the two targeted peridemeters as well as the one between the TCM and those targeted orbits. Following the campaign to observe Cerealia Facula, operations transitioned to a different cadence. To reduce hydrazine use, Dawn acquired and stored data on five consecutive peridemeters. Following the fifth, it turned to Earth and downlinked data for nearly two full revolutions, ending in time to turn again for peridemeter. That meant that for one peridemeter in every six, instruments would not be pointed at nadir. Keeping the HGA on Earth through peridemeter improved the radiometric data for gravity science compared to using a low gain antenna. More importantly, using less hydrazine extended the duration of XMO7, providing opportunities to acquire high resolution data farther south as the latitude of peridemeter continued to move. Even with RCS forces, the orbit period remained sufficiently close to the desired 3:1 resonance that Dawn continued to pass over or very near Occator on every orbit. Small shifts east or west allowed greater coverage within the crater. In addition, more opportunities to point off-nadir were built in and used to gain additional data on Cerealia Facula. 21

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

By 12 August, the peridemeter had shifted so far south that Dawn was at LAMO/XMO1 altitude when it was as far north as Occator. Prior to reaching XMO7, the project had considered investigating additional maneuvers after acquiring high-resolution data at Occator to shift the longitude to some other geologically interesting site farther south. However, it was quickly realized that the ground track would put peridemeter in Urvara Crater, 65° south of Occator. Urvara (Fig. 20) was so valuable for high-resolution study that maneuvers to other sites were not needed. Occator Crater may be ~ 20 million years old[23]. The 170-km-diameter Urvara is much older, providing an opportunity for comparison of relatively young and relatively old materials. Dawn's peridemeter continued to move south until 26 August when it reached 84°S, corresponding to the orbital inclination. By early September, lighting precluded imaging or reflectance spectroscopy below LAMO/XMO1 altitude. Nevertheless, high resolution nuclear spectroscopy and gravity measurements continued productively. Dawn's final imaging of Ceres occurred on 1-2 September. Peridemeter was then in darkness, but the spacecraft acquired images at higher altitude. Two of the last images are shown in Fig. 21. In XMO7, Dawn returned more than 3 million infrared spectra, almost 50,000 visible spectra, and more than 11,000 images. The XMO7 images brought the total for Ceres to about 70,000 and the total for Vesta plus Ceres to more than 100,000. To complete the reflectance spectroscopy and imaging science, VIR and both cameras were calibrated one final time on 29 September. 4 End of mission It was well understood from the beginning of XM2 that XMO7 would last no more than a few months before the hydrazine would be depleted, terminating mission operations. The date was uncertain because of the uncertainties in the amount of hydrazine onboard, the amount that was unusable, and the rate of use. By August, the most probable exhaustion date was found to lie between mid-September and mid-October, but earlier and later dates were quite credible. Indeed, given the uncertainty in the hydrazine onboard, the usable mass was then already consistent with 0. Unlike some spacecraft, Dawn was not expected to show specific evidence of the hydrazine nearing depletion other than the smoothly declining propellant pressure. Rather, the spacecraft was predicted to operate normally until an RCS thruster was commanded to fire when there was insufficient hydrazine available. The spacecraft would then lose attitude control. The plan was to continue acquisition of nuclear spectra and gravity measurements on every peridemeter as long as possible. As it turned out, pressure telemetry transmitted through an LGA on 23 October was interpreted to mean depletion was imminent. There was no basis for changing plans, however, so operations continued normally, with regular science data acquisition and another HGA session. Dawn was not tracked with the DSN for most of its time in XMO7, but it was tracked near every

