TECHNOLOGY EXPLAINER
Perovskite-Silicon Tandems
— stacking a second solar cell on top of silicon
A perovskite-silicon tandem builds a perovskite solar cell directly on top of a crystalline silicon one. A cell of about 1 cm² now sits in the efficiency tables at 35.2% (initial efficiency, independently certified), and a module of about 1.7 m² has been certified at 29.4%. But the number you see depends heavily on the area and the measurement category.
- What a perovskite-silicon tandem is, in three points
- Where efficiency stands — the number changes with area and measurement category
- Two-terminal and four-terminal — two ways of wiring the stack
- Inside a two-terminal cell — top cell, recombination layer, bottom cell
- Current matching — the price of a series connection
- A materials engineer's view (1): the recombination layer is a seam, not just a thin electrode
- Coating over texture — from polished wafers to industry-standard pyramids
- A materials engineer's view (2): the 200 °C ceiling and the order of deposition
- Our calculation: the gap between a small cell and a large module
- Strengths and open problems
- Glossary / References / Claim-to-source audit
Sourced = stated in published material (link given)
Our calculation = a value this article derived from assumptions it states
Not yet confirmed = a plan, target or outlook with no confirmed track record
Structural readings and materials-design interpretations are marked separately as Commentary.
Every efficiency figure is given with whether it is certified, whether it is a cell or a module, and its area.
1. What a perovskite-silicon tandem is, in three points
A tandem (multijunction) solar cell is a stack of solar cells that absorb different colours (wavelengths) of light, sharing out the sunlight between them. In a perovskite-silicon tandem, the perovskite sits on top (the side the light reaches first) and crystalline silicon sits underneath.
- Why bother: single-junction silicon cells are approaching their theoretical limit. A 2026 perspective in Energy & Environmental Science (EES) explains that tandems with two or more junctions reduce the energy lost as heat and make use of a wider slice of the solar spectrumSourced
- How far it has come: Version 68 of the efficiency tables lists a LONGi two-terminal cell at 35.2% (area 0.9994 cm², measured by ESTI, initial efficiency)Sourced. That is above the best single-junction silicon result in the same tables, 28.1%
- What makes it hard: the recombination layer that connects the two cells in series, depositing films on a textured (bumpy) silicon surface, and designing the top and bottom cells to deliver the same current. All three are materials and process problems (our commentary)
Tandem efficiency has reached the stage where it is decided less by which cell is better than by how the two are joined. The EES perspective argues that the recombination junction (recombination layer) should be treated not as a secondary interlayer but as a critical bottleneckSourced. That junction is a stack of films a few tens of nanometres thick, which puts it squarely in the materials engineer's territory.
2. Where efficiency stands — the number changes with area and measurement category
When a headline says “35% for a tandem”, the first thing to check is over what area, and who measured it. The Solar cell efficiency tables list only results confirmed by an independent test centre, and they distinguish three definitions of area: total area (t), aperture area (ap) and designated illumination area (da)Sourced. The perovskite-silicon tandem entries in Version 68 are as follows.
| Category | Efficiency | Area | Test centre (date) | Made by |
|---|---|---|---|---|
| Cell (small area) | 35.2 ± 1.1% | 0.9994 cm² (da) | ESTI (February 2026) | LONGi, two-terminal |
| Cell (large area) | 34.3 ± 1.5% | 261.0 cm² (ap) | Fraunhofer ISE (November 2025) | LONGi, two-terminal |
| Cell (large area) | 32.2 ± 1.4% | 274.0 cm² (t) | Fraunhofer ISE (July 2025) | LONGi, two-terminal |
| Module | 31.4 ± 0.9% | 1,650.4 cm² | Fraunhofer ISE (November 2025) | LONGi |
| Module (large) | 29.4 ± 0.9% | 16,766 cm² | Fraunhofer ISE (November 2025) | LONGi |
All Sourced (efficiency tables Version 68 [Ref. 1]). These perovskite-containing entries carry the table note “Initial efficiency”. They are not stabilised values.
