To address the Voc deficit issue of bromide perovskites we sought for the introduction of bulky cations from chloride salts, namely the neo-Pentylammonium chloride (hereafter called NEO) and/or iso-Pentylammonium chloride (hereafter called ISO), on the surface of 150nm thick FAPbBr3 perovskite film, as presented in Fig. 1a. In fact, NEO has previously been demonstrated as an effective 2D passivator capable of significantly increase the Voc of devices based on MAPbBr3, resulting in Voc values as high as 1.65V 17. We also investigated an isomer to NEO, namely the iso-Pentylammonium chloride that has not yet employed in PSC technology. In order to grasp information about the role of the introduction of both bulky chloride cations, morphological characterizations such like Scanning Electron Microscopy (SEM) and Atomic Force Microscopy (AFM) planar images have been performed. Low and High magnification SEM images (Figure S1) confirmed morphological changes at the surfaces of the polycrystalline layer using both NEO and ISO bulky chloride cations. It is confirmed from AFM images (Figure S2), where it is also found an increase of the surface roughness in NEO (18nm) samples with respect to the REF and ISO samples (12 nm). A combined low and high angle XRD investigation was performed, offering an important insight into the effects of NEO and ISO on the structural properties of the perovskite layers. As visible in Fig. 1b pure α-phase perovskite (see reported Miller indexes) was detected for all samples. Moreover, crystallographic FTO signature labelled accordingly to ICDD card nr. 00-003-1114 was also observed. In the inset of Fig. 1b, the low angular region of the patterns is reported, allowing to observe the eventual formation of 2D perovskites. Indeed, only ISO-containing samples evidences the presence of 2D perovskite, adopting the Ruddlesden–Popper type crystal structure. The (020) reflection, and subsequent multiple [0k0] orientations, was identified at 2θ = 4.0°, corresponding to the formation of the [PbI6]4− octahedral layers. Such orientations and inter-planar spacing typically correspond to a number of 2D layers n = 2 21–23.These experimental evidences therefore suggest that ISO can form a 3D-2D interface while NEO is expected to be simply adsorbed on the perovskite surface.
We employ photoemission spectroscopy measurements to examine the chemical changes at the surface of the FAPbBr3 film due to the treatment with the NEO and ISO surfactants, respectively, as well as the corresponding effect on the surface energetics, which denote critical parameters for the functionality of the perovskite interface24. In this regard, we determined the work function of the perovskite films by analyzing the secondary electron cutoff of the ultraviolet photoemission spectra (UPS). Our findings indicate that the addition of NEO and ISO treatments has only a minor impact on the vacuum level position of the perovskite film (as shown in Fig. 1c). Specifically, while the initial work function of the REF sample was measured at 3.85 eV, it increased to 3.9 eV for the NEO sample and decreased to 3.8 eV for the ISO sample. Notably, although no clear trend emerges for the work function, the valence band spectra of the perovskite reveal a significant shift following treatment with the molecules. While the initial valence band onset was observed at 1.5 eV relative to the Fermi level, this value increased to 1.6 eV for the NEO sample and 1.8 eV for the ISO sample. This change suggests that the Fermi level at the surface is shifted further towards the conduction band (Fig. 1d), which we project from a constant band gap of 2.3 eV in accordance with our photoluminescence data. These findings align with earlier studies showing that the Fermi level at the perovskite surface is close to the conduction band if the perovskite is deposited on an n-type substrate and exhibits a low concentration of defect states in the bulk25. However, the presence of defect states at the surface can introduce a pinning of the Fermi level to the energy level of defect states, as may be the case for the pristine FAPbBr3 surface. Our data suggests that the molecular treatment, particularly with the ISO molecule, passivates such defect states, thus shifting the Fermi level back towards the conduction band edge. Notably, this shift in the position of the Fermi level closer to the conduction band due to the molecular surfactants is not detrimental for carrier extraction and blocking, as the final energy level alignment changes once the top interface with is formed. In all cases, we do not expect any significant barrier for hole as the ionization energy of the REF, NEO and ISO samples, remain high at 5.4 eV, 5.4 eV and 5.6 eV, respectively, well above the ionization energy of PTAA, which amounts to 5.2 eV 26.
