A Million Fireflies in the Martian Sky? On Warming Mars Using Solar Sails

Eight weeks ago, the second Green Mars Workshop brought together scientists and engineers to discuss what would be needed to terraform Mars. (An overview prepared by Pioneer Labs can be found here.)

For over 50 years, there has been consensus that the first step toward globally terraforming Mars would be to warm the planet by 10s of degrees C [1]. This would melt some of Mars’ ice, a necessary but insufficient condition to open the door to resilient pioneer species that could grow and spread, eventually transforming the regolith into soil and the CO₂ atmosphere into an oxygen-rich one. This first step—warming—is the focus of the Mars Warming Research team at Astera. And while most of our work involves nanoscale aerosols, we are also investigating alternative Mars-warming strategies.

One such approach was outlined by Casey Handmer, founder and CEO of Terraform Industries, at the most recent Workshop. The idea involves using orbiting mirrors to reflect sunlight down onto the planet, a Mars-warming concept that can be traced back at least as far as Robert Zubrin’s and Christopher McKay’s 1993 work, “Technological Requirements for Terraforming Mars.” There, the authors describe how an increase of only 4° in temperature at the south pole could cause all of the CO₂ ice in the cap to sublimate, which (based on the latest radar data [3]) would double the thickness of the atmosphere¹. Zubrin and McKay estimated that this could be done using a 250 km-diameter mirror at ~200,000 km altitude on the anti-Sun side of Mars, where the solar radiation pressure would balance the force of gravity.

Handmer’s innovation presented at the Workshop was to implement the “mirror” not as a single megastructure, but rather as a constellation of many solar sails, each of which flies to Mars under its own thrust.

What is a solar sail?

A solar sail is a propellantless propulsion system: Instead of carrying propellant, solar sails absorb momentum from solar photons. The ideal solar sail is highly reflective (to optimize momentum transfer) and light (to maximize acceleration). These qualities are parameterized by the specular reflectivity ρ_s, with values between 0 and 1 (1 being the most mirror-like), and the areal density σ, typically reported in g/m².

Solar sails include JAXA’s IKAROS, launched in 2010, which demonstrated solar sail propulsion en route to Venus [4]. IKAROS had σ ≈ 1600 g/m². The Planetary Society’s LightSail 2 (2019) improved upon IKAROS’ areal density, weighing in at σ ≈ 150 g/m² [5]. A reference architecture for future Earth-orbiting reflectors (to increase power output from solar farms on the ground) envisages σ = 18 g/m² [6]. Reflect Orbital, a Hawthorne-based startup, is working to develop more efficient solar sails for that application.

Because a sail derives its propulsive force from reflected sunlight, its attitude constraints are more stringent than propellant-based systems. The orientation of a flat sail is described by the vector n, also called the sail normal. The primary attitude constraint is on the cone angle α between n and the vector s, which points from the sail to the sun. Beyond α ≈ 40°, derived thrust plummets as cos²(α).

Figure 1. Adapted from [7], showing a solar sail, its orientation vector n, its sun line s, and the cone angle α between n and s.

Once a sail is in orbit around Earth and above 800 km altitude (to minimize atmospheric drag), it can begin to accumulate momentum, spiraling outward, until it enters heliocentric orbit. There it becomes a satellite of the Sun.

Figure 2. From [8]: a 106-day solar sail Earth-escape trajectory.

Why warm Mars with sails, rather than simple mirrors?

The “orbiting mirror” strategy implemented with a constellation of sails, rather than a single large mirror (or a constellation of smaller mirrors), has many advantages. First, to put a 250-km megastructure into orbit around Mars would require in-space assembly. So far, humans have only assembled ≤ 0.1-km scale structures in space.

To put many smaller mirrors in Mars orbit requires making the mirrors either (a) in orbit near Mars (e.g., from materials sourced from Phobos and/or Deimos); or (b) on the surface of Mars and then launching them into orbit; or (c) on Earth, and then transporting them to Mars via rocket. Both (a) and (b) would require big advances beyond the current state of the art (factories in space). For (c), the cost of transporting so many mirrors all the way from Earth to Mars would be very large.

But solar sails built on Earth and launched into Earth orbit could fly themselves to Mars. This would allow Mars-warming solar sails to leverage the significant cost savings made possible by reusable orbital launch vehicles.