The text of the same paper also mentions a 221 cm² tandem cell made by Trina Solar that reached 32.15% (total area, measured by ISFH), and says this shows the performance gap between small-area and practical-size tandems is narrowingSourced. It further explains that efficiency targets in public research programmes are generally set for cells of 1 cm² or larger, and that measurement error tends to grow as cells get smallerSourced.
A company announcement that came after the tables
In July 2026 LONGi announced that its perovskite-silicon tandem cell had reached 35.5%, certified by ESTI (the European Solar Test Installation)Sourced. However, the announcement does not state the cell area, and at the time of our research we could not confirm the result in the efficiency tables. The company lists its earlier milestones as 33.9% (November 2023), 34.6% (June 2024) and 34.85% (April 2025, certified by the US NREL, two-terminal)Sourced.
• 35.5% is a company announcement, said to be certified by an independent test centre (ESTI), but the area is not given in the release
• 35.2% is the efficiency-table value, for an area of about 1 cm², and is an initial efficiency
• 29.4% is for a module of about 1.7 m², and is likewise initial performance
All three are “tandem efficiency”, yet they measure entirely different things (our commentary).
3. Two-terminal and four-terminal — two ways of wiring the stack
Tandems come in two broad forms, named after the number of electrical terminals used to take the power out.
| Item | Two-terminal (monolithic) | Four-terminal (mechanically stacked) |
|---|---|---|
| How the cells connect | The perovskite is deposited directly on the silicon and the two are connected in series through a recombination layer | Two separately made cells are stacked and their power is taken out independently |
| Advantages | A simpler structure that is easier to fit into existing production lines | Looser requirements on the recombination layer, and no need for current matching |
| Drawbacks | Requires careful design to match the top and bottom currents | Optical losses from the extra window and interface layers, and balance-of-system costs such as wiring |
All Sourced (Elsmani et al., Nanomaterials 2021 [Ref. 11]). Hooijer et al. note that there is no clear consensus on whether monolithic or mechanically stacked tandems will eventually dominate, and that the centre of gravity of research lies with monolithic (two-terminal) devices [Ref. 10].
From here on, this article concentrates on two-terminal (monolithic) tandems. The four new entries in Version 68 of the efficiency tables are all two-terminal, two-cell stacks with a perovskite top cellSourced, so this is where the race for records is being run.
4. Inside a two-terminal cell — top cell, recombination layer, bottom cell
Seen from the side the light enters, a two-terminal tandem has three parts. The exact layer combination differs from lab to lab and company to company, so the figure below is a typical configuration often used in papers, as organised by this article.
(1) Top cell: a wide-bandgap perovskite
To let light through to the silicon underneath, the top cell uses a perovskite with a larger bandgap than silicon. In the tandem with a certified efficiency of 29.15% reported in Science in 2020 by a group including HZB, the perovskite bandgap was 1.68 eVSourced. The 2023 report by Mariotti et al. (certified efficiency up to 32.5%) also uses a 1.68 eV “triple-halide” compositionSourced. For how composition relates to bandgap, see our explainer on bandgap and composition design.
(2) Recombination layer: the seam that connects the two cells in series
This layer takes the electrons coming out of the top cell and the holes coming out of the bottom cell and lets them meet and cancel out on the spot. That connects the two cells in series, so their voltages add. The earliest two-terminal cell, reported in 2015 by a group from MIT and Stanford University, did this with a silicon tunnel junction and showed an open-circuit voltage of 1.65 V and a stable efficiency of 13.7% on 1 cm²Sourced.
After a nanocrystalline silicon recombination junction (Sahli et al., 2018)Sourced, today's perovskite-silicon tandems are described as commonly using n-type Si (amorphous or nanocrystalline) / transparent conductive oxide / hole transport layer (SAM, NiOx and the like)Sourced.
(3) Bottom cell: crystalline silicon
Most record-class cells use a silicon heterojunction (SHJ) cell as the bottom cell. The LONGi cell reported in Nature in 2024 with a certified 33.89% (stabilised) was also built on an SHJ cell made from a Czochralski (Cz) waferSourced. There are also reports using the TOPCon design that dominates mass-produced silicon as the bottom cell; Nature Communications has published 31.1% (certified 30.9%) on 1 cm²Sourced.