The X-ray photoemission spectroscopy (XPS) data, depicted in Fig. 1e, indicates that the carbon and nitrogen signal of the REF and NEO do not show any marked difference. In contrast to that, the ISO sample exhibits an increase of the contribution of C-C bonds and a concomitant decrease of contribution of the C-N-C bonds to the C 1s spectra. This goes along with an increase of the ammonium content located at 402.4 eV in the N 1s spectra and a concomitant decrease of formamidinium. Similarly, we exclusively find traces of chlorine (not shown here) in the ISO samples. Thus, only for the ISO sample a significant amount of new molecular species is found at the surface. Of further note, the oxygen content is not negligible for the bare FAPbBr3 surface, but the amount decreases after NEO treatment, whereas in case of the ISO treatment we find no more oxygen at the surface. These results are in agreement with the suggestion of reduced surface defect densities and changed surface energetics via the NEO and ISO treatments. The Pb and Br core levels do not exhibit any significant changes and in particular, we observe the absence of metallic lead (Figure S3).
We consider the following reference ST-PSC stack based on nip architecture: Glass/FTO/TiO2:SnO2/FAPbBr3/PTAA/ITO (Fig. 2a). Cells have an active area of 1cm2. The compact TiO2 film deposited through spray pyrolysis is doped with Nb (2% atomic ratio to Ti in the precursor solution) to enhance its electrical conductivity 27. An ultra-thin layer of SnO2 nanoparticles was used to decorate the surface of the Nb:TiO2 layer, as recently reported by Kim and coauthors 28. This approach enabled us to benefit from the good electron extraction of SnO2 yet maintaining the structural robustness of the TiO2 layer. The choice of PTAA, which is doped with LiTFSI and 4-tBP, stems from its better compatibility with the sputtering of ITO top electrode with respect to Spiro-OMeTAD 12. In Fig. 2b we report the PCE box charts measured at Maximum Power Point (MPP) and 1Sun AM1.5G illumination conditions performed after the measurements of the J-V characteristics (see Figure S4-S6 and S.I. explanations for details on measurement conditions).
The reference devices performed quite poorly with an average PCE around 4.2% and a broad distribution of PCEs between 3.5% and 5%. As shown in Fig. 2c, the Voc of the REF PSCs is well below 1.5 V even for the most performing devices. This represent a voltage loss above 0.8V with respect to the FAPbBr3 bandgap (2.30 eV). In order to overcome this limitation, both NEO and ISO passivators have been introduced for the ultra-thin ST-PSC. By utilizing both salts as passivation agents, we successfully achieved Voc values above 1.6V, albeit with notable differences between the two. Firstly, we observed that the PTAA solution in toluene exhibited limited spreading over the perovskite surface treated with ISO, whereas NEO had a negligible effect on this parameter. It is noteworthy that at higher concentrations, both salts induced poor PTAA wettability, but at the concentration of 1mg/mL, this phenomenon was only observed with ISO.
We achieved a high Voc (up to 1.65V) using NEO, which served as an excellent starting point, but was unfortunately accompanied by a degree of hysteresis and/or instability during the initial J-V scans (Figure S7). In contrast, the ISO-passivated devices showed slightly lower average Voc likely due to poorer PTAA coverage but higher Jsc and FF values when compared to NEO (Figure S8). In addition, the devices passivated with ISO show a more stable electrical output across different J-V scans interspersed with a MPP tracking of two minutes as shown in Figure S7. The first and second J-V scan of the ISO devices practically overlap and the MPPT yields an even higher efficiency than the J-V curve. On the other hand, the first J-V curve from NEO devices usually show a marked hysteresis, although this was ameliorated in the second scan,, and a MPP tracking showing a bump in the firsts seconds followed by plateau with a slightly negative slope (Figure S9). Interestingly, after monitoring the tracking of the Voc under 1 Sun light exposure, we observe that NEO passivated device show a trend similar to the MPPT: a Voltage overshoot in the first 50s (reaching a Voc of 1.65 V) with a subsequent stabilization at lower values. On the opposite, the ISO devices reach a Voc plateau of 1.65V within the same time span (Figure S10).