Partially reusable rockets like Falcon 9, Falcon Heavy, and New Glenn have already lowered the price of launch to Low Earth Orbit, from ~$50,000/kg in the Shuttle era to ~$2,000/kg today. With fully reusable rockets like SpaceX Starship and Stoke Space Nova coming online, launch costs to Low Earth Orbit should continue to decrease dramatically: Payload Space projects < $100/kg by 2030.

Prerequisites for cost estimates

To estimate the cost of warming Mars with a constellation of solar sails, we need to know or guess (1) the total mirror surface area in orbit (less if only warming the south polar cap, more if targeting global radiative forcing); (2) the minimum viable areal density of each sail and the cost per kilogram of sail material; (3) the cost and total mass of the auxiliary equipment needed for each sail (structure, power, communications, attitude determination and control system, central processor); (4) and the rate, complexity, and cost of mass-manufacturing such highly performant spacecraft.

Though recent studies have developed more efficient orbits for warming Mars with mirrors [e.g. 9–12], we can use Zubrin’s giant mirror as a proxy for (1), i.e. total surface area, and treat it as an approximation. His proposed 250-km diameter disc has a surface area of 5×10¹⁰ m². If we assume square solar sails 100m on a side, each would contribute a surface area of 10⁴ m², and thus the full constellation would require 5×10⁶ such sails. (This is a rough approximation and would vary with more specific assumptions about orbital altitude, fraction of time in orbit reflecting light onto the region of interest, etc.)

Handmer estimates a total cost of $10bn and ten years, assuming very light 1g/m² sails. But before discussing the details, we must show that even a single solar sail can fly from Earth to Mars under its own propulsion.

Capture at Mars: it’s non-trivial, and it’s necessary

The trip that a solar sail must take to get from Low Earth Orbit to a Mars-warming orbit includes: (1) launch from the Earth’s surface into LEO onboard a (reusable chemical) rocket (~hours); (2) deploy and raise orbit, spiraling up, eventually escaping Earth’s gravity (~months); (3) continue to raise orbit, now as a satellite of the sun, targeting eventual position and velocity alignment with Mars (~2 years depending on launch time and areal density); (4) undergo capture at Mars (~months); and (5) shape and lower the orbit at Mars to tighten capture, or transfer to a more warming-useful nearby orbit (~months).

The first of these segments is straightforward, as launch vehicles now deliver payloads into orbit close to every other day. The second and third segments—namely, raising orbit around Earth to reach escape, and slowly pushing outward during heliocentric orbit to reach Mars—are also fairly well understood, thanks to a combination of demonstration missions like IKAROS and robust simulations (e.g., Fig. 5). But the fourth segment, capture at Mars, is less well understood.

That a solar sail flown to Mars under its own propulsive force can become gravitationally captured by the red planet has never been demonstrated in flight, and it has received limited treatment in astrodynamics literature. Previous missions to Mars orbit have almost all used a maneuver called a Hohmann transfer: (1) a first-stage chemical rocket is used to escape from Earth’s surface into orbit; (2) a second-stage chemical rocket is used to impart enough momentum to the spacecraft such that it escapes Earth’s gravitational influence and enters into orbit around the sun on an elliptic trajectory (heliocentric reference frame) intersecting Mars’ orbit; and (3) after a ~6–8 month coast period, a third burn is used to circularize the orbit and lower the spacecraft’s velocity relative to Mars, allowing for capture. Depending on the mission, capture might be initially weak, in which case a technique known as aerobraking, or exploiting the drag in the Martian atmosphere, is often used to further reduce the spacecraft velocity and tighten the orbit.

But a Hohmann transfer is off the table for our solar sail, assuming that the rocket takes the sail only as far as Low Earth Orbit. This leaves aerobraking/aerocapture, ballistic capture, and sail-assisted capture. Aerocapture is typically associated with high friction and plasma, and since the sail will ideally be as thin as possible, it may deform or melt. Thus, shielding would probably be required, adding mass and complexity to the spacecraft, which is undesirable. (Perhaps the trajectory could be designed such that the aerobraking would be mild and compatible with the raw sail material, and thus no shielding would be needed, but until that has been worked out, we will focus on ballistic and sail-assisted capture.)

