Scalable two-terminal all-perovskite tandem solar modules with a 19.1% efficiency

Monolithic all-perovskite tandem photovoltaics promise to combine low-cost and high-efficiency solar energy harvesting with the advantages of all-thin-film technologies. To date, laboratory-scale all-perovskite tandem solar cells have only been fabricated using non-scalable fabrication techniques. In response, this work reports on laser-scribed all-perovskite tandem modules processed exclusively with scalable fabrication methods (blade coating and vacuum deposition), demonstrating power conversion efficiencies up to 19.1% (aperture area, 12.25 cm2; geometric fill factor, 94.7%) and stable power output. Compared to the performance of our spin-coated reference tandem solar cells (efficiency, 23.5%; area, 0.1 cm2), our prototypes demonstrate substantial advances in the technological readiness of all-perovskite tandem photovoltaics. By means of electroluminescence imaging and laser-beam-induced current mapping, we demonstrate the homogeneous current collection in both subcells over the entire module area, which explains low losses (<5%rel) in open-circuit voltage and fill factor for our scalable modules. All-perovskite tandem photovoltaics hold technological potential yet their upscaling is not trivial. Here Nejand et al. fabricate mini-modules using scalable methods and laser-scribed interconnections, achieving a 19.1% efficiency over an aperture area of 12.25 cm2.

While laser-scribed interconnection schemes of single-junction perovskite solar modules were successfully implemented in the past at very high geometrical fill factors of >98% (ref. 29 ), there is yet no report addressing the very specific challenges of the monolithic all-perovskite tandem module. Similar to single-junction perovskite solar modules, scaling 2TPT-SCs will adapt the established thin-film module interconnection concept of elongated PSC stripes that are separated by equidistant interconnections formed by three scribing lines 16,[30][31][32][33][34][35][36] . However, realizing laser-scribed interconnection lines in 2TPT-SMs necessitates processing in an inert atmosphere and a scribing process that is compatible with a tandem solar cell layer stack containing two types of perovskite thin films.
Next to the module interconnection, 2TPT-SM device architectures processed with industrially scalable deposition methods need to be developed. So far, all 2TPT-SCs have required several spin-coating steps, which limited the device area to a laboratory scale of up to around 12 cm 2 (ref. 25 ). For large areas, scalable methods such as thermal evaporation 37 , blade coating 35,38,39 , spray coating 40 and slot-die coating 41,42 have been investigated for perovskite photovoltaics. Among them, blade coating (PCE: 17.8%) 39 and slot-die coating (PCE: 20.8%) 42 have demonstrated remarkable potential in processing single-junction perovskite solar modules. The key challenge in processing 2TPT-SMs with scalable solution-based methods is the sequential layer deposition considering the similar solubility of the WBG and NBG perovskite layer in the top and bottom cell, respectively. The latter poses a major hurdle to the sequential deposition of the 2TPT device architecture as it induces facile degradation of the underlying layer (for example, the WBG perovskite) when processing the solution of the NBG bottom solar cell.
