23.5% efficient monolithic perovskite/organic tandem solar cells based on an ultra-thin metal-like metal-oxide interconnect.


 Multi-junction solar cells provide an avenue to overcome fundamental efficiency limits of single-junction devices. The facile bandgap tunability of metal-halide perovskite solar cells renders them attractive building blocks for multi-junction architectures. Combinations with crystalline silicon and copper indium gallium selenide (CIGS) cells have been reported. All-perovskite tandem cells have likewise shown promising results. Meanwhile, narrow-gap non-fullerene acceptors (NFA) have revived the area of organic solar cells (OSCs) and unlocked skyrocketing efficiencies. Organic and perovskite semiconductors share similar processing technologies, which renders them attractive partners in multi junction architectures. As of yet, perovskite/organic tandem cells show subpar efficiencies of 20 per cent, limited by the low open circuit voltage (Voc) of wide-gap perovskite cells and losses introduced by the interconnect between the sub-cells. Here, we demonstrate two-terminal p-i-n perovskite/organic tandem cells with an efficiency of 23.5 per cent and a high Voc of 2.15 volts, operating near the levels predicted by a semi-empirical model. The perovskite sub-cells with optimized charge extraction layers afford an unsurpassed combination of a high Voc and fill-factor. The organic back-cells provide a high external quantum efficiency in the near-infrared. In surprising contrast to paradigmatic concerns about limited photostability of non-fullerene cells, we evidence an outstanding operational stability if excitons are predominantly generated on the NFA, which is the case in a tandem cell, where the illumination is spectrally filtered by the perovskite cell. A novel interconnect based on an ultra-thin (1.5 nanometers) metal like indium oxide layer offers unprecedented low optical/electrical losses. This work sets a new milestone for perovskite/organic tandem devices, that outperform the best p-i-n perovskite single junctions and are at par with perovskite/CIGS and all-perovskite multi-junctions. Perovskite/organic tandem architectures bear a realistic potential to reach an efficiency above 31%.


voltage (Voc) of wide-gap perovskite cells 8 and losses introduced by the interconnect between the sub-cells. 9, 10
Here, we demonstrate two-terminal p-i-n perovskite/organic tandem cells with an efficiency of 23 This work sets a new milestone for perovskite/organic tandem devices, that outperform the best p-i-n perovskite single junctions 12 and are at par with perovskite/CIGS and all-perovskite multi-junctions. 13 Perovskite/organic tandem architectures bear a realistic potential to reach an efficiency above 31%.
In general, multi-junction solar cells are designed as a series connection of wide-bandgap and narrow-bandgap sub-cells with complementary absorption spectra. An improved overlap with the solar spectrum and reduced thermalization losses are the keys to overcome the Shockley-Queisser efficiency limit of single-junctions. 14 Hybrid metal-halide perovskites have received tremendous attention as photo-active materials in solar cells. 15 Their typical ABX3 crystal structure comprises methylammonium (MA + ), formamidinium (FA + ), or Cs + ions on the A-site, Pb 2+ ions on the B-site, and halide ions, such as Ior Br -, on the X-site.
Some members of this family, e.g. FAxCs1-xPb(IyBr1-y)3, afford tunability of the bandgap energy (Eg) between 1.5-2.3 eV, mainly by variation of the I/Br ratio, 16,17 rendering them especially attractive for the design of multijunction cells.
All-perovskite tandem cells require narrow-gap materials (Eg < 1.3 eV), where the lead is partially replaced by tin. 18
The EQE spectrum of these devices extends beyond 900 nm (Figure 1b, Figure S1).
Adding a certain concentration of fullerene molecules into the PM6:Y6 photo-active layer to form a so-called ternary system, i.e. PM6:Y6:PC60BM (1:1.2:0.2), improves the blend morphology, which results in enhanced charge transport and reduced non-radiative recombination. 27,28 Thus, a notable boost in cell characteristics is achieved with a PCE up to 17.5% (for device statistics see Figure S2). We want to highlight a notably enhanced EQE of > 85% for the ternary cells in the range of  > 650 nm, which is the spectral region of operation when combined with a widegap front-cell in a tandem. The absorption spectra of PM6 and Y6 (Figure 1c) show that for  > 650 nm excitons are predominantly generated on the Y6, which will be shown to be the key that unlocks outstanding device stability.
Stability of NFA cells under continuous operation is still a serious concern and a subject of vigorous scientific research. 29 To assess the stability of our binary and  Figure S3). This degradation motif has previously been explained by a photoinduced reorganization in the donor/acceptor blend and the formation of microscopic aggregates of NFA molecules, which leads to a reduced electron mobility and enhanced recombination. 30 Owing to their improved blend morphology, the decay of the ternary cells is notably slower than that of the binary cells. 29 However, most strikingly, under NIR illumination, where excitons are solely generated on the Y6, the devices did not show any burn-in and no degradation even under long-term continuous operation for more than 1,000 h. These findings indicate that the detrimental morphological changes, discussed above, would require excitation of the donor polymer PM6 and that they can be substantially mitigated if predominantly the Y6 NFA is excited. On the other hand, we found that upon continuous illumination with the white LED under inert conditions the photoluminescence quantum yield of PM6 shows a notable degradation, while the Y6 is less affected ( Figure S4). Therefore, photo-induced degradation of PM6 could likewise play a significant role in addition to possible morphological changes. In any event, our findings contradict the paradigmatic association of non-fullerene solar cells with operational instability, and they present the especially encouraging prospect that the long-term operational stability of perovskite/organic tandem cells will not be limited by the narrow-gap OSC. This is in notable contrast to allperovskite tandems, where the stability issues of Sn-based narrow-gap PSCs are still a very serious issue.
The PM6:Y6 organic system provides an energy-gap of 1.33 eV, 26

