surya.narada.systems

R1: the design point, resolved

Sweep executed 18 August 2026. The R1 annealing baseline left the program with a design point — 100 µm n-type — that was undemonstrated on both of its axes. This page resolves it against the full evidence: manufacturability, the annealing cliff, architecture, and system-level anchors.

design point, adoptedGa-doped p-type SHJ at 60–90 µm, with a tracked path to 20–40 µm. Doping: p-type gallium — n-type is dropped; Cz n(P) shows no low-temperature annealing while p-type recovers, and Ga avoids boron–oxygen LID. Architecture: SHJ — symmetric, low-temperature, tray-based processing that tolerates thin wafers; TOPCon is the tracked alternative (CEA's 60 µm space-TOPCon matched PERC post-irradiation at 3× lower mass) but is less adapted below ~60 µm. Thickness: 60–90 µm is the band where manufacturing heritage and certified annealing (97% recovery at 90 µm, 80 °C, 10¹⁴ e/cm², ISFH-certified) already meet; 20–40 µm is where the strongest self-curing lives and where Solestial already flies, with module-level polymer support replacing wafer handling.

Manufacturability: the handling cliff was a snapshot

  • The cliff claim, traced — The "10% breakage at 100 µm → 96% at 70 µm" figure traces to one paper: Harrison et al., CEA-INES SHJ pilot line, EU PVSEC 2016 — production mode on 156 mm wafers, with handling, wet processing, and printing as the killer steps. It is a state-of-the-line snapshot from 2016, not a material law.
  • The cliff, mitigated — Terheiden et al. 2015: <5% breakage at 100 µm through full cell processing with engineered handling (Hanwha Q CELLS pilot line). Cariou et al. 2023: CEA itself now runs 60 µm on the same line class. Wu et al. 2024: edge-reinforcement rings give ~0% breakage on 28 µm free-standing wafers vs >85% for all-thin controls.
  • Below the wafer entirely — Kerfless routes bypass sawing: porous-silicon layer transfer delivered a 43 µm cell at 19% (Petermann 2012) with near-100% detachment yield demonstrated by 2015. Bond-to-substrate heritage: AstroPower grew 50 µm Si on ceramic; Wu 2024 holds free-standing 4.7 µm 4-inch wafers.
  • Flown practice — Solestial ships ~20 µm SHJ cells packaged in polymer/ceramic films — module mass 508 g/m², rollable to a 5 cm bend radius, flown on Transporter-10 (March 2024). The thin Si is handled at module level inside the film stack, not as a free wafer.

established Handling yield is engineering, not physics: the same institution that published the 96% figure in 2016 now manufactures 60 µm cells on the same line class. The directions page's breakage bullet is hereby corrected in place — the numbers were real, their interpretation as a fixed constraint was not.

The annealing cliff moves

DatasetWhat it shows
Solestial/ASU, 2023–2420 µm full recovery (120 mV/hr Voc, electrons; 7.5 mV/hr, protons), 40 µm intermediate (graphical only), 80 µm near-zero — at 10¹⁵ e/cm² on Cz p(B), 1–2 cells per condition
CEA/INES, Oct 202590 µm and 60 µm Ga-doped SHJ recovered 97% of initial performance after 10¹⁴ e/cm², cured at 80 °C, certified by ISFH-Caltec
Read togetherThe cliff is not a thickness wall. Between the two datasets the fluence dropped 10×, the doping changed, and the processing changed — and 90 µm went from "near-zero" to "97% recovered." Doping and processing set the design point first; thickness second.
Mechanism statusREDR theory (Lang–Kimerling) predicts the annealing rate scales with the injection-to-defect ratio — which thinning raises — but no published model computes this against thickness. Hydrogen's role is explicitly unresolved: hydrogenation supercharges recovery in Ga-PERC (Li 2025), yet CEA reports bulk hydrogenation did not influence post-irradiation lifetime before self-healing (Enjalbert, SiliconPV). The p/n asymmetry is consistent with the defect census: the p-type lifetime killer (BᵢOᵢ-class interstitial complexes) anneals under forward current, while n-type's E-center and divacancy need 150–400 °C.