22

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

peridemeter. About 20 minutes after peridemeter on 31 October, as the DSN was tracking the downlink carrier for gravity science, variations in the received power showed evidence of the loss of attitude control. Upon losing attitude control, the spacecraft executed several fault responses, but with no hydrazine, none of them could be effective. The design of the fault protection system was such that the spacecraft's transmitter would be powered off until it was able to attain a stable attitude with the solar arrays illuminated, but that condition could not be achieved. (In addition, with reasonable attitude rates after depletion, the time-averaged power from the solar arrays would not be enough to charge the battery.) The observed behavior of the downlink, culminating in its early termination on that DSN pass, was consistent with the depletion of the hydrazine. Although the hydrazine had already lasted longer than expected by that time, two subsequent DSN passes at different stations (and, as it happened, at different complexes) were used to search for the signal in the very unlikely case that there had been a different reason for the lost downlink and that the spacecraft had subsequently been able to recover. When no signal was observed on a pass that concluded about 18 hours later, no reasonable doubt remained. Dawn had, by then, returned four times as much GRaND data in XMO7 as had been required, including 140 hours of nuclear spectra from below LAMO/XMO1 and 50 hours from below 100 km. In addition, about 6400 two-way Doppler measurements (more than 106 hours) from below the LAMO/XMO1 altitude were acquired on 103 orbits. As the previous gravity data, the accuracy was 0.02-1 mm/s. 5 Ceres highlights The wealth of data Dawn acquired at Ceres yielded many discoveries about the first dwarf planet discovered or visited. More than 125 papers analyzing or interpreting the data have been published already and many more are nearing publication. There have been more than 400 conference presentations as well. There are too many findings even to summarize here. In addition, some of the data were acquired so recently, especially the high resolution observations in XMO7, that it is too soon to turn them into knowledge. Nevertheless, here we describe selected highlights. Hundreds of faculae have been catalogued on Ceres. Occator Crater (Fig. 22) has the most prominent, with regions so reflective that they were visible as a single bright spot in HST's images. Ceres' average albedo is 0.094[19], but the albedo of Cerealia Facula is about 0.6 [25]. The 92-km-diameter crater is relatively young, with a most likely age of about 20 million years[23]. The total area of the faculae is small, so age-based crater counting is less accurate, but Cerealia Facula could be less than 6 million years[26] and Vinalia Faculae may even be less than 1 million years[27], suggesting that the material has been emplaced recently. Cerealia Facula's central pit is surrounded by a dense network of faults which may be a result of long-term tectonic uplift (Figs. 10, 22, and 24). Subsurface briny water came to the surface after the impact (and, it appears, long after the impact). Ice is not stable at such a low latitude on Ceres. It would sublimate quickly, but the dissolved salts 23

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

would remain, and they constitute the highly reflective material. The faculae are sodium carbonate, a salt observed only on Ceres, Earth, and Enceladus. (When it was deposited, it may have been sodium bicarbonate.) It is also interesting to note that Occator Crater contains the highest concentration of any kind of carbonates known on kilometer-scales anywhere in the solar system except Earth[28]. Another carbonate, dolomite, is omnipresent on Ceres and mixed with serpentine[29]. Those minerals are formed by chemical interactions between rocks and water under pressure. Their formation underground and subsequent transport to the surface are unlikely given how widespread they are. The temperature difference between the equator and high latitudes is large enough that convective forces should have yielded a latitude-dependent abundance that is not observed. There are several alternative explanations. One, supported by other lines of evidence as well, is that early in its history, Ceres had enough radiogenic heating and water to have a global liquid ocean. An ocean ~ 5 km deep would have provided sufficient pressure to make the minerals Dawn identified[30], although it could have been much deeper. As Ceres aged and cooled, some of the water would have ended up in minerals and some would have frozen, creating an ice shell. The ice would have subsequently been lost in a geologically short time both through sublimation and by impacts. Some of the minerals that had been in that ancient ocean remain on the surface[31]. Phyllosilicates are another class of minerals ubiquitous on Ceres that form through the interaction of water and rock[32]. The phyllosilicates show distinctive evidence of ammoniation. Ammonia should have been common in the solar nebula, but conditions would have been too warm for it to condense and be incorporated into planetesimals at Ceres' current heliocentric distance. It isn't clear whether this indicates Ceres formed farther from the Sun and subsequently moved to its present location or rather that it formed near where it is now but accreted material that had formed farther away[29]. Many lines of evidence point to Ceres containing a substantial fraction of water, about 25% by mass, most or all of which is frozen. The dwarf planet is partially differentiated, with a rocky interior and volatile rich shell of water ice, phyllosilicates, carbonates, salts, and clathrate hydrates[33]. Whether liquid persists in Ceres' interior is an open question. Thermal models allow it but do not guarantee it. Ammoniated salts lower the freezing point of water. Ceres’ crust would need to contain a large fraction of insulating materials, such as gas hydrates, to slow the heat loss and maintain a relict ocean at the base of the crust (~40 km deep), which has been suggested from geological analyses[34]. There is one particularly interesting structure that indicates subsurface liquid may still be present. The tallest mountain on Ceres is Ahuna Mons, a 4-km-high cryovolcano[35]. (See Figs. 21 and 23.) It formed from a highly viscous but partially melted cryomagma composed of brines, carbonates and other minerals, and water ice. The edifice formed in a few hundred to a few hundred thousand years and is 50–240 million years old. The likely presence of a subsurface melted phase so recently suggests that there may be some now as well. More than 20 other Cerean domes have been identified as older cryovolcanoes, with present shapes explained by relaxation of structures that originally were shaped like Ahuna Mons[36]. On average, 24

