Perovskite/organic tandem solar cells pair a wide-bandgap perovskite front cell with a near-infrared organic bottom cell. Both absorbers can be solution-processed at low temperature, which makes the pairing attractive for lightweight, flexible and low-cost modules. The hard part is the perovskite. Raising its bandgap above 1.6 eV requires mixing bromide into the lattice, and mixed-halide films tend to separate into I-rich and Br-rich domains. This segregation costs voltage and undermines stability.
Segregation is usually described as a light-induced process, but it starts earlier. Br-rich phases nucleate faster than I-rich ones, so the film is already inhomogeneous by the time it has formed. Illumination then makes things worse. Our aim in this work was to remove the problem at its source by synchronizing the crystallization of the two halides.
Three bandgaps in parallel
There is no single wide-bandgap perovskite. Tandems on silicon, CIGS, narrow-gap perovskites or organics each need their own bandgap, and each bandgap means a different Br/I ratio and different crystallization behavior. Early on we decided to study three compositions at once: 1.62 eV, 1.68 eV and 1.88 eV, with bromide fractions from 10% to 50%. This tripled the workload, and for long stretches it felt like running three projects side by side. The benefit came later. Every mechanistic claim could be checked against the other two compositions immediately, and trends that would have been ambiguous in one system were clear across three.
A selective Lewis base
Lewis-base additives are a standard way to slow perovskite crystallization. Their weakness is that they bind every Pb2+ center in much the same way, so the head start of the Br-rich phase survives. We wanted an additive that prefers Pb in a Br-rich environment.
Hard-soft acid-base (HSAB) chemistry suggested a route. XPS shows that Pb in PbBr2 has a higher Pb 4f7/2 binding energy than in PbI2 (139.0 versus 138.5 eV), which indicates lower electron density and a harder Lewis acid. A harder Lewis base should bind it preferentially. We tuned the hardness of the oxygen donor by changing the anchoring group from boronic to carboxylic to phosphonic acid, then extending the alkyl chain. The series ran from 2-aminoethylboronic acid (2BA) to 6-aminohexylphosphonic acid (6PA), the hardest donor of the set.
Proving the selectivity was the slowest part of the project. Both halides compete for the same lead centers in one solution, and the spectroscopic differences between them are small. We spent months running XPS on films and 207Pb NMR on precursor solutions across all three compositions, repeating measurements whose shifts sat near the resolution limit. The shifts were small but reproducible: 6PA coordinates Pb more strongly in Br-rich environments than in I-rich ones.
In situ GIWAXS during spin coating showed what that selectivity does. In control films, diffraction appears earlier as the Br content rises, and the 1.68 eV and 1.88 eV compositions develop a separate Br-rich phase; at 1.88 eV it shows up before the antisolvent is even dropped. With 6PA, nucleation is delayed in all three compositions and the films crystallize into the target phase with minimal secondary phase.
The devices followed the films. Single-junction cells reached 25.1%, 24.6% and 18.9% at 1.62, 1.68 and 1.88 eV, against 24.2%, 23.6% and 17.6% for the controls. After 1,500 hours at 1 sun and 65 °C, they kept 95%, 94% and 90% of their initial efficiency. The controls started degrading within 200 hours and finished at 15% or below.
Building the tandem
We used the 1.88 eV perovskite as the front cell and a PM6:BTPSe-Ph4F:MO-IDIC-2F ternary blend as the organic bottom cell; on its own this near-infrared cell delivers 19.2%. The lesson of this stage was that two well-optimized sub-cells do not add up to a good tandem. The perovskite thickness sets how much light reaches the organic layer. The interconnecting layers affect both the films grown on top of them and the current matching between sub-cells. The organic layer thickness trades near-infrared absorption against charge collection. In the end we optimized the full stack layer by layer, and kept coming back to parameters we had considered settled.
The best tandem reached 28.1% efficiency, with a certified value of 27.0%, a record for perovskite/organic tandems. In third-party testing under the ISOS-L2 protocol at 65 °C, an encapsulated device retained 91% of its efficiency after 1,000 hours. What we value most is that a simple selectivity argument from coordination chemistry held at every bandgap we tried. We expect it to be useful wherever mixed-halide perovskites are grown.
More details are in our paper, “Phase-Homogeneous Mixed Halide Perovskites for Stable Tandem Photovoltaics” (https://doi.org/10.1038/s41586-026-10929-2).