In response to the above-mentioned challenges in upscaling 2TPT PV, this study reports on laser-scribed 2TPT-SMs processed exclusively with scalable fabrication techniques. The study introduces the scientific and technological advances that led to these prototypes. First, we introduce our reference device architecture, which benefits from reduced optical losses and shows close performance in small-scale 2TPT-SCs (aperture area of 0.1 cm 2 ) compared to the record devices reported in the literature 26 . Second, we report on a high-throughput laser-scribing process in an inert atmosphere and demonstrate high geometric fill factors (GFF) in small-scale spin-coated 2TPT-SMs (aperture area of 2.25 cm 2 , referred to as 'spin-coated 2TPT-SMs'). Using time-of-flight secondary ion mass spectrometry (ToF-SIMS), we reveal the highly selective laser patterning of the scribing lines. Finally, we demonstrate The disadvantage of the percolated Au layers is a reduced transmission compared to ITO recombination layers ( Supplementary  Fig. 10). According to our observations, even the conductivity of thin ITO layers of thickness ~15 nm (~450-500 Ω sq −1 sheet resistance) is sufficient to induce notable shunt resistance losses. For this reason, a percolated Au recombination layer is employed here despite the high parasitic absorption losses. These absorption losses are attributed to plasmonic losses of percolated Au layers in the recombination junction (SnO x /Au/PEDOT:PSS), even at the lower nominal thicknesses of Au down to 0.3 nm (Supplementary Note 3 and Supplementary Figs. [6][7][8][9][10][11][12]. However, neither a percolated Au recombination layer nor a thin ITO recombination layer (~15 nm) are optimal and substantial future improvement potential lies in the development of improved recombination or tunnel junctions for all-perovskite tandem PVs. Figure 1e shows an all-laser-scribed spin-coated 2TPT-SM comprising four serially interconnected cell stripes with a total aperture area of 2.56 cm 2 and a 94.7% GFF (Fig. 1d,e and Supplementary  Fig. 21a,b). It should be noted that the low stability of mixed Sn/ Pb-based perovskite narrow-bandgap bottom solar cells poses a general challenge to the industrial upscaling of all-perovskite-based tandem photovoltaics. Exposure of the devices to the ambient atmosphere needs to be avoided at any stage. For this purpose, a custom-built laser-scribing set-up is employed for the fabrication of the monolithic interconnections of 2TPT-SMs; the set-up allows patterning in an inert atmosphere to prevent O 2 exposure and degradation of the NBG bottom subcell during the scribing process.
Compared to other thin-film module technologies (GFF up to 96% (ref. 44 )), as well as single-junction perovskite modules (GFF ~95% (refs. 16,[30][31][32][33][34][35][36]45 )), the achieved GFF of 94.7% is very respectable. For details on the laser-scribing process, we refer to the experimental section and Supplementary Fig. 21. As shown in Fig. 1e, the V oc of each cell stripe of the spin-coated tandem module corresponds to around 2.01 V, while the total V oc of the module equates to the expected 8.0 V (Fig. 1e). This implies a minimal loss in V oc of only 10 mV per cell stripe (0.5% relative). Moreover, our modules retain a high FF of 75%, demonstrating a negligible detrimental effect of introduced module interconnections (compared with single-junction solar cells showing FF of ~75%; Fig. 1b). Our in-house measurement of the champion module (without encapsulation and masking, measured in ambient atmosphere) with an aperture area of 2.56 cm 2 (defined by laser scribing) exhibits a high PCE of 22.2% (23.7% active area PCE) with V oc = 8.0 V, short-circuit current I sc = 9.57 mA and FF = 75% (Fig. 1e). The MPP tracking of the modules under continuous illumination (1 Sun) shows a SPCE of 21.4% (and stable power output of 54.7 mW) for 15 hours (Fig. 1f). Moreover, the module shows stable power output at the maximum power point for 3.5 hours and 85°C in an inert atmosphere (Fig. 1g).
The good performance of the spin-coated 2TPT-SM was certified by the accredited CalLab PV Cells of Fraunhofer ISE. Given the sensitivity of the NBG perovskites towards ambient atmosphere and the requirements of the certification standards, the module was encapsulated and covered with an additional external mask. The encapsulated and masked champion module (aperture area of 2.43 cm 2 ) exhibited an in-house PCE of 19.8% (I-V measurement, backward scan) and an externally measured PCE of up to 18.3% (Fraunhofer ISE CalLab PV Cells). The certified PCE determined by steady-state MPP is 17.99% ± 0.63% (Fraunhofer ISE CalLab PV Cells, see Supplementary Fig. 22). The apparent difference between the certified and in-house PCE is attributed to degradation losses that originate from the encapsulation process (which is conducted in ambient atmosphere) and the extended time of transfer (see Supplementary Fig. 22, Supplementary Table 2 and Supplementary Note 6 for a detailed discussion). Therefore, to reach higher stability for the 2TPT-SMs, proper encapsulation under an inert atmosphere is required. Further advances in the stability of all-perovskite tandem photovoltaics will be inspired by the research and development of additives and passivation layers, such as the latest work of Lin et al., which demonstrated high stability by grain-boundary passivation 26 .
