Science upside for Starship – Casey Handmer's blog


This blog is a direct follow up of Starship Is Still Not Understood, and is part of the series on popular misconceptions in space journalism.

I think it is relatively straightforward to think of cool things to do with SpaceX Starships, so recent posts have focused on trying to understand the more mixed consequences for incumbent industrial organizations that are not ideally positioned to exploit the coming advances. It is, however, a fun exercise to enumerate all the ways in which Starship and related technologies can help execute bold, ambitious missions of scientific discovery.

While I no longer work for Caltech/JPL/NASA, as always this blog represents only my own opinions and should not be construed as official policy or even particularly heavy criticism. This is not a zero sum game, as there is a lot of upside here. Better technology can help everyone.

Let’s ask a bunch of scientists and engineers and get a laundry list of possible missions to try with Starship. Many of these may not fully utilize the ultimate logistic capacity of the system, but that’s okay. We’re going to focus on how Starship can help specific examples, rather than continuing to harangue future mission designers that they should think in terms of X Starships per year, rather than X Starships per mission.

This blog is also particularly timely as the Astrophysics Decadal Survey was released earlier this month, embodying a series of brutally tough zero sum choices driven by cost disease and a rather meager budget. The decadal process is not perfect but it’s a lot better than the alternative. It represents an ongoing, deliberative process in which the relevant academic community (there is also a planetary science and earth science decadal) develops and presents a consensus around which to collect funding and advocacy strategies. There are missions, such as the Mars helicopter or Europa lander, which are not in the decadal, but they are very very rare.

While I am not qualified to disagree with the specifics of any of their recommendations, all of which represent a treasure trove of potential new knowledge for humanity, the growing time frame involved cannot be ignored. Unable to cram major missions into even ten years, the most recent decadal instead spread the scope into the next next decade. When budgeting and process are considered, the new missions won’t start development for almost two more years, while the Luvoir/Habex hybrid telescope, a 6 m class space telescope mission to study a couple of dozen nearby Earth-like planets, will not launch before 2042 at the absolute earliest. It is wild to me that a process begun before my children were born may not bear fruit until they are completing their PhDs, if they choose to validate my career mistakes by repeating them. If the program is delayed at all, which is quite likely, postdoc astrophysicists reading the decadal this week may have already retired by the time it flies. We are quite literally running up against the limitations of human life expectancy.

To summarize the logistical benefits of Starship, we are now within a few months of the first orbital test flight of a prototype fully reusable launch system. The timing and probability of ultimate success is uncertain but it is safe to say that SpaceX has assembled a competent team and adequate resources, and is acting like they intend to succeed.

While traditional rockets are typically expendable and can launch up to 5 T probes to deep space for a few hundred million dollars, Starship promises the ability to deliver ~100 T of cargo to any planetary surface in the solar system for as little as $50m including refilling tanker flights. Caveats abound, but the key features of the system are a reusable booster and orbital stage, a tanker refueling system to “reset” the upper stage in LEO or even higher orbits, and a heat shield/landing system able to burn off kinetic energy on worlds big enough to retain an atmosphere, or land propulsively on the smaller moons. Most importantly, Starship is designed to support rapid turnaround, so in principle science launches have access to a cheap, abundant launch system. With a design capacity of one million tonnes annually to LEO, there is ample capacity to support the dreams of a generation of scientists who would like to oversee a step change in our capacity to answer big questions. There is absolutely no benefit to developing missions for 2042 hamstrung by the launch constraints of 2002.

One important caveat about cost. There is a difference between cost and price and it is highly likely that SpaceX will retain its hard-fought launch margins unless a competitor forces prices down, or a particular mission has strong alignment with SpaceX strategic objectives, such as building a Moon or Mars base. On the other hand, if a zero discount Starship launch is a significant line item in a new space telescope, this blog’s advocacy will have succeeded beyond its wildest dreams. How do we go about saturating it launch availability? How can we innovate around instrument development to bring their costs in line with coming reductions in launch cost?

For the following list of mission concepts I will provide a summary of the current state of the art and then describe possible future improvements enabled by Starship.