Ballistic capture

Ballistic capture was invented by Edward Belbruno and first demonstrated in 1991, when the Japanese lunar probe Hiten used the technique to salvage its mission and enter lunar orbit with minimal additional fuel expenditure despite the failure of its supporting Hagoromo orbiter. The technique exploits what are known as “weak stability boundary” (WSB) regions around celestial bodies—fractal, ever-shifting regions in which gravitational forces from multiple bodies nearly balance—and enables capture without auxiliary propulsion or drag [13].

Figure 3. From [13]. Top: the structure of a ballistic capture orbit for Mars is shown. “SOI” stands for Mars’ gravitational “sphere of influence”. Bottom left: a ballistic capture at Mars is shown in the Mars-centered Mars-Sun rotating reference frame. Bottom right: the same trajectory is shown in the Mars-centered inertial reference frame.

Ballistic capture is a very sensitive technique: Slight changes in initial conditions can yield dramatically different outcomes. This can be largely attributed to the chaotic nature of the weak stability boundary. To identify viable ballistic capture trajectories in the restricted “Sun, Mars, Spacecraft” two-dimensional three-body problem, we used a grid search approach. The grid search sweeps over many candidate initial orbits, propagating the spacecraft forward and backward in time from that orbit subject to the gravitational forces of Mars and the Sun. Propagated forward through time, the spacecraft should remain close to Mars, hence “capture.” Such an orbit is called “forward stable.” But propagated backward through time from that same forward-stable initial orbit, the spacecraft should escape from Mars. Only if it escapes when propagated backwards is the candidate orbit “reachable” from an initial position far from Mars. Such an orbit is called “backwards unstable.”

Even in just two dimensions, Cartesian state vectors have four elements (position x, position y, velocity along x, velocity along y), rendering grid searches computationally expensive. But orbits do not require full state vectors for complete description. Instead, a two-dimensional orbit around Mars can be parameterized using r_p and θ, where r_p is the altitude at periapsis, and θ the angle between the line of periapsis and the Sun-Mars line [e.g., 14]. This reduces the dimensionality of the grid search to two, at which point all that remains is to sweep the candidate orbits looking for those which are both forward stable and backwards unstable.

Figure 4. Top: A polar plot showing candidate orbits parameterized by θ and r_p (AU). Blue circles represent forward-stable orbits. Smaller red circles represent backwards-unstable orbits. Regions of overlap (red plotted on top of blue) indicate viable ballistic capture trajectories. Bottom left: One of the forward-stable, backwards-unstable orbits identified by the grid search, propagated forward in time (blue curve) and backward in time to escape (orange curve) in the Sun-Mars rotating frame centered at Mars. In the legend, “SOI” refers to Mars’ gravitational sphere of influence. Bottom right: the same orbit propagated forward and backward in time, but in the Mars inertial reference frame. Code and plots credit: Yuji Takubo in collaboration with the Mars Warming Research team at Astera.

Heliocentric transfer

Once a viable orbit has been found, propagating it backwards in time until the spacecraft “escapes” yields a state that can be targeted as an insertion point for capture (e.g., Fig. 4, lower panel, green dots). A combination of simulations and optimizers can then be used to calculate the interplanetary trajectory the sail must fly after escape from Earth to reach the desired insertion point at the right time.

To perform such trajectory optimization, we developed solar sail simulation software in Python. Given a start time and sequence of k sail orientation angles, the simulation propagates the sail from a just-escaped-Earth initial state over the specified mission duration, subject to the Sun’s gravitation and radiation pressure. For example, if the mission duration is 100 days and 10 orientations are provided, the sail applies the first orientation for the first 10 days, the next orientation for the next 10 days, and so on.

Mathematically, in two dimensions, the simulation can be thought of as a function

\(f(t_0, n_0, n_1, ..., n_k, d) = [x_f, y_f, v_{x,f}, v_{y,f}]\)

where t₀ is the start time; n₀ ... n_k the sequence of k orientation angles; and d the total flight duration. The outputs (x_f , y_f ) and (v_x,f , v_y,f ) represent the sail’s final position and velocity, respectively.