5. Current matching — the price of a series connection
In a two-terminal tandem the two cells are connected in series. The current through a series circuit is capped by the weaker cell. The review by Elsmani et al. names current matching between the cells as the main constraint of the two-terminal designSourced. In the earliest 2015 cell, too, the perovskite side was reported to be limiting the currentSourced.
Real cells bear this out. The 261 cm² tandem cell in Version 68 of the efficiency tables has an open-circuit voltage of 2.013 V, a short-circuit current density of 20.04 mA/cm² and a fill factor of 85.0%Sourced. A single-junction silicon cell in the same tables (175.37 cm², 27.8%) has an open-circuit voltage of 0.7469 V and a short-circuit current density of 42.76 mA/cm²Sourced.
- Checking the efficiency: 2.013 V × 20.04 mA/cm² × 0.850 = 34.29 mW/cm². Against an incident 100 mW/cm² that is 34.3%, matching the tableOur calculation
- Same for the 274 cm² cell: 2.004 V × 19.89 mA/cm² × 0.809 = 32.25 mW/cm² → 32.2% (matches the table)Our calculation
- Current: 20.04 ÷ 42.76 = about 0.47. The tandem carries a little under half the current of single-junction silicon
- Voltage: 2.013 ÷ 0.7469 = about 2.7 times
Assumptions and limits: the two cells differ in area, maker and measurement date; this is not a like-for-like comparison of a tandem and a single-junction cell built on the same silicon. It is meant only to give a sense of the orders of magnitude.
- Assumption: hold the combined current of the two cells at 40.0 mA/cm², with (1) top 20.5 / bottom 19.5 and (2) both at 20.0 (hypothetical)
- Assumption: take the operating voltage at the maximum power point as 1.7 V (hypothetical; the open-circuit voltage of the tandems in the tables is about 2.0 V)
- Difference in series current: 20.0 − 19.5 = 0.5 mA/cm²
- Difference in output: 0.5 × 1.7 = 0.85 mW/cm², or about 0.85 points of efficiencyOur calculation
- Cross-check: the estimate by Hooijer et al. (0.1 mA/cm² → 0.15 to 0.2 points) is reproduced by 0.1 × 1.5 = 0.15 and 0.1 × 2.0 = 0.2Our calculation
Assumptions and limits: in a real two-terminal cell the cell with the surplus current operates away from its maximum power point, so the loss also depends on changes in fill factor. The only aim here is to convey the order of magnitude: a 1 mA/cm² difference in current is worth roughly one point of efficiency.
The levers for matching the currents are the thickness and bandgap of the top cell and control of light absorption and reflection in each layer. Hooijer et al. point out that making the protective layer on the recombination junction (SnO2 and similar) thicker increases parasitic absorption and upsets current matchingSourced. In other words, a contest over 1 mA/cm² can be decided by choosing a film thickness a few nanometres different (our commentary).
6. A materials engineer's view (1): the recombination layer is a seam, not just a thin electrode
Hooijer et al. list the requirements for an ideal recombination junction as followsSourced.
- Aligned energy levels across the electron transport layer, the hole transport layer and the conductive interlayer (a transparent conductive oxide or similar)
- Fast recombination of electrons and holes, keeping resistance and voltage loss low
- High transparency, to pass light to the cell below
- Chemical and thermal stability, plus the chemical and mechanical robustness to survive solution processing carried out afterwards
The last item is the one materials engineers will find most familiar. The same paper cites a measurement in which the C60/SnO2 interface had a fracture energy of only 1.2 J/m², calling it a weak point against the stresses expected from thermal cycling and outdoor useSourced. It also notes a case in which deliberately engineering the interface raised this to 160 J/m²Sourced.