Using ToF-SIMS, we provide a three-dimensional (3D) depth profile of the material composition at the 2TPT-SMs interconnection, highlighting the selective laser-ablated material removal at the P1, P2 and P3 interconnection line (Fig. 2a-c). By overlaying the depth profiles of all materials ( Supplementary Fig. 23) and the underlying planar glass layer, the 3D depth profile (Fig. 2c) illustrates the material composition at the interconnection lines and the selective laser-ablated material removal. The dimension of the P1-P3 scribing investigated here by ToF-SIMS exhibits a very similar width (dead area) compared to the previous analysis via scanning electron microscopy (SEM), of 220 µm (Fig. 2b). Considering the dead area width of 210 µm, which can be estimated by ToF-SIMS, we determine a GFF of 94.7% (±0.1%).
Having investigated the selective laser-ablated material removal by means of 3D profiles and SEM surface scans, we assess the homogeneity of the tandem module with regard to: (1) photocurrent generation in the subcells over the entire module area using LBIC mapping; and (2) individual defects and shunts in the top and bottom subcells of the module by electroluminescence imaging.
To study the homogeneity of photocurrent generation in each subcell of our tandem module, we employed an adaptable configuration  of LBIC mapping ( Fig. 2d-f). The layer stack of modules for the LBIC study is identical to the tandem modules as shown in Fig. 1a (employing Au as a recombination layer). Conventional LBIC is an established method to investigate the homogeneity of perovskite layers in single-junction solar cells and modules 46-50 . To discriminate between the photogenerated current in the top and bottom subcells, the LBIC set-up is expanded by a second excitation laser, and now comprises laser sources of wavelength 530 nm and 850 nm, respectively. First, exciting the module with the 530 nm laser, we track the LBIC signal in the top subcell, as photons of this wavelength are fully absorbed in the WBG perovskite absorber layer (Fig. 2d,e). The corresponding LBIC map shows uniform photocurrent generation in the entire active area of the module. Second, using the 850 nm laser, we selectively excited the bottom subcell, as photons of this energy are not absorbed by the WBG perovskite top subcell, but only by the NBG perovskite bottom subcell (Fig. 2d,f). Similar to the LBIC map of the top subcell, the LBIC signal of the bottom subcell also exhibits a homogeneous signal over the entire module area. Interestingly, some defects are apparent as dark spots in the LBIC signal of the top subcell but appear bright in the LBIC signal of the bottom subcell. These defects are attributed to pinholes in the top subcell and inhomogeneity in the perovskite thin-film thickness. Overall, the LBIC data prove an even photocurrent generation and extraction over the entire active area for both the top and bottom subcells of our module.
To study the distribution of local defects and confined local charge barriers in our tandem modules, we further employ electroluminescence imaging, which exhibits only indirect information on current collection but at much higher resolution (for our set-up a pixel coarsely represents about 13 µm × 13 µm) compared to LBIC (our LBIC resolution is limited by laser spot diameter; 51 ~200 µm and 500 µm in our measurements). To distinguish the electroluminescence signal of the two subcells in our tandem module (1.78 eV WBG perovskite top subcell ( Supplementary Fig. 24) and 1.26 eV NBG perovskite bottom subcell 52 ), the electroluminescence signal is imaged with: (1) a neutral density filter to track the full electroluminescence signal; (2) a longpass filter (~760 nm) for imaging the bottom subcell; and (3) a shortpass filter (~775 nm) for imaging the top subcell (   bottom subcells). We attribute these defects to either local variations in charge-carrier injection (due to different series resistance) or local variation in out coupling of luminescence from the device (for example, due to variations in surface roughness and thickness). At the same time, most features are visible in both electroluminescence images, indicating that the fabrication defects in the top subcell affect the quality of the bottom subcell too. However, some features are only present in images that display the particular defects or inhomogeneity in each subcell (Fig. 2h,i). Although electroluminescence and LBIC analyses agree in terms of homogeneity in the active area as well as laser-scribed interconnection lines, electroluminescence imaging allows us to reveal individual fabrication defects of the modules in both deposition and laser-scribing steps, facilitating process optimization towards high-efficiency 2TPT-SMs (Fig. 2e,f,h,i). Overall, we demonstrate in this section the ability to scale up the spin-coated 2TPT-SCs to modules with only a 6% rel drop in PCE, which is in very good agreement with the GFF loss associated with the module interconnection.

