Efficient and Thermally Stable Wide Bandgap Perovskite Solar Cells by Dual‐Source Vacuum Deposition

Wide bandgap perovskites are being widely studied in view of their potential applications in tandem devices and other semitransparent photovoltaics. Vacuum deposition of perovskite thin films is advantageous as it allows the fabrication of multilayer devices, fine control over thickness and purity, and it can be upscaled to meet production needs. However, the vacuum processing of multicomponent perovskites (typically used to achieve wide bandgaps) is not straightforward, because one needs to simultaneously control several thermal sources during the deposition. Here a simplified dual‐source vacuum deposition method to obtain wide bandgap perovskite films is shown. The solar cells obtained with these materials have similar or even larger efficiency as those including multiple A‐cations, but are much more thermally stable, up to 3500 h at 85 °C for a perovskite with a bandgap of 1.64 eV. With optimized thickness, record efficiency of >19% and semitransparent devices with stabilized power output in excess of 17% are achieved.

formulations with double and triple Asite cations have been reported. [50][51][52][53] However, in most cases they are found to be thermally unstable when stressed at 85 °C unless complex compositions involving four Asite cations are used. [53] Interest ingly, and in apparent contradiction with what is observed for perovskites processed from solution, [54] vacuumdeposited pure methylammonium (MA + ) perovskites are much more stable than those based on multication formulations. Methylammo nium lead iodide (MAPbI 3 , or MAPI) deposited by coevapora tion has been found stable for thousands of hours when con tinuously stressed at 85 °C. [55,56] Pure MAbased wide bandgap perovskites, obtained by coevaporation of MAI, PbI 2 and PbBr 2 have also been reported, [57,58] along with materials obtained via sequential hybrid processes, [59] however, their thermal stability has not yet been assessed. In this work, we report a simplified process for the preparation of pure MAbased perovskite films by the simultaneous sublimation of MAI and a presynthesized mixed halide lead precursor. By changing the I/Br content in the Pb(I 1n Br n ) 2 precursor, the bandgap of the MAPb(I 1x Br x ) 3 wide bandgap perovskites can be readily tuned. Vacuum depos ited solar cells obtained with these materials have similar or even larger PCEs as those including multiple Acations, but are much more thermally stable, up to 3500 h at 85 °C for a perov skite with a bandgap of 1.64 eV. With optimized thickness, we achieved record PCE > 19% for a bandgap of 1.64 eV. We applied these wide bandgap perovskites in semitransparent (ST) device configurations, reaching stabilized power output in excess of 17%. Note that the concept of semitransparency as used in our manuscript, refers to devices that are bifacial, that is, where both electrodes are transparent. The devices have actually very low transparency in the visible part of the electromagnetic spectrum, as the thick perovskite films absorb most of the abovebandgap photons. Proofofconcept 4Ttandem devices using ST perovskite cells in combination with mainstream silicon solar cells are also presented.

Results and Discussion
In order to simplify the deposition of MAPb(I 1x Br x ) 3 to a dual source process, we prepared a mixed halide Pb(I 1n Br n ) 2 pre cursor by adding PbI 2 and PbBr 2 in stoichiometric amounts in a crucible and melting them at 350 °C in nitrogen at atmos pheric pressure. In this way, we obtained a mixture which sub limes without substantial changes in its stoichiometry during the duration of the typical deposition process (Figures S1-S3, Supporting Information). The mixture is homogeneous in appearance and contains a mixed Pb(I 1x Br x ) 2 phase, although not quantitatively as the precursors are still present after melting. [60] Similar approaches have been previously used by us and others to reduce the number of deposition sources in the vacuum processing of multication/halide perovskites. [53,61] MAPb(I 1x Br x ) 3 perovskite films were prepared by dual source deposition starting from MAI and mixtures of increasing bromide/iodide ratios ( Figure S4, Supporting Information), specifically Pb(I 0.9 Br 0.1 ) 2 , Pb(I 0.8 Br 0.2 ) 2, and Pb(I 0.7 Br 0.3 ) 2 . Previous works highlighted the difficulties associated with the sublimation control of MAI, [62,63] hence we have adopted a recently reported protocol based on a two source/two sensors system, which ensures a reliable and reproducible deposition process. [64,65] The corresponding Tauc plots obtained from the absorption spectra (Figure 1a) show the expected blueshift of the absorption cutoff when increasing the bromide contents, leading to wider bandgap perovskites with bandgaps (E g ) of 1.64, 1.68, and 1.71 eV. Photoluminescence (PL) spectra (Figure 1b) also shows the shift of the bandgap with increasing bromide content. The PL peaks for perovskites obtained from Pb(I 0.9 Br 0.1 ) 2 , Pb(I 0.8 Br 0.2 ) 2 , and Pb(I 0.7 Br 0.3 ) 2 precursors are cen tered at 755, 740, and 724 nm, respectively, in agreement with the optical bandgap estimated via the Tauc plot (1.64, 1.68 and 1.71 eV). The PL spectra obtained upon excitation with a 515 nm laser at carrier concentration equal to 1 sun illumination con sist of a single signal with no detectable lowenergy component, suggesting that halide segregation is absent in asprepared films. Even longer measurements (30 min) under continuous laser illumination do not show the appearance of low energy features ( Figure S5, Supporting Information).