Wide-gap perovskite sub-cell
We selected FA0.8Cs0.2Pb(I0.5Br0.5)3 as suitable perovskite composition with a bandgap of 1.85 eV ( Figure S6). Early studies have shown, that for PSCs with Eg > 1.75 eV, the Voc did not concomitantly increase with Eg, which has frequently been attributed to photo-induced halide-segregation in the perovskite into bromine-and iodine-rich domains. 31 Yet, more recently, recombination losses at the interfaces of the wide-gap perovskite and the adjacent charge extraction layers have been found to be predominately limiting the VOC. 8,32 As such, we aimed to minimize these interfacial losses in order to narrow in on the highest possible Voc in our PSCs.
In However, PTAA comes with serious limitations due to the tradeoff between efficient hole transport, which requires the PTAA to be as thin as possible, and selectivity, that has been found to impose a lower limit to the PTAA thickness. 33 To overcome the issues associated with PTAA, we use [2-(3,6-Dimethoxy-9H-carbazol-9yl)ethyl]phosphonic Acid (MeO-2PACz) 13 as HEL, that forms a dense, pinhole-free self-assembled monolayer on the ITO bottom electrode. In a direct comparison perovskite layers on MeO-2PACz as HEL afford a 90 meV larger QFLS compared to their analogues on PTAA (Figure 2a). Interestingly, in stark contrast to PTAA, with MeO-2PACz as HEL, we also did not observe any notable halide segregation under a one sun equivalent illumination on a timescale of several minutes, despite using a Br:I ratio of 0.5:0.5 ( Figure S7). This is rather unexpected, as literature suggests notable halide segregation for Br concentrations in this range. 34 Our findings demonstrate that the proper choice of HEL allows to mitigate halide segregation in perovskites even with elevated Br-concentrations.
Before studying the impact of the electron extraction layer (EEL), we implemented passivation strategies, such as the addition of excess lead 35 or the modification of the perovskite surface by the organic halide salt phenethylammonium iodide (PEAI), 36 which promotes the formation of a 2-dimensional perovskite capping layer (see Figure S8 -S10). As evidenced by photoelectron spectroscopy, the insertion of PEA + leads to some notable lowering of the perovskite valence band maximum ( Figure S11), which prevents photo-generated holes from reaching the EEL and thereby improves the selective extraction of electrons.
In the absence of an EEL the surface passivation does not affect the QFLS (Figure 2a), which indicates that defects at the surface or grain boundaries do not impose a limit in this scenario. The situation changes if we complete our p-i-n PSCs by adding PC60BM and Al doped ZnO nanoparticles (AZO NP) as EEL (Figure 2b). 37 Note, for the integration of the PSCs in the tandem cells, an additional SnOx layer, grown by low temperature atomic layer deposition (ALD), is used that serves as permeation barrier, which not only improves the long term stability but also protects the layers underneath against chemical attack by the solvents of subsequent wet chemical processes. 38,39 In striking contrast to PTAA, with MeO-2PACz as HEL, the addition of the EEL infers  The respective references can be found in Table S1. f, normalized PCE vs. time of the PSCs illuminated with a white LED and operated in the maximum power point.

Low-loss recombination interconnect
The interconnect is a key component of monolithic ( (Figure 3b). This is due to a Schottky barrier of 0.6 eV that forms at the SnOx/MoOx interface (see Figure 3c, d and Figure S13).
To render the interconnect ohmic, thin layers ( 1 nm) of a metal (Ag or Au) are frequently inserted between top-and bottom-cell. 10 (Figure 3d,e). As the metallic InOx layer is ultra-thin, it still provides a very high sheet resistance > 10 6 /sq, which is a prerequisite for future largearea scalability as it is of critical importance to avoid shorting of the sub-cells in case of local shunt paths. 5 High carrier densities typically infer optical absorption, which is another important reason to keep the thickness of the interconnect to a minimum. 43 Most strikingly, our ultra-thin InOx interconnect with a transmittance near unity does not introduce notable optical losses (Figure 3g), which boosts the EQE of the organic back-cell and the overall Jsc of the tandem by about 1.5 mA/cm 2 compared to the case of an interconnect based on 1 nm of Ag, which affords a PCE of only about 20% ( Figure   3 h, Figure S20). ALD allows for large-area, high-throughput processing (even at atmospheric pressure), 44 and enables conformal coating of textured surfaces that frequently occur in light trapping concepts. 2 Hence, we foresee that the applicability of our interconnect is not limited to perovskite/organic tandem cells but it may also be favorably used in other tandem cells. Note, in the first 2 nm of the molybdenum oxide layer we found a mix of oxidation states of the molybdenum ranging from Mo 2+ to Mo 6+ (details in Figure S15-S19).
The first 2 nm of the molybdenum oxide are characterized by the presence of h, resulting EQE spectra of the organic back-cell with InOx or Ag as interconnect demonstrating the significant current losses induced by only 1 nm Ag.

Monolithic perovskite-organic tandem cells
Drawing from the significant progress outlined above, we prepared monolithic   designed, conducted and evaluated the PLQY/QFLS studies. AA and SA provided the expertise in the processing of the self-assembled monolayers. DH and KM contributed temperature dependent J/V characterization. LM, AH and FS designed and conducted the GIWAXS studies. All authors discussed the results and were involved in the writing.

Supplementary Information available
Competing financial interests: The authors declare no competing financial interests.