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

a new Cerean cryovolcano may have formed every 50 million years for more than 1 billion years. In sections 3.6 and 3.10, we mentioned the observation of ice in Juling and Oxo Craters, but ice was observed on the ground at multiple other locations as well[37]. It should not be stable near the surface except at high latitudes and in persistently shadowed regions (PSRs), so it must have been exposed or delivered recently in those locations. It also was indeed found in some PSRs[38]. (Ceres' obliquity is only 4°. It may vary over a range of 2° to 20° in as little as 12,000 years, so many regions that are persistently shadowed throughout a Cerean year are not permanently shadowed on geological timescales[39].) A large number of geological features best explained as ground ice flows also have been identified[40]. Neutron measurements that probe to decimeters show that Ceres is rich in hydrogen, a strong indication that water is abundant[41]. Some is ice and some is bound in hydrated minerals. The measured concentration increases with increasing latitude. Lower temperatures at higher latitudes would allow ice to persist closer to the surface for geologically long times without sublimating away. Studies using models of the time-dependent flux of impactors combined with the measured size distribution of craters conclude that Ceres is deficient in large craters[42]. Vesta, the Moon, and other bodies seem to preserve their craters for much longer. Investigations of crater morphologies show crater relaxation over time. These provide support for ice also being present at greater depths, which would lower the viscosity. The infrared signature of alkanes, a class of organic materials, was found in a region of ~ 1000 km2 in and near Ernutet Crater[43]. This represents the largest area of extraterrestrial organics discovered in the solar system except for Titan. A smaller area near Inamahari Crater, ~ 400 km away, also has spectra consistent with organics. Dawn's remote sensing instruments did not detect any evidence of the exospheric water vapor observed by Herschel, despite the dedicated search in the first science orbit (section 3.2). Küppers et al. [17] noted that some searches with Herschel, the International Ultraviolet Explorer, and the Very Large Telescope had seen evidence for water vapor production and others had not. One interpretation was that the phenomenon was dependent on heliocentric range, although that was not definitive. During Dawn's approach to Ceres, GRaND detected energetic electrons. In survey orbit, energetic electrons were observed in the same location on three consecutive revolutions. These serendipitous findings led to a new explanation for Ceres' transient exosphere: it is driven largely by solar energetic proton events that sputter water from the surface. This mechanism accounts for the electrons, which were reflected from the bow shock, and correlates with solar proton activity both for prior detections and non-detections of water vapor[44]. XM2 data are too new for new, meaningful conclusions to have been reached about Ceres. The improvement in spatial resolution from LAMO/XMO1 at 385 km to XMO7 peridemeter at 35 km was the largest of the mission at Vesta or Ceres. To conclude this section, we present in Figs. 24-31 a selection of the high-resolution images taken with the 93-µrad/pixel cameras and wait for further analyses before offering any interpretations or new conclusions about Ceres. 25

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

6 Conclusion Dawn mission operations during the 3.6 years in orbit around Ceres yielded a wealth of data from 10 science orbits. The different orbital phases had different objectives, different challenges, and thus different designs, but all were productive and successful. When Dawn arrived at the dwarf planet, only the first four of the orbital phases had been planned. Despite the prior loss of two reaction wheels and the loss of another while at Ceres, the spacecraft operated far longer than anticipated. Dawn was designed to use three reaction wheels (and carried four), but it used only two for 13% of its operational life and zero for 50% of its operational life. After more than 11 years of flight operations, including 14 months in orbit around Vesta, Dawn's mission concluded. (Some lessons from the mission are described by Rayman[45].) Upon expending the last of the hydrazine, the spacecraft became an inert celestial monument to human creativity, ingenuity, and curiosity, a lasting reminder in orbit around the dwarf planet it unveiled that our passion for bold adventures and our noble aspirations to extend our reach into the universe can take us very, very far beyond the confines of our humble planetary home.