System anchors

OptionSpecific powerCostNote
III-V rigid panel100–150 W/kg$200–400/WIncumbent benchmark (Solestial SmallSat 2024 table)
III-V flexible module250–350 W/kg$400–600/WIncumbent, flexible
Commodity-Si flexible50–200 W/kg$5–20/WTerrestrial-derived, no radiation engineering
Solestial thin-Si module500–600 W/kg (target)$20–40/W (target)Company targets, marked post-2025; measured anchor: 508 g/m² module mass, >18% AM0 BOL mini-module
NASA SBIR target200 W/kg array BOL$40/WTechPort 125378; ≥10× cost reduction vs state of practice

Honesty check on the anchors: the 500–600 W/kg and $20–40/W figures are company targets, not measurements. What is measured or certified: 508 g/m² module mass (self-published), >18% AM0 BOL mini-modules, 97% recovery at 90 µm (ISFH-certified), 60 µm beating PERC EOL at every fluence (peer-reviewed). The gap between target and measured module-level W/kg is itself a watch item.

What remains unknown

  • Cliff location at the design point — No published thickness-resolved annealing rates at 80–100 µm at matched fluence and processing. One ground campaign — irradiate-and-anneal across 60/80/90/100 µm at 10¹⁴ and 10¹⁵ e/cm² — resolves it. CNES/CEA have opened a PhD on exactly this mechanism space (project 26-254, posted January 2026).
  • Peer-reviewed thin-Si annealing article — The core dataset remains a workshop poster + conference paper + press releases. SPRAT-29 is August 2026; Guillot et al. (Solar RRL 2025, thickness × composition in irradiated Ga-SHJ) is published but was bot-walled at sweep time.
  • Module-level specific power, measured — No peer-reviewed measured W/kg for a coverglass-free thin-Si module exists; 500–600 W/kg is a company target, 508 g/m² is self-published.
  • ATOX protection for coverglass-free thin Si — No published ground-test (plasma asher) or flight erosion data on modern thin-Si polymer-encapsulated modules. Heritage says dielectric coatings work (LDEF); thin-Si-specific data does not exist.
  • In-orbit recovery telemetry — CEA's October 2025 headline says "in orbit" — the underlying flight dataset is not yet published. Company claims (Transporter-10 "nominal") stand without public curves.

What this changes

The R1 direction is no longer "characterize annealing at a guessed design point" — the design point is set, and the work splits three ways. Near-term: the single decisive ground campaign (thickness-resolved irradiate-and-anneal at 60–100 µm, matched fluence and processing) to convert the adopted band from evidence-backed to established. Watch: SPRAT-29 (August 2026), the CEA in-orbit dataset behind the October 2025 headline, Guillot et al. full text, Solestial flight curves. Standing risk: every quantitative annealing number above traces to two groups (Solestial/ASU and CEA/INES) — independent replication remains the win condition on the directions page, and it is not yet met.

Perimeter and provenance

Three parallel literature sweeps, 20 new sources (plus six carried from the annealing baseline), admitted to the Surya epistemic graph as event ev-000000000009 with assessmentsurya-assessment-r1-design-point-2026-08, judgmentestablished-with-gap. Sweep-level gaps, flagged honestly: Harrison 2016 is paywalled — its figures arrive via Wu et al. 2024's paraphrase; Guillot (Solar RRL) and Le (Materials Today Energy) full texts were bot-walled; the Solestial SmallSat comparison table was reconstructed from a jumbled PDF text layer; the 40 µm annealing point exists only as a curve on a poster; SPRAT-27 has no published proceedings. The adopted design point does not rest on any of the bot-walled items.

Sources