Perovskite Tandems Hit 34.6%: The Silicon Era Is Ending

A single semiconductor can only convert photons within a narrow energy band efficiently. Silicon (1.1 eV bandgap) captures red and infrared brilliantly but wastes high-energy blue photons as heat. Perovskite tunes anywhere from 1.5-2.3 eV by adjusting the halide composition. Stack a 1.7 eV perovskite on top of silicon, and the perovskite absorbs blue/green light while silicon mops up red/IR. Combined, they harvest 86% of the solar spectrum.

The Shockley-Queisser limit (29.4% for silicon) assumes one junction. Two junctions raise the limit to 45%; three to 51%. Oxford PV's 34.6% is comfortably inside the two-junction limit and there's headroom to 38-40% before triple-junctions become necessary.

Perovskite's original Achilles heel was moisture and UV degradation. 2018-era cells lost 30% efficiency in 6 months of outdoor exposure. Three innovations changed this: (1) 2D-3D perovskite heterostructures that resist ion migration; (2) atomic layer deposition (ALD) of alumina passivation layers under 20 nm thick; (3) butyl-rubber edge seals borrowed from automotive glass.

Oxford PV's 5-year accelerated lifetime testing (equivalent to 25 years outdoor) shows 91.4% of initial efficiency retained — better than the 87% guarantee on most silicon panels. NREL's parallel testing on 144 commercial-format cells confirmed the result with 0.38%/yr median degradation.

Perovskite cells are made by solution processing — essentially inkjet printing on glass — at 100°C, vs the 900°C diffusion furnaces silicon requires. Capex per GW of production capacity is roughly £140M for perovskite vs £620M for silicon. That cost gap is why China's silicon stranglehold may be the wrong asset to own a decade from now.

Oxford PV's Brandenburg plant cost £104M for 100 MW; the next-phase 1 GW expansion is costed at £680M and ships from Q3 2026. By 2030, BloombergNEF projects 28 GW of global tandem capacity — still <8% of total PV manufacturing, but enough to set the price floor.

Q3 2026: First commercial deployments on rooftop, primarily commercial and utility scale. Q1 2027: Residential rooftop modules from First Solar, Meyer Burger, and a UK-assembled product from Oxford PV's partnership with British Solar Renewables. Q2 2027: driveway.solar will ship perovskite-tandem floor tiles at 280-310 W/m² — pending UKCA certification and field validation.

Early adopters will pay a 28-35% premium per watt. Levelised cost of energy still wins because the higher output amortises the cost faster, especially in lower-light UK conditions where the perovskite top cell continues working at dawn and dusk when silicon barely produces.

Three risks: lead toxicity (perovskite cells contain ~0.5 g/m² of lead — well below EU RoHS exemption limits but a recycling consideration); long-term outdoor data (we only have 8 years of real-world results, vs 40 for silicon); and Chinese capex catching up on solution processing, which could compress margins before Western manufacturers reach scale.

None of these are deal-breakers. All of them deserve serious engineering attention. The trajectory is clear: by 2032, single-junction silicon will look the way thin-film amorphous looks today — a transitional technology that did its job.