Upscaling all-perovskite tandem solar modules
Having developed an interconnection scheme with high GFF, next we present a scalable 2TPT-SM architecture that enabled a prototype module (aperture area of 12.25 cm 2 ) fabricated with scalable thin-film deposition methods (Fig. 3a,b). The scalable modules employ the same architecture as the spin-coated modules (see also Supplementary Note 7). The fabrication of the modules is conducted with a combination of blade-coating (2PACz, WBG perovskite, PEDOT:PSS, NBG perovskite and PCBM) and vacuum-deposition (LiF, C 60 , SnO x , Au, BCP and MgF 2 ) methods (Fig. 3a).
While several of these deposition steps are adapted from the literature (Methods), the processing of the NBG perovskite bottom solar cell required a major innovation to address the challenge of processing the NBG perovskite layer on top of the WBG perovskite top cell without dissolving the underlying layer (see the next section). Prototype modules reached a PCE of 19.1% (aperture area of 12.25 cm 2 , seven cell strips and 94.7% GFF; Fig. 3c) and 18.3% MPP tracking (corresponds to 224 mW) for 20 hours with less than 7% rel drop (Fig. 3d). As shown in Fig. 3c and Table 1, in the current blade-coated module, the single 2T tandem cell strips with 1.75 cm 2 active area on the module show almost the same FF, V oc and I sc (with an average PCE of 19.4%), which results in a 19.1% PCE for modules. Compared to our spin-coated modules with 2.56 cm 2 aperture area, the 2TPT modules processed with scalable deposition methods exhibit only a loss of 4% in FF and 100 mV in voltage, which we attribute to inhomogeneity and variations in perovskite thin-film morphology from the blade-coating step ( Fig. 3c and Table 1).
To visualize the defects that we associate with the upscaling losses, we conducted photoluminescence imaging as well as electroluminescence imaging, as introduced previously. As shown in Fig. 3e-g, using shortpass (775 nm) and longpass (760 nm) filters, the defects are apparent in the WBG and NBG solar cell and consequently are expected to impact the FF and V oc . As illustrated in the electroluminescence (EL) and photoluminescence imaging, electroluminescence imaging can visualize the defects in more detail compared to photoluminescence imaging due to its high sensitivity to thickness variation, shunts and out coupling ( Fig. 3e-g). Position-dependent variations in the thickness of both WBG and NBG perovskites are clearly visualized by EL imaging. Interestingly, the 2T module shows a remarkable electroluminescence signal at a starting voltage of ~12 V (Fig. 3f) with emission from the WBG (688 nm) and NGB (977 nm) perovskite layer, which itself highlights the high quality and homogeneity of our 2TPT modules (Fig. 3g). To investigate the photocurrent generation in the subcells over the entire module, LBIC measurement on a different module is conducted (Supplementary Fig. 29). The black arrows indicate defects and/or inhomogeneities in the bottom and top subcells.

Non-destructive scalable processing of tandem architecture
The key challenge in processing 2TPT modules with scalable solution-based methods is the very similar solubility of the WBG and NBG perovskite layers in the top and bottom cells, respectively 6 . To prevent degradation of the WBG perovskite top solar cell during the subsequent deposition step of the NBG perovskite layer, in addition to the use of robust interlayers, the resting time of the NBG perovskite solution on the sample needs to be minimized. As we show in Fig. 4a and Supplementary Fig. 30, notable corrosion of the WBG subcell, due to penetration of NBG solution through the interlayers, is already apparent after 6 s. For spin-coated devices, the resting time is <5 s (Supplementary Fig. 31), as most of the solution already evaporates during the early phase of the spin rotation, leading to high-performance devices (Fig. 1b). For blade-coated NBG thin films, a simple VAGC step already helps to accelerate the solvent extraction (PV0), but crystallization of the blade-coated NBG perovskite still takes longer (~18 s) than small-size spin-coated samples due to the high amount of DMF:DMSO vapour and thicker perovskite solution (Fig. 4a,c and Supplementary Fig. 31). This slightly longer resting time of the NBG solution causes penetration of solvents through the SnO x barrier layer and consequently degradation of the WBG perovskite in the top subcell ( Fig. 4a and Supplementary Fig. 30).