There are electrical targets too. As a practical guideline, the contact resistivity of each recombination junction should not exceed 1 Ω·cm²; tunnel junctions in III-V multijunction solar cells achieve 10⁻⁴ Ω·cm², while perovskite recombination junctions are orders of magnitude higherSourced.
So the recombination layer is an electrode, an optical window, a solvent barrier and a mechanical seam, all at once. Adhesion, fracture toughness, solvent barrier performance and resistance to sputtering — properties that coatings, adhesives and thin-film engineers handle every day — here directly decide efficiency and lifetime (our commentary). Each of these layers is covered in more detail in our explainers on transparent conductive films and on self-assembled monolayers.
7. Coating over texture — from polished wafers to industry-standard pyramids
Mass-produced silicon solar cells have micrometre-scale pyramid texture etched into the surface to cut reflection and trap light. Perovskite films made by coating a solution, however, are much easier to make on a flat surface.
| Report | Silicon surface | What was published |
|---|---|---|
| Sahli et al. (Nature Materials, 2018) | Textured on both sides (fully textured) | Points out that earlier two-terminal tandems polished the front of the silicon to suit perovskite deposition, which brought higher cost, reflection losses and poor light trapping. Developed a process that deposits conformally (evenly following the shape) over micrometre-scale pyramids, reaching a certified steady-state efficiency of 25.2% and a current density of 19.5 mA/cm² |
| Chin et al. (Science, 2023) | Micrometre-scale pyramids (industry standard) | Coated the perovskite layer conformally over the pyramids, controlled crystallisation with an additive and reduced recombination losses at the interface with C60. Area 1.17 cm², certified efficiency 31.25% |
| Liu et al. / LONGi (Nature, 2024) | Mild texture on the front, strong texture on the rear | A “dual texture” that raises photocurrent without compromising rear-side passivation. With bilayer passivation from an ultrathin LiF layer and organic molecules, a certified stabilised efficiency of 33.89%, fill factor 83.0% and open-circuit voltage of about 1.97 V |
All Sourced (Sahli et al. [Ref. 5], Chin et al. [Ref. 7], Liu et al. [Ref. 9]). The three differ in area, test centre and date, so their efficiencies cannot be compared directly.
The review by Elsmani et al. collects cases in which solution-processing perovskite over texture lowers the fill factor, and the point that conformal deposition by evaporation suits small-scale work but is not well suited to large-scale productionSourced. Either keep the texture low (a mild texture) or adapt the deposition to the texture: choosing the substrate shape and the deposition method together has become the core of two-terminal tandem design (our commentary). Coating methods themselves are covered in our explainer on solution-based film deposition.
8. A materials engineer's view (2): the 200 °C ceiling and the order of deposition
A two-terminal tandem builds the perovskite on top of a finished silicon cell. So every process step for the layers above has to stay within the conditions the silicon cell underneath can withstand.
Elsmani et al. explain that the hydrogenated amorphous silicon (a-Si:H) layers used in heterojunction silicon are designed to withstand only temperatures below 200 °CSourced. They also point out that 200 °C metallisation (electrode formation) is something the perovskite, not the silicon, cannot withstandSourced.
In other words, this stack has a 200 °C ceiling imposed from both below and above. Firing to densify, annealing at high temperature to improve crystallinity: tools that materials development takes for granted are off the table from the start (our commentary).
The other constraint is order. Hooijer et al. note that SnO2 is used as a buffer layer to prevent damage to the underlying layers when the top transparent conductive oxide is deposited by sputteringSourced. Because a later step damages an earlier layer, a layer whose only job is protection gets added. That layer absorbs light (Section 5) and adds another interface (Section 6). A protective film that is dense yet made at low temperature, thin, transparent and well adhered: this is where the materials side has room to contribute (our commentary).
9. Our calculation: the gap between a small cell and a large module
- Area ratio: 16,766 cm² ÷ 0.9994 cm² = about 16,800 times (roughly 17,000)Our calculation
- Efficiency gap: 35.2 − 29.4 = 5.8 points. In relative terms, 29.4 ÷ 35.2 = about 0.84Our calculation
- Cell to cell: from about 1 cm² to 261 cm² (about 260 times the area), the gap is 0.9 pointsOur calculation
Assumptions and limits: module efficiency includes losses from cell interconnection, scribing (dividing into segments) and the inactive border, so the losses differ in kind from those of a cell. All of these are also initial efficiencies and say nothing about performance after long outdoor use. Long-term durability is covered in our explainer on outdoor testing and durability assessment.