Considering the complementarity of the two passivation strategies we operate both passivation at the same time (hereafter called ISO-NEO). In ISO-NEO passivation the ISO and NEO stock solution are mixed 1:1 v/v prior to use. Interestingly, with the ISO-NEO samples we obtained a higher efficiency compared to the single passivation schemes. ISO-NEO passivation helped to solve the issue related to the poor surface wettability of PTAA solution improving all the average PV parameters (and shrinking the box chart distribution) showing an outstanding maximum Voc of 1.73V (Fig. 2c). From our knowledge, this represent the best Voc value ever reported for PSC technology using opaque or semi-transparent device stacks (Fig. 2d). The Voc tracking and the J-V/MPPT characterizations are in agreement with ISO results but showing huge improvement of all the PV parameters with respect to the REF sample (Fig. 2e). Furthermore, XRD confirms the presence of 2D perovskite also for the ISO-NEO sample (Figure S11).
Notably, all devices, including the REF sample, exhibited excellent shelf-life stability when stored without encapsulation in ambient air, with their performances even improving during the first two to four weeks of storage (Figure S12). However, when testing their MPPT stability in air, the ISO-NEO passivation exhibited superior performance compared to the PSCs using only REF, NEO, and ISO. The REF and NEO devices exhibited a clear burn-in within the first 10–20 hours before stabilizing their power output to approximately 40–50% for REF and 60–80% for NEO devices of the initial efficiency after 100 hours. Conversely, the ISO and ISO-NEO devices retained more than 85% of the initial efficiency, with ISO exhibiting a slow, linear decreasing trend without any evidence of burn-in. The burn-in degradation is a common behavior in perovskite and organic solar cells, for which several mechanisms have been proposed 30,31. In our case, the origin of the burn-in is likely to originate at the interface between PTAA and the perovskite, and the introduction of ISO represents a winning strategy in this regard. Prolonged light soaking test at MPP have been performed using ISO-NEO passivation resulting in a negligible relative PCE variation (-6%) after 400hours of ageing (Figure S13).
The impact of the bulky cations at the perovskite/HTL interface on the optoelectronic and transport properties of full stacks (FTO/Nb:TiO2/SnO2/FAPbBr3/PTAA) has been investigated by two spatially32 resolved multidimensional imaging systems: a Time-Resolved Fluorescence Imaging (TR-FLIM) set-up33 and a spectrally resolved Hyperspectral Imager. In particular, we aimed at investigating the level of coverage of the perovskite absorber by the passivating agents and their impact on quantitative parameters related to the main photovoltaic figures of merits such as quasi-Fermi level splitting and carrier decay times. We recently used the same approach to investigate carrier recombination dynamics in high efficient inverted PSCs dual passivated by organic cations34.