To further accelerate the extraction of solvents, a moderate nitrogen flow (260 cm 3 min −1 , PV2) is applied in addition to the vacuum extraction step (Fig. 4b,c), leading to overall faster mass transport of solvents away from the sample surface 53 . As a result, the resting time of the NBG drops to <5 s, which prevents degradation of the underlying WBG top solar cell (Fig. 4a). Importantly, this nitrogen gas flow-assisted crystallization of the NBG perovskite on the top subcell produces a microscopic, defect-free morphology (PV2; Fig. 4c,f), while in the absence of nitrogen gas flow (PV0; Fig. 4c,d) or insufficient nitrogen gas flow (PV1; Fig. 4c,e) a large number of defects appear that lead to degradation of the WBG perovskite top solar cell by the solvents of the NBG perovskite solution.
The importance of nitrogen gas flow in the VAGC of NBG perovskite thin films is further highlighted by comparing the performance of isolated small-area 2TPT-SCs (aperture area: 0.1 cm 2 ) processed from a large-area substrate (>30 cm 2 ) in the same architecture as the scalable 2TPT-SMs ( Supplementary Fig. 32). As shown in Supplementary Fig. 32, 2TPT-SC with NBG perovskite thin films processed only by VAGC show considerably lower overall performance compared to VAGC with gas flow. Furthermore, the different PCE of the devices at different locations on the large-area substrate is attributed to the defects and variation in the thickness of the blade-coated layers at different locations. Therefore, we show that scaling up the 2TPT-SM architecture by scalable deposition methods is feasible as long as very fast extraction of solvents of the bottom subcell is guaranteed. In this work, we employ a combination of VAGC with nitrogen gas flow to realize fast drying of the solution-processed NBG perovskite thin film in a mild vacuum range of 10 4 -10 5 Pa. Similar to gas quenching, the technique realizes a fast mass transport of evaporated solvents away from the surface of the wet film 54 . In view of future in-line large-area production, gas quenching, as well as combinations of vacuum quenching with gas flow, would need to be realized shortly after the slot-or blade-coating step to ensure sufficiently fast drying.
Overall, this work highlights the facile upscaling route of the 2TPT-SCs from 0.1 cm 2 up to 12.25 cm 2 in 2TPT-SMs. Having employed only scalable deposition techniques (blade coating and vacuum deposition) for all 12 layers of the fabricated 2TPT-SMs, the feasibility of upscaling this technology is proven. The demonstrated GFF of our tandem module is comparable to other thin-film PV technologies, such as tandem thin-film silicon photovoltaics (~98% GFF) 29 and CIGS photovoltaics (~93% GFF) 55 . It is noteworthy that the laser set-up applied in this work uses a rather conventional and inexpensive nanosecond laser that is less complex than the widely used picosecond or femtosecond lasers for thin-film solar module patterning in laboratory-scale modules. The technological bottleneck of the current development remains the limited stability of the 2TPT-SMs 18,24 . Fortunately, given recent advances in engineering the composition 7,20,24 as well as defect passivation of NBG solar cells 21,26 , there is encouraging progress in improvement in the stability of the NBG bottom subcell.

Conclusions
This work advances two-terminal all-perovskite tandem photovoltaics by introducing a scalable high-efficiency, two-terminal, all-perovskite tandem module. The upscaling of laboratory-scale,  Atomic layer deposition of the SnO x . SnO 2 layers were fabricated by atomic layer deposition (ALD) in a Picosun ALD system (Picosun R200) at 90 °C using the precursors tetrakis(dimethylamino)tin(IV) (TDMASn) (99.99%-Sn, Strem Chemicals) and water. High-purity argon (Ar, 99.999%) was used as carrier gas and purge gas. The line flows of TDMASn and water were set to 120 and 150 standard cubic centimetres per minute (sccm), respectively. The TDMASn source container was preheated for 1 hour at 70 °C, to ensure thermal equilibrium, before the deposition and the water precursor was unheated and pulsed at room temperature. A 35 nm thick SnO x film was deposited by 300 cycles using these settings: TDMASn (pulse time 1.6 s, purge time 12 s) and water (pulse time 0.1 s, purge time 16 s). Due to the machine layout, a boosting system was utilized instead of the conventional bubbler. In this system, the line flow was increased to 500 sccm immediately before pulsing, to facilitate an increased precursor concentration. The increased time of this process is the cause of the abnormally high TDMASn pulse time (normal pulse times are <0.5 s).