Acknowledgments Dawn's many successes in reaching Ceres and conducting complex operations to acquire a trove of important data there were achieved through the skill and dedication of the members of the operations team, with support from the instrument and science teams. Their impressive work is acknowledged with the greatest respect and gratitude. The work described in this paper was carried out in part at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

References 1. Kuchynka, P., W.M. Folkner, A new approach to determining asteroid masses from planetary range measurements, Icarus 222 (2013). https://doi.org/10.1016/j.icarus.2012.11.003. 2. Rayman, M.D., R.A. Mase, Dawn's exploration of Vesta, Acta Astronautica 94 (2014). https://doi.org/10.1016/j.actaastro.2013.08.003. 3. Polanskey, C.A., S.P. Joy, C.A. Raymond, Efficacy of the Dawn Vesta Science Plan, Paper 1275915, SpaceOps 2012 Conference Proceedings, 2012. https://doi.org/10.2514/5.9781624102080.0501.0516 4. Russell, C.T. et al., Dawn at Vesta: testing the protoplanetary paradigm, Science 336 (2012) (and references therein). https://doi.org/10.1126/science.1219381

26

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

5. Rayman, M.D., T.C. Fraschetti, C.A. Raymond, C.T. Russell, Dawn: A mission in development for exploration of main belt asteroids Vesta and Ceres, Acta Astronautica 58 (2006). https://doi.org/10.1016/j.actaastro.2006.01.014. 6. Russell, C.T. et al., Dawn Mission to Vesta and Ceres: Symbiosis between Terrestrial Observations and Robotic Exploration, Earth Moon Planets 101, p. 65 (2007). https://doi.org/10.1007/s11038-007-9151-9

7. Rayman, M.D., K.C. Patel, The Dawn project's transition to mission operations: On its way to rendezvous with (4) Vesta and (1) Ceres, Acta Astronautica 66 (2010). https://doi.org/10.1016/j.actaastro.2009.05.012. 8. Rayman, M.D., R.A. Mase, The second year of Dawn mission operations: Mars gravity assist and onward to Vesta, Acta Astronautica 67, p. 483 (2010). https://doi.org/10.1016/j.actaastro.2010.03.010 9. Rayman, M.D., R.A. Mase, Dawn's operations in cruise from Vesta to Ceres, Acta Astronautica 103 (2014). https://doi.org/10.1016/j.actaastro.2014.06.042. 10. Rayman, M.D., R.A. Mase, Preparing for Dawn's Mission at Ceres: Challenges and Opportunities in the Exploration of a Dwarf Planet, 65th International Astronautical Congress, 29 September–3 October 2014, Toronto, Canada, Paper IAC-14.A3.4.5x21592. 11. Rayman, M.D., S.N. Williams, Design of the First Interplanetary Solar Electric Propulsion Mission, Journal of Spacecraft and Rockets 39 (2002). https://doi.org/10.2514/2.3848. 12. Rayman, M.D., T.C. Fraschetti, C.A. Raymond, C.T. Russell, Coupling of system resource margins through the use of electric propulsion: Implications in preparing for the Dawn mission to Ceres and Vesta, Acta Astronautica 60 (2007). https://doi.org/10.1016/j.actaastro.2006.11.012. 13. Bruno, D., Contingency Mixed Actuator Controller Implementation for the Dawn Asteroid Rendezvous Spacecraft, AIAA 2012-5289, AIAA SPACE 2012 Conference & Exposition, 11–13 September 2012, Pasadena, CA. https://doi.org/10.2514/6.2012-5289 14. Polanskey, C.A., S.P. Joy, C.A. Raymond, M.D. Rayman, Architecting the Dawn Ceres Science Plan, SpaceOps 2014 Conference Proceedings, 2014. https://doi.org/10.2514/6.2014-1720 15. Thomas, P.C. et al., Differentiation of the asteroid Ceres as revealed by its shape, Nature 437 (2005). https://doi.org/10.1038/nature03938. 16. McFadden, L.A., et al., Dawn mission's search for satellites of Ceres: Intact protoplanets don't have satellites, Icarus 316, p. 191 (2018). https://doi.org/10.1016/j.icarus.2018.02.017 17. Küppers, M. et al., Localized sources of water vapor on the dwarf planet (1) Ceres, Nature 505 (2014). https://doi.org/10.1038/nature12918. 18. Polanskey, C.A., S.P. Joy, C.A. Raymond, M.D. Rayman, Dawn Ceres Mission: Science 27