The relative halide concentrations in the perovskite samples were estimated by energy dispersive Xray (EDS) spectroscopy. From the bromide/iodide ratio we calculated the bromide frac tion x in the chemical formula MAPb(I 1x Br x ) 3 , and compare it with the initial bromide content n in the mixed halide precursor Pb(I 1n Br n ) 2 ( Figure 1c). We observed a systematic reduction of the amount of bromide during the conversion to perovskite, which is expected as the lead halide precursor reacts with MAI. The final Br content is 70%-80% of the initial values, leading to perovskite with nominal compositions of MAPb(I 0.92 Br 0.08 ) 3 , MAPb(I 0.86 Br 0.14 ) 3 , and MAPb(I 0.79 Br 0.21 ) 3 , corresponding to optical bandgaps of 1.64, 1.68, and 1.71 eV, respectively. The Xray diffraction (XRD) of the three materials shows similar patterns (Figure 1d), compatible with a cubic perovskite, with an intense signal at 2θ = 14.05°, corresponding to the (002) perovskite plane. [56] Conversion to the perovskite phase seems quantitative in all cases, as only the perovskite films with 1.68 eV bandgap show a minor contribution of residual PbI 2 , denoted by the reflection at 2θ = 12.5°. The origin of this weak but appreci able signal is not clear yet, but it seems to not be detrimental to the device functioning, as also observed previously for pure MAPbI 3 . [56] The surface morphology of the MAPb(I 1x Br x ) 3 films were imaged with scanning electron microscopy (SEM, Figure S6, Supporting Information). The lowest bandgap perovskite films show a compact morphology, with randomly oriented grains with a size in the 100-200 nm range. When more bromide is added to the material, the grain size is slightly reduced and the morphology appears more disordered, with other small features formed on the perovskite surface. In general, the materials are very compact with no voids, as com monly observed for vacuum deposited perovskites. [46,47,66] The perovskite films were tested in fully vacuumprocessed solar cells in the pin configuration. Devices were built on indium tin oxide (ITO) patterned glass substrates, and a thin (3 nm) layer 2,2",7,7"Tetra(N,Ndiptolyl)amino9,9spirobi fluorene (SpiroTTB) was used as the hole transport layer (HTL), as it can be used as a standalone material without any high work function interlayers at the ITO/HTL interface. [67,68] A 450 nm thick wide bandgap perovskite was evaporated onto the SpiroTTB, and coated with carbon 60 fullerene (C 60 , 25 nm) and bathocuproine BCP (8 nm) as electron transport layers (ETLs). The thickness of the perovskite film was chosen to ensure sufficient current collection, although, with alternative vacuumdeposition methods, much thinner films with similar current collection efficiency have been reported. [69] The devices were finished with a 100 nm thick silver electrode. The solar cells were encapsulated with a 20 nm thick Al 2 O 3 film deposited by a low temperature atomic layer deposition (ALD) process. [56] Asprepared solar cells were characterized by measuring their current density versus voltage (J-V) curves under simu lated solar illumination (Figure 2a). The corresponding photo voltaic parameters, PCE, fill factor (FF), opencircuit voltage (V oc ) and short circuit current density (J sc ) are reported in Figure 2c-f and in Table 1. For solar cells with varying perov skite bandgaps, we observed the expected trend of the V oc and J sc , that is, the photovoltage increases with the bandgap while the photocurrent decreases. The J sc decreases mono tonically from 18.5 to 15.0 mA cm −2 when the bandgap is increased from 1.64 to 1.71 eV. The V oc was found to be 1.13 V for the 1.64 eV bandgap, and larger ≈1.17 V for the perovskite with 1.68 and 1.71 eV bandgap. The FF was found to be mod erate, increasing from ≈65% to 70% with increasing perovskite bandgap. Overall, we obtained PCE of ≈13.5% and 14.0% for E g of 1.64 and 1.68 eV, respectively, and of ≈12.5% for the wider bandgap material (1.71 eV). As the aim of this work was to assess the thermal stability of MAIbased vacuum deposited wide bandgap perovskites, we stressed the devices at 85 °C on a hot plate in ambient light and measured their performance periodically. The J-V curves under illumination and the corre sponding parameters measured after 1 h at 85 °C are reported in Figure 2b,c-f, respectively. In general, we found a substan tial improvement of the device performance, which is mainly driven by a better rectification, as the FF increases to 76%, 77%, and 74%, on average, for perovskite solar cells with bandgaps of 1.64, 1.68, and 1.71 eV, respectively. We also observed an increase in the V oc