First, photometrically calibrated and spectrally resolved maps were acquired on a reference sample and on three different stacks with NEO, ISO and ISO-NEO cations added at the interface between the perovskite and the PTAA. We then performed fitting with the generalized Planck’s law to obtain local estimates of quasi-Fermi level splitting (QFLS) and band gap energy Eg - details of this fitting being given in the supporting information. The averaged recorded spectra are reported in Fig. 3a. We observe a significant increase of the PL maximum intensity of the passivated samples compared to the reference, as well as a slight blue shift. The ISO and ISO-NEO samples exhibit similar PL average spectra. To have further insight on the carrier recombination dynamics we determined the local decay time on the different stacks. The precise methods and details are given in the supporting information. The resulting averaged decays are displayed in Fig. 3b where we observe an impressive increase of the decay times from the reference sample with a time constant around 27 ns to the ISO at 135ns, the NEO at 170 ns and at last the ISO-NEO with 230ns. In Fig. 3c we display maps of QFLS obtained at 1 Sun equivalent illumination for the different passivation strategies. We observe a gradual improvement from the reference (i) with a QFLS of 1.79 eV to the NEO (ii) at 1.81 eV and to the ISO-NEO and ISO (iii) and (iv) with a QFLS of around 1.83 eV, leading to a 17% reduction in voltage loss from the radiative limit of 2.02 eV. This demonstrates the beneficial effect of these compounds in minimising non-radiative losses at the absorber/HTL interface. These findings are in line with the UPS data (Fig. 1c) which suggested lower defect densities at the surface upon integration of the molecular surfactants with the ISO showing the most pronounced effect. Therefore, the formation of a 2D perovskite layer, that was evidenced by XRD analysis in the case of ISO and ISO-NEO, resulted in a significant improvement of the optoelectronic properties of the stack. The high level of QFLS can be directly related to the record Voc of 1.73 V obtained for a champion device passivated with ISO-NEO. We can notice a significant difference between the Voc and the observed QFLS. Sputtering damage from ITO deposition, which has already been reported for semi-transparent devices35, or a not perfect energetic alignment between the perovskite and the selective contacts, which could introduce differences between the QFLS and the actual Voc of the solar cells 36, could be possible explanations for such discrepancy. The latter hypothesis is corroborated by the reduction of the difference between average QFLS and Voc when the more performant passivation layer is deposited. Indeed, this parameter decreases from 390meV for the reference, to 260meV for NEO and to about 200meV for ISO and ISO-NEO compositions, which is nearly half of the value for the reference. Moreover, the Jsc significantly increased from an average value of 5.5mA/cm2 for the reference to 6.5mA/cm2 for the ISO-NEO composition, indicating a better carrier extraction in the passivated devices. Furthermore, we plot in Fig. 3d the maps of fitted bandgap after the cations addition. The bandgap slightly blue shifted from 2.28 eV for the reference device to about 2.30 eV for the passivated devices, suggesting an incorporation of some Cl of the cations in the bulk of the absorber thus increasing the gap. The resulting change in bandgap is small compared to the improvement of the QFLS. To evidence this, we plot in Fig. 3e the difference of the two quantities. This could be interpreted as the logarithm of the carrier density under operation as we expect:
We observe that in spite of the slight increase in Eg, the difference Eg-Δµ is significantly more favorable in the passivated devices, as shown in Fig. 3e. With a decrease of Eg -Δµ of ~ 1kT, the carrier density under solar operation is increased ~ 15%. These mappings also confirm the fact that the passivation is relatively homogeneous as the dispersion (6*standard deviation/mean) of the QFLS is 5.6% for NEO but as low as 3.5% for ISO-NEO. Moreover, we calculate the Urbach energy from the absorptivity curves, as reported in Figure S14. The absorptivity decay is purely mono-exponential, revealing an Urbach absorption below the bandgap with a related energy of 16 meV, confirming the usual low thermal and structural disorder in the perovskite absorbers37.
We observe a strong correlation between the electrical data (Voc) and the two optical characterizations described above. In Figure S15 we plot Voc and Eg-QFLS versus the decay times. These three parameters gradually improve from the reference sample to the single cation case (only ISO or only NEO), until reaching the maximal values for the mixed ISO-NEO passivation. The trends are thus similar, confirming the positive effect of such passivating agents on the optical and electrical properties of the devices. However, as previously reported by Zhu et al.17, we can observe a discrepancy between the rise in terms of QFLS (+ 40 meV) and the increase in terms of the Voc (+ 230 mV). This indicates that the improvement in device performance, and in particular of the Voc, is only partially due to interfacial passivation and that the introduction of the cations also improves the electrical behavior of the devices as a result of a possible higher shunt resistance, which is further supported by a significant rise in the FF. In Figure S16, we compared our results (in terms of QFLS and Voc) with the other ST-PSCs shown in literature concerning the qVoc/Eg (%) vs. Eg, a figure of merit introduced by Ruhle in 2016 38. However, we confirmed that our results represent the state of art for ST-PSC reaching 75.21% and 79.56% in terms of Voc and QFLS, respectively.