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

Operations Performance, SpaceOps 2016 Conference, 16–20 May 2016, Daejeon, Korea, Paper AIAA 2016-2442. https://doi.org/10.2514/6.2016-2442 19. Schröder, S.E., et al., Ceres' opposition effect observed by the Dawn framing camera, Astron. Astrophys. 620 (2018). https://doi.org/10.1051/0004-6361/201833596 20. Raponi, A. et al., Variations in the amount of water ice on Ceres’ surface suggest a seasonal water cycle, Sci. Adv. 4 (2018). https://doi.org/10.1126/sciadv.aao3757. 21. Grebow, D., N. Bradley, B. Kennedy, Stability and targeting in Dawn's final orbit, 29th AAS/AIAA Space Flight Mechanics Meeting, 13-17 January 2019, Ka'anapali, HI, Paper AAS 19254. 22. Scully, J.E.C. et al., Introduction to the special issue: The formation and evolution of Ceres’ Occator crater, Icarus 320 (2019) 1. https://doi.org/10.1016/j.icarus.2018.02.029, and other papers in that special issue. 23. Neesemann, A. et al., The various ages of Occator crater, Ceres: Results of a comprehensive synthesis approach, Icarus 320 (2019), 60. https://doi.org/10.1016/j.icarus.2018.09.006. 24. Jaumann, R. et al., Topography and Geomorphology of the Interior of Occator Crater on Ceres, #1440, 48th Lunar and Planetary Science Conference, 20–24 March 2017, Houston, TX. 25. Schröder, S.E. et al., Resolved spectrophotometric properties of the Ceres surface from Dawn Framing Camera images, Icarus 288 (2017), 201. https://doi.org/10.1016/j.icarus.2017.01.026. 26. Nathues, A. et al., Evolution of Occator Crater on (1) Ceres, Astron. J. 153 (2017), 112. https://doi.org/10.3847/1538-3881/153/3/112. 27. Neesemann, A. et al., Revisiting the Cerealia and Vinalia Faculae on Ceres, EPSC Abstracts 12 (2018), EPSC2018-1254. 28. DeSanctis, M.C. et al., Bright carbonate deposits as evidence of aqueous alteration on (1) Ceres, Nature 536 (2016), 54. https://doi.org/10.1038/nature18290 29. DeSanctis, M.C. et al., Ammoniated phyllosilicates with a likely outer Solar System origin on (1) Ceres, Nature 528 (2015) 241. https://doi.org/10.1038/nature16172 30. Castillo-Rogez, J.C. et al., Insights into Ceres’s evolution from surface composition, Meteorit. Planet. Sci. 53 (2018), 1820. https://doi.org/10.1111/maps.13181. 31. Castillo-Rogez, J.C. et al., Loss of Ceres' Icy Shell from Impacts: Assessment and Implications, #1903, 47th Lunar and Planetary Science Conference, 21-25 March 2016, Houston, TX. 32. Ammannito, E. et al., Distribution of phyllosilicates on the surface of Ceres, Science 353 (2016) 1006. https://doi.org/10.1126/science.aaf4279. 33. Park, R.S. et al. A partially differentiated interior for (1) Ceres deduced from its gravity field 28

