Space-grade solar power sits behind four documented chokepoints: the germanium substrate (a structural, already-exercised export-control exposure), gallium (a smaller, logistics-manageable one), ceria-doped coverglass (a heritage monopoly under demolition), and the qualification lock (the filing-cabinet moat that keeps new materials grounded for a decade). Add the economics — $20/W merchant cells, $600–950/W retail panels, a proven ~$5/W floor for space-ified silicon — and the question becomes concrete: what would it take to generate watts in orbit without any of that?

Each direction below is defined by two things: which constraint it escapes, and what must be true for it to win. They are not competitors; they are a portfolio with different clocks.

R1 — Space-ified terrestrial silicon

The workhorse. Near-term, highest confidence.

Ultrathin crystalline silicon with polymer encapsulation, no coverglass, radiation self-annealing — the route Solestial and Source Energy are already flying. Nothing here needs inventing; it needs characterizing.

Open research:

  • Annealing kinetics at real operating temperatures as a function of wafer thickness and doping. The mechanism exists in the literature but is not yet creditable at the 100 µm n-type design point.
  • Handling yield below 100 µm: breakage climbs from ~10% at 100 µm to ~96% at 70 µm. Carriers, edge reinforcement, and kerfless methods are the battleground.
  • Atomic-oxygen protection without glass: ETFE, SiO₂ nanolaminates, sacrificial coatings.

Wins if: end-of-life degradation becomes predictable and boring — an actuarial input rather than a research question.

R2 — Perovskite and perovskite/Si tandems

The upside bet. Mid-term.

Self-healing under irradiation, specific power above 1 kW/kg, projected $5–15/W at maturity. The tandem escapes every materials chokepoint at once: no germanium, no gallium worth counting, no glass.

Open research:

  • Multi-year stability under combined UV, ATOX, and thermal cycling. Single-stressor data exists; the coupled dataset does not, anywhere.
  • Encapsulation as the actual product — the cell is easy, keeping it alive is the company.
  • Qualification protocols. None exist for this class. Writing the test standard is itself a research direction with unusual leverage: whoever defines the protocol defines the market entry cost for everyone else.

Wins if: a credible 5-year LEO degradation model survives independent replication.

R3 — Substrate liberation: ELO/IMM

The geopolitical hedge. Mid-term.

Epitaxial lift-off and inverted metamorphic growth separate the III-V junction from the germanium wafer. The wafer becomes a reusable template instead of a consumable. This is the only direction that attacks the germanium chokepoint while keeping III-V performance.

Open research:

  • ELO throughput and yield at industrial scale — today it is a laboratory courtesy.
  • Substrate reuse cycle count. ~10 cycles is the threshold where germanium exposure drops by an order of magnitude and the chokepoint becomes a line item.
  • IMM mechanical reliability without the handle wafer: what holds the cell together once the germanium leaves.

Wins if: reuse economics close at constellation volumes. Watch MicroLink and the Rocket Lab/AZUR process disclosures.

R4 — III-V on alternate lattices

A watching brief, not a program.

Metamorphic III-V on silicon, GaAs substrates (recyclable but still gallium), sapphire. Honest assessment: this trades a structural chokepoint for a defect-physics problem. Threading dislocations and CTE mismatch over 30,000 thermal cycles are soluble in principle and unrewarding in practice so far.

Wins if: someone else’s metamorphic-buffer breakthrough makes it cheap to revisit. Check annually, fund never.

R5 — In-space cell manufacturing

The long game. Far-term, and uniquely aligned with this program.

Deposit thin-film cells directly onto deployed membranes in orbit. Vacuum is free up there; launch mass collapses to feedstock; coverglass becomes optional when nothing has to survive the fairing. This is the only direction that escapes substrate, coverglass, and launch mass in one move.

Open research:

  • Deposition rate and film quality in uncontrolled vacuum, on non-ideal substrates, with thermal drift.
  • Contamination of the host platform — the coater and the array must not poison each other.
  • Repair and re-coat as a service model: a coating pod visiting a membrane is an ISAM mission profile, and it composes with the desktop’s attachment architecture.

Wins if: deposited specific power beats launched specific power after accounting for the coater’s own mass and operations. Nobody is motivated to build this except programs like this one.

R6 — System architecture as research

Cross-cutting. Cheap to do now, compounding returns.

The ISCR analysis showed the move: integrate solar, radiator, and compute so the dedicated panel structure disappears (~500 W/kg at array level). Sister directions:

  • High-voltage array design to cut harness mass.
  • Mass-production panel standardization — the Starlink lesson is that the cost was never the cell.
  • Qualification economics: cheap, fast radiation-test campaigns that let new materials accumulate credible heritage without a decade-long dossier. This attacks the filing-cabinet moat directly, and it benefits every other R-direction.

R7 — Non-photovoltaic solar: the closing argument

One sweep, then retire.

Solar-thermal Stirling, thermionics, thermophotovoltaics with concentrators. Almost certainly dominated by photovoltaics on both $/W and W/kg — but the honest closing argument with current numbers has never been written. A single reading sweep retires the question permanently instead of leaving it as background noise in every trade study.

Sequencing

Direction Clock Posture
R1 space-ified silicon 0–3 yr Buy and characterize, don’t invent
R2 perovskite tandem 3–7 yr Fund the coupled-environment dataset and the test protocol
R3 ELO/IMM 3–7 yr Track reuse cycle count; engage when it crosses ~10
R4 alternate lattices — Watching brief
R5 in-space manufacturing 7–15 yr Ours to build; composes with the attachment roadmap
R6 architecture continuous Background duty; feeds all others
R7 non-PV one sweep Close it and move on

R2 and R3 are complementary bets: R2 escapes the materials, R3 escapes the geography. R1 carries the program while they mature. R5 is the destination the attachment architecture was quietly pointing at all along.