Fabrication of the control two-terminal all-perovskite tandem solar cell with Arch. 2. Two-terminal all-perovskite tandem solar cells with Arch. 2 architecture
were fabricated in the p-i-n architecture. Glass substrates coated with a 140 nm thick ITO front electrode (sheet resistance 15 Ω sq −1 , Luminescence Technology) were used. The substrates were cleaned consecutively in deionized water, acetone and isopropanol for 10 min each in an ultrasonic bath. Immediately before deposition of the HTL, substrates were treated using oxygen plasma for 3 min. As HTL, poly(bis(4-phenyl)(2,4,6-trimethylphenyl)amine (PTAA, 2 mg ml −1 in toluene) was used. For deposition of the PTAA, 60 µl of the PTAA solution in toluene was deposited onto the substrate (16 mm by 16 mm) by spin coating at 6,000 r.p.m. for 30 s, followed by annealing at 100 °C for 10 min. Subsequently, the samples were transferred to a glovebox for deposition of the perovskite thin film. A 15 nm thick C 60 thin film was deposited by thermal evaporation (OPTIvap, CreaPhys) as an ETL. Subsequently, a 35 nm thick SnO x thin film was deposited by ALD. As a recombination layer, different thicknesses of Au, of 0.3, 0.6, 1.2 and 1.8 nm, were deposited on the SnO x layer by thermal evaporation with a low rate of 0.1 Å s −1 to reach the maximum possible uniformity of Au on the substrates. After filtering the aqueous PEDOT:PSS dispersion (Ossila, AI 4083) with a 0.45 µm polyvinylidene fluoride (PVDF) filter, the solution was deposited on the substrates directly by spin coating at 6,000 r.p.m. for 30 s dynamically to avoid degradation of the sublayers. Next, the substrates were annealed on a hotplate at 120°C for 20 min. Afterward, fullerene (C 60 , ~25 nm) and BCP (~7 nm) were thermally evaporated as the electron transport material, followed by the deposition of a Cu (~150 nm) or Au (~60 nm) thin film as the rear electrode. The active area of the two-terminal all-perovskite tandem solar cells was 0.1 cm 2 .

Fabrication of the two-terminal all-perovskite tandem solar cell with modified architecture (Arch. 1).
Similarly to the control sample with Arch. 2, the samples with modified architectures of two-terminal all-perovskite tandem solar cells (Arch. 1) were fabricated in the p-i-n architecture. Glass substrates coated with a 230 nm thick IO:H front electrode were used 56 . Cleaning the substrates was conducted in the same way as for the control samples. For HTL, 2PACz (0.375 mg ml −1 in ethanol) was used. For deposition of the 2PACz thin film, 70 µl of the 2PACz solution in ethanol was deposited on the substrate by spin coating at 3,000 r.p.m. for 30 s followed by annealing at 100 °C for 10 min. After deposition of the WBG perovskite, a 1 nm thick LiF thin film was deposited on the WBG perovskite. As for the control samples, a 15 nm thick C 60 thin film was deposited by thermal evaporation as an ETL and, subsequently, a 35 nm thick SnO x thin film was deposited by ALD. As a recombination layer, a 15 nm thick ITO thin film was deposited on SnO x via d.c. sputtering. In the modified architecture, after filtering the aqueous PEDOT:PSS with 0.45 µm PVDF filter, the PEDOT:PSS was diluted with ethanol or isopropanol with a 3:1 ratio. The diluted solution was deposited on the substrates with the same spin-coating and annealing parameters as the control samples. After deposition of the NBG perovskite thin film using the VAGC method, a thin film (~5 nm) of PCBM (Sigma-Aldrich, 99.5%) was deposited onto the perovskite thin film using a 5 mg ml −1 solution in 1,2-dichlorobenzene (Sigma-Aldrich, anhydrous, 99%) in the modified architecture. After deposition of the C 60 , BCP and rear contact layers, as done for the control samples, finally, a 125 nm thick MgF 2 thin film was thermally evaporated on the front side of the glass. The active area of the tandem solar cells was the same as for the control samples.
