IZ← Ian Zhang / Projects Live solver ↗

Independent design study 2024–26

Can the sun help fly a hydrogen freighter?

Only a little. I wrote a geometric-programming model of a 5‑tonne regional cargo plane that runs on hydrogen fuel cells, then asked how much of its cruise power solar cells could supply. At sensible wing shapes the answer is 2 to 3 percent. Past that, the cells need more area than the aeroplane has.

Take-off mass vs aspect ratio500 km · 40 m/s · 5 t
Nine disciplines, one program
AEROAerodynamicsSTRWing-box structureEMPEmpennageFCFuel cellsLH₂Hydrogen tanksPVSolar cellsMISMission energyT/OTake-offRESDiversion reserve

01 · The question

Why a geometric program?

Sizing an aircraft is a loop inside a loop. A heavier battery wants a bigger wing, the bigger wing wants a stronger spar, and the spar adds weight again. Iterate by hand and you stop when you get tired, not when you reach the best design.

A geometric program writes every relationship as a posynomial. That turns the whole loop into a convex problem, so the solver returns the global optimum rather than a local one that happens to look fine. Where the physics refuses that form, I fitted posynomial surrogates to data instead, aerofoil polars from XFLR5 among them.

The objective is simple: minimise take-off mass for a 5-tonne payload flown 500 km at 40 m/s, then see how much solar the optimum will accept.

Free variables
≈80
Coupled disciplines
9
Design point
500km · 40m/s
Payload
5t

02 · Explore

Pick a wing. Watch the aircraft change.

Aspect ratio (span² ÷ wing area) turned out to be the lever that matters most. Every number below is a full solve from the paper's sweep at the design point.

Where the mass goes

    03 · The solar limit

    Solar runs out of aeroplane.

    The optimiser may add cells until they need more area than the wing and fuselage tops provide. That cap arrives early. At 40 m/s and A = 11, solar tops out near 2.5 percent of cruise power. A long, thin wing (A ≈ 25) reaches about 6 percent, but by then the aircraft weighs 67 tonnes.

    Highest solar share of cruise powerby cruise speed
    Solar only: area needed ÷ availableby cruise speed

    Flying on sunlight alone needs about ten times the area the airframe offers at best, and the gap widens with speed. The A = 25 and 33 designs never close at all.

    04 · Sensitivities

    What the take-off mass really depends on.

    The solver reports how much the optimum moves when each input moves. At A = 11, +1.15 for payload means 1% more payload costs 1.15% more take-off mass. Payload and structure dominate; range barely registers. This is a structure-limited aircraft, not an energy-limited one.

    Effect of a 1% change in each input on take-off massA = 11

    05 · Better cells?

    Better solar cells don't rescue it.

    Today's high-end cells convert about 25 percent of sunlight. Raise that and the solar share climbs in a straight line, but from a low base: at A = 11 it reaches 4.3 percent at 40 percent efficiency and 10.7 percent at an impossible 100.

    Highest solar share vs cell efficiencycell mass per m² held fixed

    06 · Limits

    What the model covers

    The model covers aerodynamics, a wing box sized for bending, tail booms, fuel-cell and hydrogen-tank mass, solar generation, the mission energy balance, take-off and an emergency diversion. Its answer is a hydrogen fuel-cell freighter in which sunlight trims a few percent off cruise energy, and nothing more.

    It is built to be reused. Change the payload class, the mission or the technology numbers and the same program re-solves in seconds, which is what the live solver lets you do.

    Where it stops being trustworthy

    • Drag polars and component masses are fitted curves, so they hold only inside the data behind them.
    • Airframe mass comes from a payload-driven regression of existing aircraft, not a part-by-part build-up.
    • Climb has no minimum gradient, so the optimiser picks a very shallow one. The solar results depend only on cruise and are unaffected.
    • Past A ≈ 30 the program gets badly scaled. Read those curves as bounding sketches, not exact answers.

    Your turn

    Change any assumption yourself.

    The live solver runs the same model on a server. Move payload, range, speed, cell efficiency or the sun's position and it re-optimises the aircraft in a few seconds.