The authors of the efficiency tables write that the large-area tandem cell results show the performance gap between small-area and practical-size devices is narrowingSourced. At the cell level, scaling up to 261 cm² keeps the gap within one pointOur calculation. The gap opens up at the module stage.
10. Strengths and open problems
(1) The numbers in this article say nothing about long-term durability
The perovskite entries in the efficiency tables carry the note “Initial efficiency”Sourced. Al-Ashouri et al. reported in 2020 that an unencapsulated tandem kept 95% of its initial efficiency after 300 hours of operation in airSourced, but that is a different thing from assessing a practical lifetime of several decades. How many years these cells last outdoors could not be confirmed in the primary sources within the scope of this article.
(2) The practical ceiling is still an outlook
Hooijer et al. note that while efficiencies close to 35% have been reported at the 1 cm² scale, the practical limit is expected to exceed 38 to 39%Not yet confirmed. In its own announcement LONGi puts the theoretical efficiency limit of perovskite-silicon tandems at 43%Not yet confirmed. Both are calculated outlooks, not demonstrated values.
(3) Two-terminal or four-terminal: not yet settled
Research is centred on two-terminal devices, but there is said to be no clear consensus on which will eventually dominateNot yet confirmed.
(4) Timing and scale of mass production are outside this article
This article does not cover mass-production plans, production capacity or prices, because company announcements on these vary widely in character. At the time of our research (September 2026), full-scale mass production of perovskite-silicon tandems could not be confirmed in independent primary sources.
- In Version 68 of the efficiency tables: 35.2% for a cell of about 1 cm² and 29.4% for a module of about 1.7 m² (both initial efficiency, independently measured)Sourced
- A two-terminal tandem is a series connection, so the top and bottom currents must match. The tandem carries a little under half the current of single-junction silicon at about 2.7 times the voltageOur calculation
- The recombination layer is commonly n-type Si / transparent conductive oxide / SAM or similar, and must deliver low contact resistance, transparency, solvent resistance and mechanical strength all at onceSourced
- The technology has moved from polished wafers to deposition on industry-standard pyramid textureSourced
- Neither the silicon underneath nor the perovskite can go much above 200 °C. Dense protective films and transparent electrodes made at low temperature are the materials-side issue (commentary)
- The gap between cell and module is 5.8 points, and it opens up at the module stageOur calculation
11. Glossary
- Tandem (multijunction) solar cell
- A structure that stacks solar cells absorbing different wavelengths so that they share the work of converting sunlight.
- Two-terminal (monolithic)
- Top and bottom cells built directly on one substrate and connected in series. Two terminals.
- Four-terminal (mechanically stacked)
- Separately made cells stacked on top of each other, with power taken from each. Four terminals.
- Top cell / bottom cell
- The upper cell that the light reaches first, and the lower cell that receives what is left.
- Recombination layer (recombination junction)
- The layer between the two cells where electrons and holes meet and cancel, connecting the cells in series.
- Tunnel junction
- A thin, heavily doped p-n junction through which charge passes by tunnelling. One kind of recombination layer.
- Current matching
- Equalising the currents of series-connected top and bottom cells. The weaker one sets the overall current.
- Texture
- Micrometre-scale pyramid-shaped relief formed on a silicon surface to reduce reflection.
- Conformal deposition
- Depositing a film that follows the relief of the substrate with an even thickness.
- SHJ (silicon heterojunction)
- A cell structure in which the crystalline silicon surface is covered with thin amorphous silicon films. Made with low-temperature processes.
- TOPCon
- A silicon cell structure that passivates the surface with an ultrathin oxide and a polycrystalline silicon layer.