The role of the optimizer is to find a sequence of orientations n₀ ... n_k and total time d such that the final state of the sail matches the desired insertion state as closely as possible, in both position and velocity. The optimizer guesses a set of values for these parameters, plugs them into the simulation to see how they turn out, and then updates its guess based on the results. If this problem involved only a few parameters, it would be solvable with just another grid search. But primarily because the sequence of sail orientations is long (twice as long in three dimensions), more advanced optimization algorithms become necessary. For this work we are using the Interior Point Optimizer (IPOPT) algorithm, widely used in astrodynamics, via its cyipopt Python implementation.

With appropriate choices for launch date and total flight time, we have already been able to achieve < 1 km position error and < 1 m/s velocity error with respect to candidate ballistic capture insertion points using as few as twenty sail orientation changes over a ~two year heliocentric flight (Fig. 5).

Figure 5. 2D (left) and 3D (right) representations of an optimized 620-day solar sail trajectory targeting a ballistic capture insertion node. The blue arc represents Earth’s orbit around the sun, which is represented with a yellow circle. The black X is the sail’s initial position (coincident with Earth’s initial position); the small blue open circle marks Earth’s position at the end of the flight. The red start represents the desired final position. Twenty equally temporally spaced orientation segments are used, and sail orientation is denoted with arrows at the midpoint of each segment in the 2D figure. Note: The Z axis in the 3D plot uses different scaling than the X and Y axes, and thus the inclination of the orbit appears exaggerated. Errors w.r.t. target < 1 km in position and < 1 m/s in velocity.

Connecting the dots

We now have simulations reproducing both (a) the near-Mars-into-ballistic-capture segment, and (b) the heliocentric near-Earth-to-near-Mars segment. We are currently working on implementing the third core trajectory segment, namely, (c) spiraling up from and escaping Low Earth Orbit. Since this “spiral up” has been demonstrated before [e.g., 8], we are confident it can be done. Once we can simulate these three segments, we can connect the ends together (just as we did in Fig. 5, in which the end state of the heliocentric transfer corresponds to the initial state for ballistic capture), and demonstrate end-to-end simulated solar sail flight from Earth to Mars.

Worth noting is that a solar sail, which is very light and possesses a large reflective surface area, behaves quite differently from a ballistic particle. Our ballistic-capture trajectory segment currently includes solar gravity and Mars gravity, but no solar radiation pressure force. This simplification enables a more rapid search across the dense grid of θ and r_p parameters, but comes at the cost of realistic simulation. We are now working to address this by modeling “sail-assisted capture.” Depending on sail performance, we anticipate that the final capture may appear more like an elegant “spiral down.”

Figure 6. A preliminary “spiral down” sail-assisted capture trajectory at Mars. Trajectory found by propagating sail motion backwards in time, starting from a low orbit near Mars, subject to the following analytical control law: at any given time, orient the sail in the direction that will maximally increase its energy, but without exceeding |α| > 50°. In forward propagation, this should minimize energy, leading to capture. In the left plot, pink arrows denote the sail normal at various timesteps, and the sail approaches from the anti-sun side of Mars. In the legend, “SOI” refers to Mars’ gravitational sphere of influence. Code and plots credit: Yuji Takubo in collaboration with the Mars Warming Research team at Astera.

Another unresolved detail is maneuvering the sail into a Mars-warming orbit after it has become captured at Mars. We plan to turn our attention to this after simulating the end-to-end capture trajectory in a consistent and thorough solar system model.

Zooming out

Determining whether and at what cost Mars can be warmed using solar sails isn’t just an orbital mechanics problem. Trajectory optimization is just one piece of the puzzle. Unlike every single-spacecraft Mars missions (e.g., Pathfinder, Spirit, Opportunity, Mars Reconnaissance Orbiter, Curiosity, Perseverance, etc), even if we had a solar sail right now that could fly itself from Low Earth Orbit into a Mars-warming orbit, we would still need ~10⁶ more copies of that sail.

Recent advances in manufacturing—specifically in cellphone manufacturing, such as high-quality camera miniaturization and system-on-a-chip performance gains, but also in automation—should be leveraged to address this challenge. Optimal application of those advances toward efficient solar sail production should be studied. Perhaps with only modest hardening to accommodate vacuum conditions, cellphone cameras could be used as star trackers, for instance, contributing to attitude determination. Lighter and more efficient structures to support larger, more maneuverable solar sails should also be re-designed and re-engineered for mass-manufacture.

The scale of this endeavour requires contributions from many scientific and creative fields, from orbital mechanics to origami. We hope many humans from across many disciplines will feel inspired to help advance this work.