for the 1.64 and 1.71 eV bandgap materials, up to 1.16 and 1.21 V, respectively. For all bandgaps, the photocurrent was also found to increase by 1-2 mA cm −2 . As a consequence, PCEs for the solar cells with 1.64, 1.68, and 1.71 eV bandgaps were found to reach average values of 17.7%, 16.7%, and 15.1%, respectively, after 1 h thermal stressing at 85 °C. The origin of the efficiency improvement is not clear yet. In a control experi ment we performed annealing of the sample immediately after cosublimation of the perovskite, and found no improvement in the device performance ( Figure S7, Supporting Information). This indicates that most likely the performance enhancement is related with interface effects at the perovskite/ETL and/or ETL/Ag interface.
In order to assess the quality of the widebandgap MAPb(I 1x Br x ) 3   similar to previously reported mixed cation/mixed halide vacuum deposited wide bandgap perovskites. [52] The bandgaps estimated from the derivative of the EQE spectra ( Figure S8, Supporting Information) agree well with those estimated with Tauc plots in Figure 1a (within a maximum deviation of 10 meV). In order to compare the limitations of solar cells based on perovskites with different bandgaps with each other, we calculated the ratio between the measured performance indicators as obtained from the J-V curves and their maximal obtainable values in the radiative limit (Figure 3b). The radi ative limit of the V oc was calculated via the EQE response, [70] while the FF and J sc were obtained from detailed balance calculations for the specific material bandgap. [71] The V oc loss was found to be rather independent of the bandgap, at 84%-86% of the theoretical maximum. One can notice a larger FF loss (80%) for the wider bandgap material, and a clear increase of the photocurrent losses when increasing the bandgap, as the J sc decreases from 83% to 76% of the radiative limit. This results in larger PCE losses for the wider bandgap perovskite solar cells, standing at 52% of the maximum achievable PCE. For the lowest bandgap, the PCE is at ≈60% of its corresponding theoretical maximum. These differences might originate from variations in the charge diffusion length when increasing the bromide content in the MAPb(I 1x Br x ) 3 perovskite.
In addition, we measured the absolute intensity PL spectra of the three MAPb(I 1x Br x ) 3 perovskites film deposited on  glass ( Figure 3c). The PL peak centers were calculated via a bi Gaussian fit of the spectra, which corresponds to the optical bandgap of the perovskites. The PL spectrum of the perovskite with 1.68 eV bandgap shows a low energy component which is, however, not associated with halide segregation but rather originates from photon reabsorption and cavity effects. [44,72] This is also supported by the fact that the PL of the same samples deposited on ITO do not show the low energy compo nent (Figure 1b), and that it is absent for the perovskite with the wider bandgap, where more bromide is inserted. We cal culated the PL quantum yield (PLQY) and the corresponding quasiFermi level splitting (QFLS PLQY ), which sets the upper limit for the attainable V oc . In Figure 3d, these parameters are compared to the QFLS in the radiative limit (QFLS RAD ). The PLQY for the bare perovskite films was found to be relatively low and decreased slightly from 0.14% to 0.11% for the lower to the wider bandgap materials, respectively. The corresponding QFLS PLQY values are only marginally higher (50-100 meV) com pared to the qV oc measured in full devices, where the perovskite is sandwiched between transport layers and contacts. This suggests that the nonradiative losses (difference between QFLS RAD and QFLS PLQY ) are associated with recombination in the bulk of the perovskite film, and only to a less extent with surface/interface recombination. A similar behavior was pre viously observed for other vacuum deposited wide bandgap perovskites. [52] We further tested the properties of the wide bandgap perov skite solar cells upon storing them in an inert atmosphere at 85 °C and periodically measured their J-V characteristics (corresponding parameters in Figure 4). This protocol is sim ilar to the ISOS protocols reported elsewhere, [73] although it is carried out in inert atmosphere and not in (humid) air as sug gested in the ISOSD2 and ISOSD3 test protocols. This is because testing stability in air is actually much more related to testing the encapsulation of the device, and not the semicon ducting materials of the solar cell. As our focus is on the mate rial science, we believe that only working in inert atmosphere can give valuable information.