- SAM (self-assembled monolayer)
- A film one molecule thick that bonds to the underlying surface. Used as a hole-extraction layer.
- Initial efficiency
- Efficiency just after fabrication, not a value after changes caused by light or heat over time.
- Designated illumination area (da) / aperture area (ap) / total area (t)
- The area definitions used in the efficiency tables. They differ in what the mask covers.
12. References (primary sources)
- Green, M.A. et al. “Solar cell efficiency tables: Version 68”, Joule 10, 102494 (2026) https://doi.org/10.1016/j.joule.2026.102494
- LONGi “35.5%! LONGi Once Again Breaks World Record for Crystalline Silicon-Perovskite Tandem Solar Cell Efficiency”, 15 July 2026 https://www.longi.com/en/news/crystalline-silicon-perovskite-tandem-solar-cell-new-world-efficiency-2026/
- LONGi “34.85%! LONGi Breaks World Record for Crystalline Silicon-Perovskite Tandem Solar Cell Efficiency Again”, 16 April 2025 https://www.longi.com/en/news/silicon-perovskite-tandem-solar-cells-new-world-efficiency/
- Mailoa, J.P. et al. “A 2-terminal perovskite/silicon multijunction solar cell enabled by a silicon tunnel junction”, Applied Physics Letters 106, 121105 (2015) https://doi.org/10.1063/1.4914179
- Sahli, F. et al. “Fully textured monolithic perovskite/silicon tandem solar cells with 25.2% power conversion efficiency”, Nature Materials 17, 820–826 (2018) https://doi.org/10.1038/s41563-018-0115-4
- Al-Ashouri, A. et al. “Monolithic perovskite/silicon tandem solar cell with >29% efficiency by enhanced hole extraction”, Science 370, 1300–1309 (2020) https://doi.org/10.1126/science.abd4016
- Chin, X.Y. et al. “Interface passivation for 31.25%-efficient perovskite/silicon tandem solar cells”, Science 381, 59–63 (2023) https://doi.org/10.1126/science.adg0091
- Mariotti, S. et al. “Interface engineering for high-performance, triple-halide perovskite–silicon tandem solar cells”, Science 381, 63–69 (2023) https://doi.org/10.1126/science.adf5872
- Liu, J. et al. (LONGi) “Perovskite/silicon tandem solar cells with bilayer interface passivation”, Nature 635, 596–603 (2024) https://doi.org/10.1038/s41586-024-07997-7
- Hooijer, R. et al. “Why more junctions do not yet deliver: interconnection challenges in perovskite multijunction solar cells”, Energy & Environmental Science 19, 4582–4596 (2026, open access) https://doi.org/10.1039/d6ee01631f
- Elsmani, M.I. et al. “Recent Issues and Configuration Factors in Perovskite-Silicon Tandem Solar Cells towards Large Scaling Production”, Nanomaterials 11, 3186 (2021, open access) https://doi.org/10.3390/nano11123186
- Luo, Y. et al. “Inductive effects in molecular contacts enable wide-bandgap perovskite cells for efficient perovskite/TOPCon tandems”, Nature Communications 16, 4516 (2025, open access) https://doi.org/10.1038/s41467-025-59896-8
13. Claim-to-source audit
| Claim in the text | Basis | Label |
|---|---|---|
| The perovskite-silicon tandem entries in Version 68 of the efficiency tables: 35.2 ± 1.1% (0.9994 cm², designated illumination area, ESTI, February 2026, LONGi, two-terminal, Voc 1.998 V); 34.3 ± 1.5% (261.0 cm², aperture area, FhG-ISE, November 2025, Voc 2.013 V, Jsc 20.04 mA/cm², FF 85.0%); 32.2 ± 1.4% (274.0 cm², total area, FhG-ISE, July 2025, Voc 2.004 V, Jsc 19.89 mA/cm², FF 80.9%). Modules at 31.4 ± 0.9% (1,650.4 cm²) and 29.4 ± 0.9% (16,766 cm², “about 1.7 m²” in the text), both LONGi, FhG-ISE, November 2025. That these carry the note “initial efficiency / initial performance”. Single-junction silicon at 27.8% (175.37 cm², Voc 0.7469 V, Jsc 42.76 mA/cm²) and 28.1% (139.70 cm²). That the four new entries are all two-terminal, two-cell stacks with a perovskite top cell. Trina Solar's 32.15% on 221 cm² (ISFH). The statement that the gap between small-area and practical-size performance is narrowing. That efficiency targets in government research programmes are generally set for 1 cm² or larger. The definitions of the three area categories | Efficiency tables Version 68 (Joule, 2026). Reference 1 https://doi.org/10.1016/j.joule.2026.102494 | Sourced |