Fabrication of the spin-coated two-terminal all-perovskite tandem solar modules with 2.56 cm 2 aperture area. For fabrication of the primitive modules with the spin-coating method and an aperture area of 2.56 cm 2 , 30 mm by 30 mm IO:H substrates were used (to have extra space for glass-glass encapsulation). We followed all procedures described for the fabrication of small-area (0.105 cm 2 ) 2T tandems with our modified architecture, except for using a higher amount of solution to reach uniform coverage. Furthermore, the modules use a percolated Au as a recombination junction (nominal thickness 1.2 nm). While such percolated Au layers introduce some optical losses, they exhibit a very low lateral conductivity. The architecture of the spin-coated 2TPT-SMs was the same as shown in Fig. 1a, but with a SnO x /percolated Au recombination junction (see also Supplementary  Fig. 21 and Supplementary Note 7). The laser-scribing procedure described in the next section was used in the fabrication of our modules. The final modules were encapsulated with butyl rubber at ambient atmosphere for further LBIC and electroluminescence characterizations in the atmosphere.
Laser scribing of module interconnection lines. Fabrication of monolithically interconnected two-terminal modules on substrates of 30 mm by 30 mm was facilitated by integrating three scribing lines (P1, P2, P3) in our device stack. On the basis of the modified architecture, P1 was employed to ablate the IO:H, P2 for ablating the multilayers of 2PACz/WBG perovskite/LiF/C 60 /SnO x /Au/PEDOT:PSS/ NBG perovskite/PCBM/C 60 /BCP and P3 to ablate the rear metal contact. The widths of P1, P2 and P3 were 60, 60 and 40 µm, respectively ( Supplementary Fig. 21). The entire scribing width encompassing buffer regions was 240 µm. A module layout with a 2.56 cm 2 aperture area was used. A custom-built laser-scribing set-up developed in conjunction with Bergfeld Lasertech was employed, integrating a 1 ns Nd:YVO 4 laser (Picolo AOT 10-MOPA, InnoLas Laser) with available 1,064 nm and frequency-doubled 532 nm wavelengths, variable pulse repetition rates and power output. All three interconnection lines (P1, P2 and P3) were laser scribed using the same ns laser (wavelength of 532 nm). For alignment and laser spot motion, an X-Y directional scanner system combined with a camera was utilized. All scribing processes were performed from the film side and followed a predefined scribing layout. The scribing part of the system was incorporated in a nitrogen-filled glovebox. P1 laser scribing was performed at a pulse repetition rate of 10 kHz, 50 mm s −1 scribing speed and 2.00 J cm −2 fluence. P2 laser scribing was performed at a pulse repetition rate of 10 kHz, 33 mm s −1 scribing speed and 0.45 J cm −2 fluence. P3 laser scribing was performed at a pulse repetition rate of 10 kHz, 100 mm s −1 scribing speed and 0.40 J cm −2 fluence.
Upscaling the two-terminal all-perovskite tandem solar modules with a 12.25 cm 2 aperture area. For the fabrication of two-terminal all-perovskite tandem solar modules with 12.25 cm 2 aperture area, only fully scalable methods were used in the deposition of the entire layer. For this purpose, a combination of vacuum deposition and blade coating (blade gap of 100 μm) was employed. Substrates of 6.4 × 5 cm 2 IO:H were used such that the inhomogeneous region in the blade-coating procedure (7 mm from the beginning and 7 mm from the end of the blade-coated substrates) can be cut away. The final substrate size for our modules was 5 × 5 cm 2 , in which a 0.5 mm distance from each edge was reserved for the glass-glass encapsulation. As shown in Fig. 3b, the modules have an aperture area of 12.25 cm 2 (3.5 cm × 3.5 cm), which contains seven cell strips (with an active area of 1.75 cm 2 on each). The fabrication of two-terminal tandems was conducted as follows.