Adv. Funct. Mater. 2023, 33, 2214357   Figure 3. a) EQE spectra in the bandgap region and calculated bandgap and Urbach energies. b) The ratio of measured V oc , J sc , and FF to their maximum theoretical (radiative) limit (lines are guides to the eye). c) Absolute calibrated PL spectra of perovskite films on glass under laser excitation at 515 nm and 1 sun equivalent light intensity. d) Comparison of the measured qV oc , QFLS calculated from PLQY (QFLS PLQY ) and QFLS in the radiative limit (QFLS RAD ) as a function of the perovskite bandgap, which is estimated from a bi-Gaussian fit of the PL spectrum. Analysis is based on devices stressed at 85 °C for 1 h. In general, all the MAPb(I 1x Br x ) 3 solar cells were found to be more thermally stable than other previously reported vacuumdeposited wide bandgap devices, [53] and comparable to MAPbI 3 solar cells. [55,56] The time to reach 80% of the initial efficiency (t 80 ) was found to exceed 3500 h for the 1.64 eV solar cells and to diminish for perovskite cells with wider bandgap. The 1.68 eV showed a t 80 of ≈2600 h (which is still remarkable), while the wider bandgap cells (1.71 eV) had a t 80 of ≈850 h. The main drivers for the device degradation are the current density, monotonically diminishing during thermal stress, and the FF, which was found to decrease faster for wider bandgap devices. The V oc , on the other hand, was found to be rather stable during the thermal stress test. The FF reduction might be also related to the thermal degradation of the organic transport layers, and/ or loss of ohmic contact at the electrode interface. In general, however, perovskite formulations with increasing bromide content becomes less thermally stable, as clearly demonstrated recently by highthroughput screening experiments. [74] We also performed measurements under continuous illumination, tracking the maximum power point of a solar cell with bandgap of 1.68 eV ( Figure S9, Supporting Information). This initial assessment of light soaking stability suggests that the devices are also stable under continuous illumination, as only a minor decrease of the power output is observed after 6 days.
We prepared ST solar cells with the perovskite with 1.64 eV bandgap, obtained by replacing the Ag top contact with ITO, deposited by pulsed laser deposition (ITO PLD ), as recently reported. [75] LiF (100 nm) was used as the antireflecting coating (ARC). We chose the most efficient formulation (1.64 eV), and optimized its thickness to maximize the power output of the ST solar cells. We found that the J sc of asprepared cells can be increased up to >19 mA cm −2 , on average ( Figure S10a, Sup porting Information), by increasing the perovskite thickness to 800 nm. However, the V oc was found to scale inversely with thickness ( Figure S10b, Supporting Information), indicating again that nonradiative recombination in the bulk of the perovskite is the main voltage loss for this material. The best tradeoff was found for solar cells with the 1.64 eV absorber with 700 nm thick films, leading to a PCE for asprepared ST solar cells of 15%, on average ( Figure S10d, Supporting Infor mation). With the same bandgap and thickness, we also pre pared a reference opaque cell with an Ag electrode, which showed a record efficiency of 19.4% after 24 h thermal stress at 85 °C ( Figure S11, Supporting Information). As a proofof concept, we tested the potential of these MAPb(I 1x Br x ) 3 perov skite solar cells in 4Ttandem devices in combination with a mainstream passivated emitter rear cell (PERC) crystalline Si (cSi) device. For these cells, the top BCP/ITO PLD transparent electrode was substituted with a more NIR transparent top contact, consisting of a SnO 2 buffer layer (40 nm) prepared by spatial ALD (sALD) and a sputtered ITO contact (180 nm), as previously reported. [76,77] An antireflective coating (ARC, MgF 2 ) was deposited on both sides of the devices (glass: 100 nm; ITO: 160 nm). Figure 5a shows the J-V curves for ST perovskite solar cells completed with sALD and sputtered ITO. The aspre pared devices are already well performing without extra post annealing, due to the thermal budget during the sALD process (100 °C for 20 min). Thanks to the use of MgF 2 ARC, the J sc was found to be 19.6 ± 0.2 mA cm −2 in the ST configuration. A FF of 74.3% ± 2% and a slightly lower V oc (as compared to refer ence opaque cells) of 1.13 ± 0.01 V were measured. The latter is related to the different top contact used for these ST devices. The overall PCE was 16.5 ± 0.5% on average, and >17% for the best pixel, as measured from its stabilized power output (inset of Figure 5a). In order to estimate the efficiency of the