Figure 7. From Casey Handmer’s presentation at the Green Mars Workshop, 2025. An illustration of what a constellation of Mars-warming solar sails might look like at sunrise, viewed from the surface of the red planet.

Ari Essunfeld

Thanks to Edwin Kite and Yuji Takubo for reading a draft.

References

[1] Center, A.R., Averner, M.M., MacElroy, R.D., 1976. On the Habitability of Mars: An Approach to Planetary Ecosynthesis, NASA SP. Scientific and Technical Information Office, National Aeronautics and Space Administration.

[2] Zubrin, R., Mckay, C., 1993. Technological requirements for terraforming Mars, in: 29th Joint Propulsion Conference and Exhibit. Presented at the 29th Joint Propulsion Conference and Exhibit, American Institute of Aeronautics and Astronautics, Monterey, CA, U.S.A. https://doi.org/10.2514/6.1993-2005

[3] Bierson, C.J., Phillips, R.J., Smith, I.B., Wood, S.E., Putzig, N.E., Nunes, D., Byrne, S., 2016. Stratigraphy and evolution of the buried CO₂ deposit in the Martian south polar cap. Geophysical Research Letters 43, 4172–4179. https://doi.org/10.1002/2016GL068457

[4] Tsuda, Y., Mori, O., Funase, R., Sawada, H., Yamamoto, T., Saiki, T., Endo, T., Kawaguchi, J., 2011. Flight status of IKAROS deep space solar sail demonstrator. Acta Astronautica 69, 833–840. https://doi.org/10.1016/j.actaastro.2011.06.005

[5] Mansell, J.R., Bellardo, J.M., Betts, B., Plante, B., Spencer, D.A., 2023. LightSail 2 Solar Sail Control and Orbit Evolution. Aerospace 10, 579. https://doi.org/10.3390/aerospace10070579

[6] Viale, A., Çelik, O., Oderinwale, T., Sulbhewar, L., McInnes, C.R., 2023. A reference architecture for orbiting solar reflectors to enhance terrestrial solar power plant output. Advances in Space Research 72, 1304–1348. https://doi.org/10.1016/j.asr.2023.05.037

[7] Pezent, J.B., Sood, R., Heaton, A., 2021. Configuration space and stability analysis of solar sail near-vertical earth-trailing orbits. Advances in Space Research 67, 2981–2994. https://doi.org/10.1016/j.asr.2020.10.011

[8] Cabalga, G.J.M., 2020. Numerical optimization of solar sail trajectories to Mars. KTH Royal Institute of Technology School of Engineering Sciences.

[9] McInnes, C.R., 2009. Mars Climate Engineering Using Orbiting Solar Reflectors, in: Badescu, V. (Ed.), Mars. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 645–659. https://doi.org/10.1007/978-3-642-03629-3_25

[10] McInnes, C.R., 2006. Planetary Macro-Engineering Using Orbiting Solar Reflectors, in: Badescu, V., Cathcart, R.B., Schuiling, R.D. (Eds.), Macro-Engineering, Water Science and Technology Library. Springer Netherlands, Dordrecht, pp. 215–250. https://doi.org/10.1007/1-4020-4604-9_11

[11] Çelik, O., McInnes, C.R., 2022. An analytical model for solar energy reflected from space with selected applications. Advances in Space Research 69, 647–663. https://doi.org/10.1016/j.asr.2021.10.033

[12] Salazar, F.J.T., McInnes, C.R., Winter, O.C., 2016. Intervening in Earth’s climate system through space-based solar reflectors. Advances in Space Research 58, 17–29. https://doi.org/10.1016/j.asr.2016.04.007

[13] Topputo, F., Belbruno, E., 2015. Earth–Mars transfers with ballistic capture. Celest Mech Dyn Astr 121, 329–346. https://doi.org/10.1007/s10569-015-9605-8

[14] Luo, Z.-F., Topputo, F., Bernelli-Zazzera, F., Tang, G.-J., 2014. Constructing ballistic capture orbits in the real Solar System model. Celest Mech Dyn Astr 120, 433–450. https://doi.org/10.1007/s10569-014-9580-5

1. The south polar ice record on Mars holds vast quantities of valuable climatological and geological information. Thorough sampling ought to be conducted prior to any large-scale sublimation effort.