| That at a conference on 14 July 2026 LONGi announced a perovskite-silicon tandem cell efficiency of 35.5% (certified by ESTI). The progression 33.9% (November 2023), 34.6% (June 2024), 34.85% and 35.2%. That the release does not state the cell area. That it puts the theoretical efficiency limit at 43% | LONGi news release (15 July 2026). Reference 2 https://www.longi.com/en/news/crystalline-silicon-perovskite-tandem-solar-cell-new-world-efficiency-2026/ | Sourced |
| That 34.85% was a two-terminal cell certified by the US NREL (April 2025) | LONGi news release (16 April 2025). Reference 3 https://www.longi.com/en/news/silicon-perovskite-tandem-solar-cells-new-world-efficiency/ | Sourced |
| That a two-terminal monolithic tandem was made with a silicon tunnel junction, giving on 1 cm² a Voc of up to 1.65 V and a stable efficiency of 13.7%, with the perovskite side limiting the current | Mailoa et al. (Applied Physics Letters, 2015). Reference 4 https://doi.org/10.1063/1.4914179 | Sourced |
| That earlier two-terminal tandems polished the front of the silicon, at a disadvantage in cost, reflection loss and light trapping. Conformal deposition over micrometre-scale pyramids. A nanocrystalline silicon recombination junction. Certified steady-state efficiency of 25.2% and current density of 19.5 mA/cm² | Sahli et al. (Nature Materials, 2018). Reference 5 https://doi.org/10.1038/s41563-018-0115-4 | Sourced |
| Certified efficiency of 29.15%, perovskite bandgap of 1.68 eV, a methyl-substituted carbazole self-assembled monolayer as the hole-selective layer, Voc up to 1.92 V, and 95% of initial efficiency retained after 300 hours of operation in air without encapsulation | Al-Ashouri et al. (Science, 2020). Reference 6 https://doi.org/10.1126/science.abd4016 | Sourced |
| Conformal coating over industry-standard micrometre-scale pyramids, crystallisation control with an additive, reduced recombination loss at the interface with C60, and a certified efficiency of 31.25% on 1.17 cm² | Chin et al. (Science, 2023). Reference 7 https://doi.org/10.1126/science.adg0091 | Sourced |
| A tandem Voc of 2.00 V and certified efficiency of up to 32.5% using a 1.68 eV triple-halide composition and interface modification | Mariotti et al. (Science, 2023). Reference 8 https://doi.org/10.1126/science.adf5872 | Sourced |
| A dual-texture structure on an SHJ cell from a Cz wafer, mild on the front and strong on the rear. Bilayer passivation from an ultrathin LiF layer and organic molecules. Certified stabilised efficiency of 33.89%, FF 83.0% and Voc of about 1.97 V | Liu et al. (Nature, 2024). Reference 9 https://doi.org/10.1038/s41586-024-07997-7 | Sourced |
| That multijunctions reduce thermalisation loss and use the spectrum more widely. That there is no clear consensus on whether monolithic or mechanically stacked tandems will dominate. Close to 35% at the 1 cm² scale, with a practical limit expected above 38 to 39%. The argument that the recombination junction should be treated as a critical bottleneck. That perovskite/silicon recombination junctions consist of n-type Si / TCO / hole transport layer (SAM, NiOx and the like). The requirements for an ideal recombination junction. That SnO2 is used as a buffer layer against sputter damage. That thick protective layers upset current matching through parasitic absorption. The estimate that a 0.1 mA/cm² current loss equals 0.15 to 0.2 points at 1.5 to 2.0 V. The guideline of contact resistivity at or below 1 Ω·cm² and the 10⁻⁴ Ω·cm² of III-V devices. A C60/SnO2 interface fracture energy of 1.2 J/m², and 160 J/m² with interface engineering | Hooijer et al. (Energy & Environmental Science, 2026). Reference 10 https://doi.org/10.1039/d6ee01631f | Sourced |
