Lead‐Free Perovskite‐Inspired Absorbers for Indoor Photovoltaics

With the exponential rise in the market value and number of devices part of the Internet of Things (IoT), the demand for indoor photovoltaics (IPV) to power autonomous devices is predicted to rapidly increase. Lead‐free perovskite‐inspired materials (PIMs) have recently attracted significant attention in photovoltaics research, due to the similarity of their electronic structure to high‐performance lead‐halide perovskites, but without the same toxicity limitations. However, the capability of PIMs for indoor light harvesting has not yet been considered. Herein, two exemplar PIMs, BiOI and Cs3Sb2ClxI9‐x are examined. It is shown that while their bandgaps are too wide for single‐junction solar cells, they are close to the optimum for indoor light harvesting. As a result, while BiOI and Cs3Sb2ClxI9‐x devices are only circa 1%‐efficient under 1‐sun illumination, their efficiencies increase to 4–5% under indoor illumination. These efficiencies are within the range of reported values for hydrogenated amorphous silicon, i.e., the industry standard for IPV. It is demonstrated that such performance levels are already sufficient for millimeter‐scale PIM devices to power thin‐film‐transistor circuits. Intensity‐dependent and optical loss analyses show that future improvements in efficiency are possible. Furthermore, calculations of the optically limited efficiency of these and other low‐toxicity PIMs reveal their considerable potential for IPV, thus encouraging future efforts for their exploration for powering IoT devices.


Introduction
With the emergence of Big Data and the Internet of Things (IoT), the demand for self-powered wireless devices (e.g., smart sensors for health and wellness monitoring, and for smart buildings and smart packaging) has substantially increased over the past few years. [1][2][3][4][5][6] Self-powered devices are expected to play an important role in enabling the growth of the IoT ecosystem to fulfil its future potential in terms of the number of devices and complexity, in which battery-powered devices alone are no longer expected to be practical in all cases. [7][8][9] A significant proportion of these wireless devices are deployed in indoor environments where varying light intensities and spectra are present, for which efficient and reliable indoor photovoltaics (IPV) would be attractive as a power supply. Generally, the intensity of indoor light sources is approximately three orders of magnitude lower than the standard terrestrial outdoor AM 1.5G solar spectrum. The illuminance of fluorescent light (FL) sources and white light emitting diodes (WLED) commonly found in residential and commercial buildings is in the range of 200-1000 lux. [10] The corresponding irradiance is approximately 50-300 µW cm −2 , with spectra spanning predominantly over the visible wavelength range (400 ≲ λ ≲ 700 nm). [11][12][13] As a result, although commercial crystalline silicon solar cells are suitable for harvesting terrestrial solar radiation, their bandgaps (1.12 eV) are too narrow for optimal harvesting of indoor lighting. [14] Furthermore, crystalline silicon solar cells suffer from significant Shockley-Read-Hall recombination under low light conditions. [15] Therefore, alternative absorbers with wider bandgaps are needed for IPV applications, and the technologies explored to date are based on hydrogenated amorphous silicon (a-Si:H), dye-sensitized solar cells, lead-halide perovskites and organic semiconductors. [16][17][18][19] Hydrogenated amorphous silicon has become the industry standard for IPV because its efficiency under indoor illumination, ranging from 4.4% to 9.2%, is adequate for typical electronic devices powered by photovoltaics. [15,[20][21][22][23] Among the emerging absorbers, lead-halide perovskites have been identified as particularly promising for IPV owing to their ability to achieve excellent optoelectronic properties (e.g., long diffusion lengths >1 µm) when grown by low-cost, facile methods. [24][25][26][27][28][29] Lead-halide perovskite IPV with power conversion efficiencies With the exponential rise in the market value and number of devices part of the Internet of Things (IoT), the demand for indoor photovoltaics (IPV) to power autonomous devices is predicted to rapidly increase. Lead-free perov skite-inspired materials (PIMs) have recently attracted significant attention in photovoltaics research, due to the similarity of their electronic structure to high-performance lead-halide perovskites, but without the same toxicity limitations. However, the capability of PIMs for indoor light harvesting has not yet been considered. Herein, two exemplar PIMs, BiOI and Cs 3 Sb 2 Cl x I 9-x are examined. It is shown that while their bandgaps are too wide for single-junction solar cells, they are close to the optimum for indoor light harvesting. As a result, while BiOI and Cs 3 Sb 2 Cl x I 9-x devices are only circa 1%-efficient under 1-sun illumination, their efficiencies increase to 4-5% under indoor illumination. These efficiencies are within the range of reported values for hydrogenated amorphous silicon, i.e., the industry standard for IPV. It is demonstrated that such performance levels are already sufficient for millimeter-scale PIM devices to power thin-film-transistor circuits. Intensity-dependent and optical loss analyses show that future improvements in efficiency are possible. Furthermore, calculations of the optically limited efficiency of these and other low-toxicity PIMs reveal their considerable potential for IPV, thus encouraging future efforts for their exploration for powering IoT devices.
of 36% under low-intensity indoor lighting have recently been demonstrated. [30] Despite their attractive performance, the use of lead-halide perovskites raises toxicity concerns. Although the lead concentration in lead-halide perovskites is lower than in natural soil, there is evidence that the highly-soluble lead content of lead-halide perovskites is ten times more bioavailable. [31,32] While the challenges of lead-containing compounds in outdoor solar applications are mostly downstream (e.g., recycling and disposal of cells at their end of life), the use of lead-halide perovskites in an indoor environment poses significantly higher risks. [33] In an indoor setting, the encapsulation layer would be the only barrier between the human end-user and the lead contained within the perovskite device, which increases the potential for lead exposure in case of damage to the device. [7,34] Moreover, the use of lead-halide perovskites to power wearable electronics and other indoor IoT devices faces regulatory challenges (e.g., the limit of 0.1 wt% Pb content set by the EU Restriction of Hazardous Substances (RoHS) directive). [35] Furthermore, lead-iodide based perovskites (e.g., methylammonium lead iodide) have a bandgap of approximately 1.6 eV (depending on the A-site cation), which is below the optimal bandgap for indoor light harvesting (1.9 eV). [36] Alloying with ≈70% bromide in place of iodide is needed to raise the bandgap to 1.9 eV, but the resulting materials are not photostable. Indeed, it has been shown that illuminating these mixed iodide-bromide perovskites with 15 mW cm −2 visible light for <10 min results in phase-separation and the bandgap red-shifting and settling at 1.68 eV. [36] In the broader photovoltaics research landscape, investigators have been actively searching for lead-free perovskiteinspired materials (PIMs) with the aim of mitigating the toxicity challenges of lead-halide perovskites. [37] However, this effort has been primarily aimed at outdoor solar harvesting, [37][38][39][40][41] while the potential of PIMs for IPV has been unexplored to date. Among lead-free perovskite-inspired absorbers, halides and chalcohalides based on Bi 3+ and Sb 3+ have attracted a considerable amount of attention due to the similarity of their electronic structure to that of lead-halide perovskites, namely the s-p hybridization in the valence band and the antibonding-toantibonding orbital transition across the bandgap. [42] Bismuth and antimony are also substantially less toxic than lead, and are sufficiently abundant for large-scale commercialization. [43][44][45] In devices, bismuth-and antimony-based absorbers have recently demonstrated promising performance under AM 1.5G illumination, [40,46] but have yet to reach levels comparable to their lead-halide counterparts. [44] An important limitation they face in outdoor solar harvesting is their wide bandgap of ≈2 eV. However, such bandgap values are in fact optimal for indoor light harvesting owing to the spectra of indoor light sources being blue-shifted compared to the AM 1.5G spectrum. [7,14,15] Furthermore, PIMs absorbing in the visible light spectrum come in aesthetically appealing colors, thus addressing the need for photovoltaics that can also carry out a decorative function when deployed in the objects and environments of our daily lives. [32,45,[47][48][49][50][51][52] Additionally, antimony-and bismuthbased PIMs can be synthesized using techniques that are compatible with low-cost flexible plastic substrates, thereby pointing to their integration into a wide range of IoT devices. [53][54][55][56] In this study we aim to shed light on the indoor photovoltaic capabilities and potential of lead-free perovskite-inspired absorbers. Specifically, we first investigate two exemplar absorbers, and subsequently assess the significance of our findings in the broader PIM landscape. The exemplar materials we first study in detail are bismuth oxyiodide (BiOI) and the defect-ordered perovskite cesium antimony chloride-iodide (Cs 3 Sb 2 Cl x I 9-x ). Both materials have bandgaps close to 1.9 eV without requiring any further alloying. Critically, while many Bi 3+ and Sb 3+ PIMs have a 0D crystal or electronic structure, BiOI and Cs 3 Sb 2 Cl x I 9-x have layered (2D) structures. [44,46,47,50] This allows efficient charge transport within the covalently linked planes, such that >60% external quantum efficiencies could be achieved in devices. [32,45,56] Herein, we investigate the photovoltaic performance of BiOI and Cs 3 Sb 2 Cl x I 9-x devices under FL and WLED illumination, as compared to AM 1.5G radiation, and assess whether their performance is sufficient to power thin-film-transistor (TFT) circuits for digital processing. Furthermore, we perform intensity-dependent measurements and optical loss analyses to determine the pathways for achieving future increases in efficiency. Finally, we estimate the efficiency in the radiative limit and the spectroscopically limited maximum efficiency (SLME) of a range of perovskite-inspired absorbers under AM 1.5G and indoor light conditions in order to demonstrate the potential of these materials for indoor light harvesting applications, and make recommendations for PIMs that should be investigated further in the future.