Space Telescopes

As of this writing, the James Webb Space Telescope is achingly close to launch. Begun in 1996 for a planned launch date of 2007 and cost of $500m, actual construction was completed in 2016 before five years of testing for a total cost approaching $9b. Part of the reason for the absurd complexity and expense, beyond routine contractor profiteering and questionable program management, is that the 6.5 m diameter segmented gold-coated beryllium mirror must be folded up to fit into the relatively capacious payload fairing of the Ariane 5 rocket. It is telling that the Ariane 5’s entire launch career began in 2003 and will end with JWST, nearly four years after the penultimate launch. As the JWST program rolled on eating everything in sight, subsequent mission new starts were delayed and delayed again, and subsequent mission plans have assimilated this trauma not by aggressively finding ways to recover our historical ability to deliver cool new stuff quickly and cheaply, but by fiddling with spreadsheet parameters to once again lower expectations for future project delivery competence.

Starship can’t magically generate engineers and processes that can deliver a cheaper space telescope, but it can provide a launch system that a) greatly reduces mass and volume constraints and b) reduces the potential cost for operating a serial space telescope construction and launch program, whereby design improvements and learnings can be rolled in continuously.

The first class of things Starship can do really well is launch lots of stuff. This can enable the development of a standard telescope bus, similar to those used by surveillance satellites, to which custom instruments can be added. Other possibilities that require no custom launch vehicle engineering include orbital neutrino detectors, particle accelerators, or gravitational wave observatories.

Another possibility is to support monolithic telescope design that doesn’t require a 400 step sequence to unfold. For a relatively trivial fraction of the overall telescope budget, non-recurring engineering costs could weld together an expendable Starship variant (no TPS, no flaps, no landing legs) with a 15 meter diameter payload fairing. Almost overnight, endless gnashing of teeth about the relative mirror diameters of Luvoir or Habex, or the relative difficulty of performing coronography with a segmented, non circular mirror, go away.

Starship MegaChomper would also be useful for one-off deliveries of other large space hardware to remote locations, including space station parts, light sails, or anything else that accrues substantial cost/schedule overhead to endure folding or modularization. Imagine the size of the starshade that could be fit into that thing!


Starship with 15 m fairing.

Probably the coolest telescope concept enabled by Starship, though, is the giant segmented telescope to end all giant segmented telescopes. An unmodified Starship can deliver perhaps a dozen 8 m monolithic hexagonal free-flying segments per launch to a target location such as L2, where they self assemble, calibrate, and then focus incoming light. Over a few dozen Starship flights, a truly enormous spherical mirror section >600 m in diameter and with a focal length of 20 km or so can be assembled behind a free-flying sun shade, pointed in a direction of general interest. Dozens of specialty instruments can then be launched to operate at target-specific foci, operating in an off-axis modality by default. Depending on choices about geometry, a single mirror could address O(10 degrees) of the sky at any one time. In the most extreme case a series of mirrors, possibly in a dodecahedral configuration, could enable simultaneous examination of the entire sky limited only by the number of secondary instruments.


Multiple independent free flying secondary optics and instruments (gray boxes) can observe numerous exoplanets or other astrophysical targets simultaneously with off-axis targeting.

Light sails

There are two less impractical approaches for terraforming Mars, both focused on increasing net heat retention in the atmosphere. The first is generation of powerful per-fluoro carbon greenhouse gases in giant factories on the surface. The second is mass producing light sails on Earth, launching them into LEO, then flying them to Mars where they can lurk near Mars-Sun L2 and reflect light back at the planet, reducing heat loss during the Martian night. In principle these can be any size but last time I did a trade study it supported mass production of sails ~30 m in diameter each weighing 1 kg with a cell phone based guidance computer and LCD panels for steering and trim. Each Starship could launch 100,000 of these, with a combined area of almost 100 km^2. Flying as an enormous autonomous flotilla they would reach Mars in less than a year and adopt a station magnifying the sun on the far side of the planet. Mere dozens of such Starship launches would be needed to substantially increase net insolation on Mars and begin raising the temperature, without the emplacement of any surface infrastructure.

