Silicon has a ceiling. The theoretical conversion limit for a single-junction crystalline silicon cell is about 29.43%, and the industry is close enough to see it. That is why every major manufacturer is working on perovskite-silicon tandem cells, where the theoretical limit rises above 43%.
Laboratory records are already past 35%. The more interesting question for anyone specifying a project is different: what do tandem modules actually do when they are installed outdoors, in a desert or on a plateau, for a year or two? A small but growing set of Chinese demonstration plants has started to answer it.
Why tandem works
A single junction absorbs only the part of the spectrum it is tuned for. Everything above its bandgap loses the excess energy as heat; everything below passes straight through. Stacking two absorbers with different bandgaps splits the spectrum.
The wide-bandgap perovskite top cell acts as a high-energy catcher, absorbing blue, green and red photons. The narrow-bandgap silicon bottom cell catches the near-infrared light that passes through. The result is a theoretical efficiency above 43% — a genuinely different regime from single-junction silicon.
Two configurations are in development. Four-terminal (4T) devices stack two independent cells with a transparent electrode and separate circuits; two-terminal (2T) devices are monolithically integrated with a tunnel junction. The 4T route is simpler electrically but more expensive in materials; the 2T route is more compact but demands tighter process control.
The demonstration fleet
China Three Gorges Renewables (CTG), one of the earliest state-owned developers to invest in perovskite field applications, has built a series of projects across four very different climates — desert, plateau, humid subtropical and cold northern conditions.
| Date | Site | Scale | Technology | Purpose |
|---|---|---|---|---|
| Nov 2023 | Kubuqi Desert, Inner Mongolia | 1 MW, 11,200 modules | Single-junction perovskite | First commercial megawatt-scale perovskite plant; operation under drought, wind-blown sand and high UV |
| Apr 2025 | Yicheng, Shandong | 500 kW | 4-terminal perovskite/TOPCon tandem | First square-metre-class 4T tandem demonstration plant, with side-by-side comparison arrays |
| Nov 2025 | Haixi, Qinghai | 5 MW | Perovskite | High-altitude, high-UV, extreme-temperature operating data |
The Shandong project is the most informative, because it was designed as an experiment rather than a showcase. Of its 28 arrays, two were deliberately allocated to single-junction perovskite and to pure TOPCon silicon. Tandem, single-junction perovskite and conventional silicon therefore run under identical irradiance, temperature, wind and rain, and produce current under the same conditions. Three technologies, one field, one dataset.
What the data shows
The headline result, reported by researchers at CTG's research institute, is a 10% generation gain for perovskite-silicon tandem modules compared with single-junction silicon under the same conditions at Yicheng.
The manufacturing numbers are arguably more important than the field gain. The latest production data for four-terminal perovskite/BC tandem modules reports a best mass-production efficiency of 26.49%, with the modules passing the full IEC 61730 and IEC 61215 test sequences — the two core international standards, covering safety and performance respectively. A product that has passed both is a different commercial proposition from a laboratory champion.
The square-metre-class 4T perovskite/TOPCon module and its hundred-megawatt-class manufacturing process were also included in China's National Energy Administration list of first-of-a-kind major technical equipment (fifth batch) — a signal that the technology is being treated as strategically significant rather than as a research curiosity.
The four gates still standing
None of this means tandem is ready to replace silicon. Four constraints remain, and they are the same ones the industry has been discussing for years.
1. Long-term stability
Perovskite is intrinsically sensitive to moisture, oxygen, light and heat, and early laboratory cells discoloured and degraded quickly in ambient conditions. Encapsulation improvements and material engineering have brought short-term stability to an acceptable level, but long-term behaviour is still unproven.
Silicon modules are warranted for 25–30 years and have decades of field data behind that. For tandem, the outdoor degradation law is not yet established, and reliability under extreme temperature cycling, intense UV, high humidity and salt spray needs longer real-world observation. This is the single biggest reason investors hesitate.
2. The laboratory-to-factory gap
Centimetre-scale tandem cells in research laboratories keep setting records; square-metre production modules come in significantly lower. Tandem manufacturing involves thin-film deposition, coating and encapsulation, and circuit integration — far more process steps than a silicon cell, with correspondingly more opportunities for yield loss. Industry-wide mass-production yield is still low to middling, and consistency at scale is not yet assured.
3. Cost
Four-terminal structures need transparent electrodes and independent circuits, which raises both material and process cost. Two-terminal structures are simpler but demand higher-precision materials, processes and equipment, with corresponding capital intensity. Cost per watt for tandem modules remains well above mainstream n-type silicon, and although the efficiency is higher, the levelised cost of electricity advantage is not yet decisive. Parity is some way off.
4. Supply chain and standards
Perovskite precursors, high-end transparent conductive films and precision deposition equipment are all in short supply, with few suppliers and small volumes. Technology routes are not standardised either: 4T, 2T and different silicon bottom cells produce parameters that are hard to compare, and there are no unified product, test or warranty standards. For a bankable project, that is a real obstacle.
Underneath all four is a data problem. Silicon has decades of field data; tandem has a couple of years. Investors and insurers price uncertainty, and the absence of multi-climate, multi-year performance data makes them conservative.
What it means for buyers
For the next few years, tandem belongs in demonstration and pilot projects, not in utility-scale procurement. That is not a criticism of the technology — it is where every new cell architecture starts, including PERC and TOPCon.
If you are evaluating tandem products, the questions that matter are concrete:
- Which IEC standards have been passed, for the exact bill of materials being offered?
- What warranty length and degradation curve is being offered, and by whom — a manufacturer, or a project-specific special purpose vehicle?
- How much outdoor data exists, in which climates, over what period?
- What is the cost per watt and the modelled LCOE against a current n-type silicon module on the same site?
The trajectory is clear enough. Perovskite-silicon tandem is the most credible route past the single-junction silicon limit, and the demonstration results — a 10% field gain, IEC-certified 26.49% production modules — are genuine progress. But the gap between a certified module and a 25-year asset is closed by field data, not by records. Until that data exists, silicon remains the safe specification, and tandem remains the technology to watch.
ZOEAST PV's position on tandem
ZOEAST PV keeps crystalline silicon as the bankable baseline. Perovskite-silicon tandem is evaluated for pilot and demonstration projects, and for product categories where area and weight matter more than cost per watt — flexible modules from 120 W to 580 W, and building-integrated façades where the module is a construction material rather than a commodity.
Our criterion is deliberately conservative: we will specify tandem where the module carries the full IEC 61730 and IEC 61215 sequence for the exact bill of materials, where a warranty counterparty exists beyond the project, and where field data from at least two climates is available. Until those conditions are met, silicon — TOPCon or BC — remains the specification we will put our name on.
See the flexible module range and the colour module range, or talk to us about pilot-scale evaluation.
Photo: from the original article.