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

and shape, Nature 537. https://doi.org/10.1038/nature18955 (2016). 34. Castillo-Rogez, J.C., et al. Conditions for the Long‐Term Preservation of a Deep Brine Reservoir in Ceres, Geophys. Res. Lett. 46 (2019). https://doi.org/10.1029/2018GL081473. 35. Ruesch, O. et al., Cryovolcanism on Ceres, Science 353 (2016). https://doi.org/10.1126/science.aaf4286. 36. Sori, M.M. et al., Cryovolcanic rates on Ceres revealed by topography, Nat. Astron. 2 (2018), 946. https://doi.org/10.1038/s41550-018-0574-1. 37. Combe, J.-P. et al., Detection of local H2O exposed at the surface of Ceres, Science 353 (2016) https://doi.org/10.1126/science.aaf3010. 38. Platz, T. et al. Surface water-ice deposits in the northern shadowed regions of Ceres, Nat. Astron. 1 (2016) https://doi.org/10.1038/s41550-016-0007. 39. Ermakov, A. et al., Ceres's obliquity history and its implications for the permanently shadowed regions, Geophys. Res. Lett. 44 (2017) https://doi.org/10.1002/2016GL072250. 40. Schmidt, B.E. et al., Geomorphological evidence for ground ice on dwarf planet Ceres, Nat. Geosci. 10 (2017) https://doi.org/10.1038/ngeo2936. 41. Prettyman, T.H. et al., Extensive water ice within Ceres’ aqueously altered regolith: Evidence from nuclear spectroscopy, Science 355 (2017). https://doi.org/10.1126/science.aah6765. 42. Marchi, S. et al., The missing large impact craters on Ceres, Nat. Commun. 7 (2016). https://doi.org/10.1038/ncomms12257. 43. DeSanctis, M.C. et al., Localized aliphatic organic material on the surface of Ceres, Science 355 (2017). https://doi.org/10.1126/science.aaj2305. 44. Villarreal, M.N. et al., The dependence of the Cerean exosphere on solar energetic particle events, ApJL 838 (2017). https://doi.org/10.3847/2041-8213/aa66cd. 45. Rayman, M.D., "The Successful Conclusion of the Dawn Mission: Important Results with a Flashy Title," 70th International Astronautical Congress, 21-25 October 2019, Washington, D.C., Paper IAC-19.A3.4A.1.x51360. Figure Captions Figure 1. Dawn's interplanetary trajectory. The trajectory is blue where the spacecraft thrust with its IPS and black where it coasted. Thrusting in orbit around Vesta and Ceres is not shown. The regular interruptions in thrust are for conducting activities incompatible with optimal thrusting, usually pointing the high gain antenna (HGA) to Earth. Most are exaggerated in duration in this figure. For most of the cruise to Vesta, the spacecraft thrust 95% of a typical week, and the remaining time was devoted to turning and pointing the HGA to Earth. Note the reduction in the frequency of such

29

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

telecommunications sessions following departure from Vesta, as explained in the text.

Figure 2. Dawn's trajectory to capture at Ceres and then to the first science orbit. The targeted circular orbit (labeled RC3 and described in section 3.2) is shown in red and is described in section 3.1. The lower curve shows the plan before the September 2014 interruption in ion thrust, and the upper curve shows the revised plan, which Dawn did implement. Ceres' north pole is up and the Sun is to the left (Ceres' orbital motion is into the plane of the figure). The orange and blue curves are solid for ion thrusting and dotted for coasting. Each trajectory ends in a red diamond at insertion into the science orbit. On the actual (blue) trajectory, Dawn initially was out of the plane of this figure (closer to the reader) so that Ceres was ahead of it in orbit around the Sun. Then after capture, as it continued to higher altitude and then descended, it thrust to move into the plane and raise its orbital inclination around Ceres.

Figure 3. Opnav 1 image. This was acquired at a range of 383,000 km, so each pixel is about 36 km. Note the bright spot, later determined to be a collection of bright features in Occator Crater.

Figure 4. Ceres imaged from RC3 orbit. Occator Crater (upper left), with the brightest features, is 92 km in diameter.

Figure 5. Transfer from RC3 to survey orbit. The two science orbits are shown in red, and the transfer trajectory is in blue (with solid for ion thrusting and dashed for coasting). The first two coast segments, both on the left, are for opnavs 8 and 9 (and the subsequent downlinks), the final dedicated opnavs of the prime mission.