perov skite solar cell in the tandem configuration, we measured its EQE and transmittance (Figure 5b). The EQE spectrum shows a maximum of 90% in the 500-600 nm range, with the lower values in the NIR region ascribed to nonoptimal perovskite thickness. [78] The EQE minimum in the 400-500 nm range is due to the low transmittance of the commercial glass/ITO sub strate used ( Figure S12  EQE spectra might be due to a different reflectance of the two systems. Also the transparency of the full stack was rather low in the NIR, due to the high absorptance of the substrate ( Figure S13, Supporting Information) and to unfavorable inter ference fringes. As a bottom cell, we used a PERC silicon solar cells fabricated on M2 (156.75 × 156.75 mm 2 ) ptype crystalline silicon (cSi) wafers (photographs of the semitransparent perov skite cell and of the PERC Si cell are shown in Figure S14,S15, Supporting Information, respectively). Details of the cSi solar cells fabrication are reported in the supplementary information, and the device layout is illustrated in Figure 5c. The integrated current density estimated from the EQE spectrum of the PERC Si solar cell was 38.2 mA cm −2 , and the solar cell delivered a PCE efficiency of 21.23%, as measured by its J-V characteristics (Figure 5d, PV parameters are also reported).
Due to the large area mismatch between the ST perovskite cell and the cSi solar cell, it is not possible to directly measure the two devices in a 4Ttandem configuration. Thus, to esti mate its efficiency, we used the optical transmittance of the ST perovskite device, and calculated the filtered efficiency of the bottom cSi cell, assuming the ST perovskite device is used as a top cell. [76] Details of the method used to calculate the tandem efficiency are reported in the Supporting Information. Due to the low NIR transparency of the glass/ITO substrate of the ST perovskite cell, the current density of the filtered cSi cell is strongly reduced to 12.7 mA cm −2 . Taking into account the rest of the parameters for the filtered bottom cSi (see Table 2), the efficiency of the tandem device is estimated to be 23.3%, a 2.1% absolute gain as compared to the single junction cSi. The overall 4Ttandem efficiency is limited on one hand by the low NIR transparency of the glass/ITO substrate used for the ST perovskite solar cell fabrication. If the NIR transparency were improved to above 90%, as shown previously by some of us, [76] using a validated calculation for 4Tperovskite/Si tandem cells we estimated the efficiency of our 4Ttandem with PERC Si cell to be 26.5%. [79] On the other hand, the efficiency could be  3 . Inset shows the stabilized power output for the same device. b) EQE (red) and transmittance spectra (T, blue) of the same device. c) EQE spectrum of a PERC type Si solar cell; in the inset, a schematic of the Si solar cell layout is shown. d) J-V curve measured under simulated solar illumination for the same Si solar cell. further enhanced by using a more efficient Si solar cell, such as interdigitated back contact (IBC) or heterojunction (HJT) crys talline silicon cells.

Conclusion
In summary, we present a simple process to obtain wide bandgap perovskite solar cells via dualsource vacuum depo sition. This is possible when a mixed bromide/iodide lead precursor is preformed and sublimed as a single compo nent. The bandgap of the perovskite can be simply tuned by changing the bromide content in the precursor compound. The MAPb(I 1x Br x ) 3 films have been tested in vacuum depos ited solar cells, leading to efficient devices, comparable or even superior to those obtained with mixed cation perovskite formu lations. Importantly, the MAPb(I 1x Br x ) 3 solar cells are found to be highly thermally stable, with lifetimes (estimated as the time to reach 80% of the initial efficiency) in the order of thousands of hours, depending on the bandgap. This finding agrees with what was previously observed for vacuum deposited MAPbI 3 , and is somewhat in contrast with reports on solutionprocessed perovskite cells, where methylammonium is typically used as a minority cation to ensure stability. As a proofofconcept, we fabricated MAPb(I 1x Br x ) 3 solar cells in a semitransparent con figuration obtaining stabilized power efficiency exceeding 17%, and estimated their power output in 4Ttandem devices with a mainstream PERC silicon bottom cell. Future studies will be directed towards the development of 2Ttandem devices, either with silicon or with other complementary absorbers.

Experimental Section
A detailed description of the materials and methods is available in the Supporting Information.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.