Interstellar objects

As of November 2021 there are two known interstellar objects discovered transiting our solar system. It is within our capabilities to build a generic exploration probe, the challenge is launching it quickly and fast enough to catch up with the next candidate so we can get a decent close up look. To be perfectly frank, there are concepts in study right now that don’t even need a Starship, just a steady cadence of probes launched to highly energetic Earth orbits where they can wait for activation, and upon retirement after a few years, be directed to some candidate near Earth asteroid instead. Starship simply enables mass production and launch of these probes, along with improved propellant margins and reduced mass constraints. Why not launch 50 every six months, chase down ‘Oumuamua and 2I/Borisov, and get eyes on every major asteroid inside the orbit of Mars? It would not be cheap but I know dozens of astronomers who would donate half their meager salaries in perpetuity so they didn’t have to endure That Guy dragging Jill Tarter and insisting that it was an alien artifact, ever again.

Bombard All The Planets

While we’re talking about mass production of generic probes to chase down fast-moving interstellar visitors, it’s a great time to revisit the old concept of “Bombard All The Planets”. Since the end of the Mariner program, robotic planetary exploration has generally consisted of expensive, laboriously constructed once-in-a-lifetime one-offs to Jupiter, Saturn, or Pluto. All planets have launch windows but most of them have a launch window at least once per Earth year.

Below is a plot I made in a fit of enthusiasm a few years ago showing all the launch windows to all the planets between 2000 and 2037, focused on the Falcon Heavy. As you can see, barely any launch windows (the colored blobs) have missions in them – what a waste!


Pork chop plot showing all launch windows until 2037.

A fully fueled Starship in LEO requires about 10 launches at a cost of perhaps $50m-$100m. Unlike Falcon Heavy, whose capacity for direct launch to Pluto is pretty meager, a Starship could deliver a flyby mission weighing 100 T to any of the outer planets or moons in less than 10 years. With refueling at higher energy Earth orbits and some creative use of flybys and/or aerobraking, a Starship could deliver >10 T to the surface of any of the outer planet moons with less than a decade of flight time. Starship could deliver 100 T payloads to the surface of Venus or Mars, and even Mercury could get substantial landers and rovers. In short, Starship offers an affordable conveyor belt for essentially anything mission designers can dream up and build. For substantially less than current annual SLS development cost, a planetary science-focused Starship launch program could send a fully loaded Starship to every planet at least once per year, except for Mars whose launch windows are less frequent, but which benefits from Starship baseline design and will probably enjoy its own dedicated program.

Why shouldn’t we have a dedicated orbiter, lander, rover, helicopter, and submarine on every discrete body in the solar system over, say, 100 km in diameter? Let’s build a fleet of clockwork automatons for Venus and an armada of submarines for Europa, Enceladus, and Titan. Let’s darken the Martian skies with helicopters. Let’s drive rovers across the frozen nitrogen plains of Pluto.


This is all the solid surface in the solar system. We must ROVE it.

Of course this couldn’t be done if every probe cost $1b to build. But I hold in my hand a cell phone that can wirelessly download the entire content of a large library in less than a second almost anywhere on Earth, that exceeds the computational power of the best super computer in the year 2000, that cost me less than $1000 to buy and which was not even the most highly rated smart phone in its year. It is within our capacity as a species to exploit the relaxed design constraints enabled by Starship and build a few thousand tonnes of generic space probe each year for a more reasonable price. Failure is acceptable, because new probes, instruments, and launches are continually rolling off the assembly line at a predictable and rapid pace. PIs need no longer fear that any failure will spell doom until their children are retired.

We have not visited Venus, Uranus, or Neptune since before 1990. Our solar system is a precious gift containing 8 whole planets and hundreds of moons. We have had the capacity to explore it for decades and yet it remains largely ignored. Maybe we should have evolved in a solar system with only one planet and no moons?


Robotic exploration and giant telescopes are great, but the future of space also has humans in it. Let’s talk about how Starship changes the game for human exploration. I cannot be accused of having never touched this subject before. In particular, exploiting Starship now seems to be the only way to save the Artemis program, and the NASA OIG seems to agree. But what about after Artemis? Where can humans live in space?