(1) After P1, deposition of the 2PACz was conducted by blade coating with a rate of 15 mm s −1 followed by drying in a nitrogen atmosphere and annealing at 100 °C for 10 minutes. (2) Supplementary  Fig. 25. The electroluminescence spectra of the cells were acquired using an Ocean Optics HDX spectrometer. All measurements were performed in ambient air on encapsulated samples.
Photoluminescence measurement and imaging. The photoluminescence images were acquired with a 2.1 megapixel scientific CMOS camera (Quantalux sCMOS camera, Thorlabs). The photoluminescence was filtered with two different optical filters mounted on a stationary wheel: a 775 nm shortpass (Edmund Optics) and a 760 nm longpass (Edmund Optics). The modules were excited using a bright blue LED with a 470 nm emission peak wavelength. A 665 nm longpass filter was used to cut out the blue LED interference. The set-up is shown in Supplementary Fig.  25. The photoluminescence signal was collected with a f = 5 cm lens and filtered with a 600 nm longpass. The excitation wavelength was 513 nm at a repetition rate of 20 kHz with a pulse width of approximately 260 fs. The beam area was 0.787 mm 2 and the power was 7.9 mW, leading to a fluence of roughly 50 μJ cm −2 . The sample (without contact layer and coated with ethyl vinyl acetate) was excited from both sides to obtain the final accumulative photoluminescence spectrum from WBG and NBG perovskites.

ToF-SIMS.
Time-of-flight secondary ion mass spectrometry was performed on a TOF.SIMS5 instrument (ION-TOF) equipped with a Bi cluster primary ion source and a reflectron-type time-of-flight analyser. For depth profiling, dual-beam analysis was performed in an interlaced mode. The sputter gun operated with cesium ions at 2 keV, scanned over a field of 500 × 500 µm 2 (target current 120 nA) and was applied to erode the sample, whereas the primary ion source was scanned over a concentric field of 350 × 350 µm 2 (256 × 256 data points). Spectra were calibrated on the omnipresent C − , C 2 − , C 4 − , O − and Br − peaks. On the basis of these datasets the chemical assignments for characteristic fragments were determined. For data visualization, secondary ion intensities were rendered to 3D depth profile images.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated or analysed during this study are included in the published article and/or its Supplementary Information. Additional original and unprocessed data used in the article, as well as in the Supplementary Information, are available from the corresponding author upon reasonable request, along with reference/ calibration measurements and background information.
Confirmation that the reference cell was calibrated and certified Yes No Provided in "Current-density-voltage (J-V) measurements" section Calculation of spectral mismatch between the reference cell and the devices under test Yes No Spectral mismatch was conducted on the the certified modules. As we measured the cells with one lamp solar simulator, in the J-V measurements we took the Jsc from the EQE.

Mask/aperture
Size of the mask/aperture used during testing Yes

No
The active area of the WBG and NBG single-junction perovskite solar cells is defined by a metal mask to 0.1 cm². The aperture area of the laser-scribed modules is defined by the laser scribing (2.56 cm2). For the certification, additional external manual masking of the module was required which decreased the aperture area down to 2.43 cm². For the measurements of the 12.25 cm² modules, a metal mask with a 12.25 cm² aperture area was used. For the latter device, no difference with and without a mask is apparent, highlighting the accuracy of the laser scribing.
Variation of the measured short-circuit current density with the mask/aperture area Yes No EQE measurement was used to approve the measured current destiny by J-V measurement. See Fig 1b and Fig. S19 7. Performance certification Identity of the independent certification laboratory that confirmed the photovoltaic performance 9. Long-term stability analysis

Type of analysis, bias conditions and environmental conditions
For instance: illumination type, temperature, atmosphere humidity, encapsulation method, preconditioning temperature

Yes No
Provided in "Current-density-voltage (J-V) measurements" section. See Fig. 1e and Fig. S29. The encapsulation described in the "Fabrication of the modules" section in the main text.