Figure 6. Ceres' limb imaged from survey orbit. Occator Crater is readily recognizable.

Figure 7. Transfer from survey orbit to HAMO. The two science orbits are shown in red, and the transfer trajectory is in blue. The transfer required 22 revolutions.

Figure 8. Occator Crater observed in HAMO. Cerealia Facula is in the center of the 92-km-diameter crater, and the other reflective areas are collectively named Vinalia Faculae. This is a combination of two HAMO images, one with an integration time for typical Ceres scenes and one with a short integration time for the faculae.

Figure 9. Transfer from HAMO to LAMO. The two science orbits are shown in red, and the transfer trajectory is in blue. The deterministic part of the transfer required 118 revolutions.

Figure 10. Mosaic of two LAMO images in Occator Crater. Cerealia Facula is on the left, and Vinalia Faculae is on the right. 30

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

Figure 11. Occator Crater as observed in XMO2.

Figure 12. Dawn orbit planes and relative altitudes. Looking down on Ceres' north pole, this illustrates the mean ß during imaging in the first six science orbits, numbered chronologically: 1 RC3; 2 - survey; 3 - HAMO; 4 - LAMO/XMO1; 5 - XMO2; 6 - XMO3. Orbits 1, 2 and 6 all extend off the figure to the lower right. The dotted section of XMO3 illustrates the range of altitudes in the elliptical orbit.

Figure 13. Transfer from XMO3 to XMO4. XMO3 is the inner red orbit. The three deterministic thrust segments plus the two-day window for the TCM are shown in solid blue, and coasting periods are dashed. In each figure, + is Dawn's location when Ceres was at opposition. (a) View from near Ceres' north pole with the sun to the left. (b) View from near the Sun (and hence near Cere's equator). (c) View from near Ceres' equator with the Sun to the left.

Figure 14. Juling Crater. The crater is 20 kilometers in diameter. (a) This image was acquired in XMO1. (b) This perspective was synthesized with multiple images acquired in LAMO/XMO1.

Figure 15. Transfer from XMO5 to XMO6. The two science orbits are shown in red, and the transfer trajectory is in blue.

Figure 16. XMO6 limb views. Image (a) was acquired on 16 May 2018. The limb is about 800 km away. The large crater in the foreground is 27 km in diameter. The crater seen nearly edge-on on the limb at upper left is 120 km away from it and is 31 km in diameter. Image (b) was acquired on 30 May 2018. The limb is about 730 km away. The largest crater is 35 km in diameter. The distance from the lower right corner of the picture to the limb is about 165 km.

Figure 17. XMO7 peridemeter. The peridemeter altitude was 35 km, compared to Ceres' mean radius of 470 km. Dawn revolved counterclockwise in this perspective, traveling north at low altitude on the dayside. Velocity at peridemeter was 0.47 km/s. Dawn spent about 1.25 hours below 385 km, which was the previous lowest altitude (in LAMO/XMO1). The XMO7 peridemeter altitude was 9% of the LAMO/XMO1 altitude. Fig. 18 illustrates the entire orbit. Dawn acquired this image of Ceres as part of the opposition observation in XMO4. Note Occator Crater at the center.

Figure 18. XMO7. As in Fig. 17, Dawn revolved counterclockwise in these illustrations. The peridemeter's 1.9° southward shift with each revolution is shown in (a). One revolution is isolated in (b) for clarity.

31

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

Figure 19. Transfer from XMO6 to XMO7. The two science orbits are shown in red, and the transfer trajectory is in blue.

Figure 20. Urvara Crater. Dawn acquired this image in HAMO. Urvara does not have a central peak but rather a central ridge. Note the fracturing on the crater floor east of the center. Details in the crater are shown in Figs. 30-31.

Figure 21. These two are among Dawn's last images of Ceres, acquired on 1 September 2018. Image (a) was taken at an altitude of 3,570 km almost 9 hours after peridemeter. Prominent on the limb, at a distance of 4,030 km, is Ahuna Mons, shown in detail in Fig. 23. Image (b) was acquired 80 minutes later from an altitude of 3,770 km. The range to Cerealia Facula was 3,970 km.