Starship itself can serve as a human habitat in LEO, GEO, L5, Lunar orbit, the Lunar surface, deep space, Mars, or an asteroid. Additionally, Starship could be used to launch custom-designed modules to build stations at any of these places. Starship alone has about 1000 m^3 of internal volume, which is nearly double that of the ISS. Repurposing empty fuel and oxygen tanks more than doubles that volume. Starship’s welded stainless steel construction reduces the cost and complexity of modifications, particularly ones that do not affect structural performance. Starship could launch a space station for so little money that it’s possible they could be cheap enough to be supported by non government industrial, commercial, and tourist users.

At the extreme, a Starship upper stage could be modified to form a wedge-shaped segment with a removable nose cone, then docked together to form a giant rotating wheel with artificial gravity.


32 segment ring station composed of dockable modified Starship barrels.


Speaking of asteroids, why not use Starship to improve our study of asteroids? OSIRIS-REx and Hayabusa2 are cool, but what if we could travel with 100 T of instruments, or a bunch of people, to a nearby asteroid for a while, then return to Earth. Forget ARM. Send Starship Chomper out to a nearby asteroid, take a big bite, then fly it right back to the cape.

Asteroid mining probably won’t pay, at least for Earth markets, but Starship can make conducting study and assay affordable by speculative explorers. No need for further hypotheticals about platinum asteroids. Send a Starship to the 10 most likely candidates, fly 100 T of each back to Earth, see what’s there.

While we’re sending Starships to every nearby asteroid in sight, we can also begin preventative study of all potential Earth impacting NEOs while we still have time. Precise tracking, surface study, even emplacement of contingency systems for redirection. All affordable, if we can work out how to make more than one of any given spacecraft.

Finally, I’ve always wanted to know if there are actually any vulcanoids, or small asteroids occupying a gravitationally stable region between Mercury and the sun. Let’s send a Starship down inside Mercury’s orbit with a big camera and find out.

Large scale planetary bases

Building Starship-based space stations is one thing, but Starship can also help construction of real bases on the Moon or Mars. No more decades of hand wringing over closing design trades on a $10b Moon “base” the size of a school bus. Starship is designed to enable “drag and drop” logistics. They can deliver so much stuff just unloading them at the destination could prove a major bottleneck.

There are a few dozen scientific research stations in Antarctica, mostly populated by scientists and mountaineers who actually, believe it or not, voluntarily want to be there. Surrounded by a thousand miles of frozen ice, jagged mountains, and millions upon millions of psychotically famished penguins with very sharp beaks. The largest of these stations is McMurdo Station, which houses up to 1500 people during the summer and is typically supplied by ship. Imagine a base of 1500 people on the rim of Shackleton Crater on the Moon. With a per-person mass overhead of 10 T, such a base would require only 150 Starship landings over perhaps five years to construct. Indeed, much of the base could simply be Starships with Whipple shields instead of TPS, pre-fabricated with the essential elements to support human operations. Let’s not overthink this. The Starship is a self-landing pressurized structure with >2000 cubic meters of internal volume.

The Mars base is even more ambitious but reflects SpaceX’s ultimate goal. A city of a million people living beneath a spacious transparent tensile inflatable greenhouse building Humanity 2.0. This is 99% finding and fixing widget manufacturing bottlenecks, so somewhat less exotic than the movies might have you believe. Still, a grand vision. And one that is categorically impossible without a Starship-class fully reusable high capacity high delta-V launcher.

Other In-Space Infrastructure

Starship can also be used to launch systems previously impossible due to launch cadence, mass, and volume constraints. I’m somewhat dubious about some of these applications but orbital tugs, fuel depots, space based solar power, and nuclear thermal rockets are all orders of magnitude less difficult in a world with Starship than one without.


Starlink isn’t strictly Starship logistical capacity, though it is enabled by it. Starlink is SpaceX’s orbital high speed internet megaconstellation. Every day I wake up and struggle to believe that this thing is actually real, and I’ve seen it with my own eyes. We live in the future.

This section is concerned with potential applications for the Starlink constellation that have not been possible in the past with single satellite missions.