The analysis on the passivation supports us in rationalizing the increase in PCE. However, in order to improve the light utilization efficiency (LUE = AVT*PCE), it is important to consider also the optical properties of the ST-PSC39. The ST-PSCs show an AVT in the range of 61–64%, with small sample to sample variation (Figure S17). This value combined with PCE above 7% results in LUE above 4.5%, at the state of the art of transparent photovoltaics15,16. Further improvement can be achieved by developing light management routes, minimizing the reflection at air/glass (where the light impinges on the solar cell) and at the ITO/air (where the light exits the device) interfaces 40,41. We applied MgF2 anti-reflective coating (ARC) on the glass side and a spin coated Al2O3 nanoparticles thin film on the sputtered ITO electrode. In this way, we smoothed the refractive index gradient, thus minimizing the optical reflections (Fig. 4a). Notably, the AVT of the full device can be increased from 66–67% with the application of the MgF2 and further to 70.7% with the application of Al2O3 accompanied by negligible losses in Jsc (Figure S18). The most performing device delivered a 70.69% AVT and a PCE of 8.2% (Fig. 4c), resulting in a LUE of 5.7, overcoming the best results ever reported for semitransparent PSCs, as we show in Fig. 4d. This result is particularly impressive considering that it is achieved on 1cm2 active area solar cells. We should point out also the reduction of Voc observed for the ST-PSC where ARC is used (Voc = 1.64 V) compared to those without ARC (Voc = 1.73 V). ARC increase the photon outcoupling thus reducing the photon recycling in the ST-PSC. This impacts negatively in the QFLS splitting and consequently on the Voc 42.
By comparing the integrated current density from the IPCE measured by illuminating from the two different directions we obtained a bifaciality factor of 87%. We can clearly observe that the larger current loss occurs at low wavelength, which is mainly related to the parasitic light absorption from PTAA (below 400nm). Interestingly, the illumination side has practically no effect on the J-V curve confirming that the bifaciality factor approach 90% also when considering the power conversion efficiency (Figure S19). An important feature to consider is the PV performance under artificial (indoor) light illumination. When integrated in facades as solar windows, or into alternative architectural elements, the semitransparent solar cells could also be illuminated with artificial (indoor) light. While this is negligible when compared to the illumination from the sun, during the night an efficient conversion of indoor lighting could supply low power electronics, as those comprising the IoT paradigm (alarm, sensors). Interestingly, we obtained excellent performances under artificial indoor light, with a PCE between 16–17% in the range from 200 to 1000 lux (Figure S20). Notably, we could achieve a Voc above 1.2V at 1000 lux and a power density exceeding 60 uW/cm2 (13 uW/cm2 at 200 lux). These results indicate that a 1m2 solar window can reasonably deliver a power exceeding 100mW under indoor lighting, above the requirements for most household devices or wireless communication protocols10. In Fig. 4d and Fig. 4e, we also reported our state-of-art results in LUE (AVT) and PCE (AVT) behavior for ST-PSCs introducing the theoretical limits reported from Bing et al. 43.
Finally, 20cm2-sized semi-transparent perovskite solar modules (ST-PSM) with high geometrical fill factor (up to 97.83%) have been fabricated using 3D FAPbBr3 perovskite and ISO-NEO passivation scheme (Fig. 5a-b). The results are very encouraging showing a maximum steady-state PCE of 7.3% (7.1% in average) after 120s of MPPT, Voc up to 1.65V/cell, AVT equal to 65% (Fig. 5c) showing a LUE equal to 4.74. Statistical PV parameters measured in a batch of eight ST-PSMs are reported in Figure S21. Finally, we speculate a suitable exploitation of the ST-PSMs when integrated in a BIPV window (Fig. 5d). The sketch highlighted the bifacial working operation of the ST-PSM powered BIPV window able to potentially generate electricity switching from outdoor (day) to indoor (at night, with artificial illumination) working conditions.