Figure 22. Occator Crater. This view was synthesized by D. P. O'Brien with a mosaic of LAMO images and the shape model derived from topographical imaging[24]. The central dome of Cerealia Facula is ~ 3.5 km in diameter and 0.3-0.7 km high. It is in a pit ~ 9 km in diameter and ~ 0.8 km deep.

Figure 23. Ahuna Mons. (a) Mosaic of LAMO images. The cryovolcano is on the lower right. (b) Perspective from within the nearby, 17-km-diameter, geologically unrelated crater, constructed with LAMO topography data[35]. Ahuna Mons is about 17 km across at its base. Extensive fracturing is visible on the top, and bright material and streaks from rockfalls are evident on the slopes of up to 35°. Note the sharp contact between the base and the surrounding smooth area. From the lowest point in the crater to the summit of Ahuna Mons is 7.6 km vertically across a horizontal distance of 15.0 km.

Figure 24. Cerealia Facula. This mosaic was constructed with images from multiple orbits in XMO7. The scene is about 11.5 km across. The lowest images were taken from an altitude of 34 km. Integration times were optimized for the bright material. Gaps in XMO7 coverage are filled in with LAMO data. (Later images in XMO7 filled most of the gaps from an altitude well below LAMO.) The isolated bright structure on the left is shown in Fig. 25.

Figure 25. Detail of Cerealia Facula. The main bright structure here is visible on the left of Fig. 24. It is about 1.5 km in its longest dimension. It is located on a slope that goes down to the upper right of the picture. As in the other images of the faculae, note the sharp boundaries between the bright material and the dark material. The two images of this mosaic were taken at an altitude of 34 km on 22 June in the first peridemeter after the TCM and helped in updating the pointing for the next two peridemeters.

Figure 26. Vinalia Faculae. This mosaic was constructed with images from multiple orbits in XMO7. The scene is about 13.6 km across. The lowest images were taken from an altitude of 34 km. Integration times were optimized for the bright material. Gaps in XMO7 coverage are filled in 32

Copyright 2019 California Institute of Technology. U.S. Government sponsorship acknowledged.

with LAMO data. (Later images in XMO7 filled most of the gaps from an altitude well below LAMO.) Details of the nearly square structure near the center are in Fig. 27.

Figure 27. Vinalia Faculae detail. This image was acquired from an altitude of 58 km and shows an area near the center of Fig. 26 (but rotated here about 40° counterclockwise). The scene is 4.3 km wide. The integration time was optimized for the bright material. Note the intriguing bright curve on the right side of the picture, slightly more than one-third of the way up from the bottom. Its shape suggests some kind of flow.

Figure 28. North wall of Occator Crater. This image was acquired from an altitude of 33 km and is about 3 km across. Note the many boulders that slid part way down the wall before stopping. Ceres' surface gravity is about 0.027g.

Figure 29. Boulder field in Occator Crater. This image inside Occator's eastern rim was acquired from an altitude of 48 km and is about 4.6 km across.

Figure 30. Ridge in Urvara Crater. This image of the central ridge was acquired from an altitude of 58 km and is about 5.5 km across. Note the patterns of bright material that apparently flowed downhill.

Figure 31. North wall of Urvara. This image was acquired from an altitude of 45 km and is about 4.3 km across. Boulders, and the trails they left as they fell down the wall, are visible.

Author biography: Marc D. Rayman earned an A.B. in physics from Princeton University and an M.S. and Ph. D. in physics from the University of Colorado at Boulder. With a lifelong passion for the exploration of space, he joined JPL in 1986 and has worked on a variety of planetary and astrophysics missions and technology development programs there. He served as the Dawn chief engineer, mission director, and project manager

33

Highlights for Dawn at Ceres: The First Exploration of the First Dwarf Planet Discovered • Dawn explored dwarf planet Ceres from 2015 to 2018 • Despite the loss of reaction wheel control, Dawn exceeded its objectives at Ceres • Using ion propulsion, Dawn operated in 10 distinct orbits to optimize the science • After two extensions at Ceres, Dawn's mission ended when the hydrazine was depleted • Dawn revealed Ceres to be a world rich in water, salt, and rock, with diverse geology