| That two-terminal devices fit more easily into existing production lines but are constrained by current matching. That four-terminal devices relax the recombination-layer requirements but incur optical losses from window and interface layers and balance-of-system costs such as wiring. That a-Si:H layers are designed to withstand only temperatures below 200 °C. That the perovskite side cannot withstand 200 °C metallisation. Cases of lower fill factor from solution deposition over texture, and that conformal deposition by evaporation is not suited to large-scale production | Elsmani et al. (Nanomaterials, 2021). Reference 11 https://doi.org/10.3390/nano11123186 | Sourced |
| A tandem with a TOPCon-type bottom cell reaching 31.1% (certified 30.9%) on 1 cm² | Luo et al. (Nature Communications, 2025). Reference 12 https://doi.org/10.1038/s41467-025-59896-8 | Sourced |
| A practical limit above 38 to 39%, a theoretical limit of 43%, and which of two-terminal and four-terminal will ultimately prevail | All are outlooks or calculated values, not demonstrated results. Reference 10 https://doi.org/10.1039/d6ee01631f | Not yet confirmed |
| Checking the efficiencies of the 261 cm² and 274 cm² cells as Voc × Jsc × FF. The tandem-to-single-junction current ratio of about 0.47 and voltage ratio of about 2.7. The current-matching example in Fig. 3 (20.5, 19.5 and 20.0 mA/cm², 1.7 V operating voltage, about 0.85 points). The cross-check of 0.1 mA/cm² to 0.15–0.2 points. The area ratio of about 16,800, efficiency gaps of 5.8 and 0.9 points and relative value of about 0.84. The gaps in Fig. 5 (−0.9 to −5.8 points) | Our calculation. The currents and operating voltage in Fig. 3 are hypothetical values set by this article, not measurements of a real cell. The tandem and single-junction silicon cells differ in maker, area and measurement date. Area definitions differ by row, so not all of the gap is an area effect | Our calculation |
| The cell area of the 35.5% result, and its listing in the efficiency tables | LONGi's release does not give the area and the listing could not be confirmed at the time of our research, so no area is stated (commentary by this article) | Commentary |
| Long-term outdoor durability, and the timing and scale of mass production | Not stated because they could not be confirmed in the primary sources within the scope of this article (commentary by this article) | Not yet confirmed |
| Viewing the recombination layer as an electrode, optical window, solvent barrier and mechanical seam. The reading that a 200 °C ceiling applies from both below and above. The framing that choosing the substrate shape and the deposition method together is the core of the design. The order-of-magnitude sense that a 1 mA/cm² current difference is worth about one point of efficiency. The two-way split into strengths and challenges. Grouping the layers in Fig. 2 into three parts | This article's own framing and commentary based on published content. Not views expressed by the institutions or companies | Commentary |
| That Figs. 1, 2, 3, 5 and 6 are explanatory drawings, and that the hero image and Fig. 4 are AI-generated images | A note by this article (commentary) | Commentary |
Last updated 25 September 2026. Sources are limited to primary material (the efficiency tables, peer-reviewed papers and official company announcements). Every efficiency value is given with whether it is certified, whether it is a cell or a module, its area and the test centre. The cell area of the 35.5% result, long-term outdoor durability, and the timing and scale of mass production are not stated, because they could not be confirmed in published primary sources. All figures are for explanation. Figs. 1, 2, 3, 5 and 6 are vector drawings, and the hero image and Fig. 4 are AI-generated images; none of them shows a real cross-section, micrograph or physical product.