Starlink will ultimately be a network of tens of thousands of satellites connecting to hundreds of millions of user terminals located all over the Earth. Its radio encoding scheme adapts the signal rate to measured atmospheric opacity along the signal line of sight across 10 different frequency bands in real time. Collectively, the system measures trillions of baselines of Earth’s entire atmosphere every day. This data, fed into standard tomography algorithms such as those used by medical CT imagers, can resolve essentially all weather structure in the atmosphere. No more careful scrutiny of remote weather station pressure gauge measurements. No more reliance on single mission oxygen emission line broadening. Instead, complete real time resolution of the present state of the entire atmosphere, a gift for weather prediction and climate study.

Starlink satellites are equipped with perhaps the most versatile software defined radios ever put into mass production. Each antenna allows the formation of multiple beams at multiple frequencies in both send and receive. With sufficiently accurate position, navigation and timing (PNT) data from GPS satellites, Starlink satellites could perform fully 3D synthetic aperture radar (SAR) of the Earth’s surface, with enough bandwidth to downlink this treasure trove of data. Precise ocean height measurements. Precise land height measurements. Surface reflectivity. Crop health and hydration. Seismology and accumulation of strain across faults. City surveying. Traffic measurements in real time. Aircraft tracking for air traffic control. Wildlife study. Ocean surface wind measurements. Search and rescue. Capella has produced extraordinary radar images with a single satellite. Now imagine the resolving power with birds from horizon to horizon.

Starlink SAR is great for Earth observation, but the same principle can be applied looking outwards. Starlink is a network of thousands of software defined radios with highly precise PNT information and high speed data connections. It is practically begging to be integrated into a global radio telescope. With 13000 km of baseline and the ability to point in any desired direction simultaneously, Starlink could capture practically holographic levels of detail about the local radio environment. Literally orders of magnitude better resolution than ground-based antennas like the Very Large Array. Cheaper than repairing Arecibo and independent of Earth’s rotation. Potentially capable of resolving exoplanets.

There’s no reason to do only passive radio astronomy. Starlink can exploit its exceptional resolving power and onboard amplifiers to perform active planetary radar, for examination of close-flying asteroids and transmission of radio signals to distant missions in support of the Deep Space Network. As of November 2021, all Starlink satellites are flying with lasercoms so in principle the DSN application could also support laser, as well as radio, communication with distant probes. No need to build even larger dishes than the 70 m monsters.

And while Starlink can derive PNT from the GPS constellation, it need not depend on it forever. High capacity radio encoding schemes such as QAM4092 and the 5G standard contain zero-epoch synchronization data, meaning that any radio capable of receiving Starlink handshake signals is able to obtain approximate pseudorange information. What Starlink’s onboard clocks lack in nanosecond stability, they make up in sheer quantity of connections and publicly available information about their orbital ephemerides. Already a group from OSU has demonstrated <10 m accuracy, while a group based at UT Austin is developing a related method for robust PNT estimation using Starlink hardware. It seems likely to me that Starlink could support global navigation with few to no software changes and no hardware changes, improving the resilience of satellite navigation especially in a case where the relatively small GPS constellation is disabled. I won’t go into vast detail, but GNSS signals are not only used for pizza delivery, but also support a vast array of Earth science objectives, including the monitoring of tectonic drift.

Starlink has received its fair share of criticism, drawn perhaps by its overwhelming scale and potential impacts to ground-based astronomy. But Starlink can also be the single greatest scientific instrument ever built, a hyperspectral radio eye the size of the Earth, capable of decoding information about the Earth and the universe that is right up against the limits of physics.

Will We Do All This And More?

I don’t know. Maybe eventually. Starship removes the mass and volume constraints traditionally blamed for the expense of space exploration. Does that mean that come 2022, the decadals will all be revised to reflect this new reality? I doubt it, at least not right away. The expense of space exploration is multifactorial but Starship at least hands us the key to find another way, to write inventive contracts that incentivize continuous improvement and innovation and a reduction of instrument costs commensurate to the improvement in access that Starship brings. I, for one, would be thrilled to see a global recalibration of the scope of our ambition in recognition of the fact that, as a species, we are still capable of doing awesome things quickly, cheaply, and well.