BiOI and Cs 3 Sb 2 Cl x I 9-x Photovoltaic Devices
Thin films of BiOI and Cs 3 Sb 2 Cl x I 9-x were grown by thermal chemical vapor deposition (t-CVD) and solution processing, respectively, using our previously established routes. [45,56] BiOI has a layered structure, with a tetragonal unit cell (space group P4/nmm, no. 129), as shown in Figure 1a. Cs 3 Sb 2 I 9 , on the other hand, can adopt either a 0D structure (space group P6 3 /mmc, no. 194) or 2D structure (space group P3m1, no. 164). [56] As explained in our previous work, [56] the 2D phase can be stabilized by introducing Cl (Figure 1a). Stabilizing the layered phase is important for improving the charge transport properties of the material. [56] X-ray diffraction (XRD) patterns of the thin films grown show both materials to be phase pure with the layered structures ( Figure 1b). The BiOI grown on NiO x /ITO had a compact morphology (Figure 1c), with a film thickness of ≈700 nm and grains of comparable size, which adopt a {012} preferred orientation. This preferred orientation is favorable for charge extraction in a vertically structured device. [57] Cs 3 Sb 2 Cl x I 9-x had smaller grains (≈250 nm in size) with a lessdense morphology when grown on mesoporous TiO 2 /FTO, which we nonetheless found to allow devices with high photoconversion efficiency. [56] Critically, both BiOI and Cs 3 Sb 2 Cl x I 9-x have bandgaps close to the optimum value of 1.9 eV for IPV applications (1.93 eV for BiOI and 1.95 eV for Cs 3 Sb 2 Cl x I 9-x ), as shown in our previous works. [45,56] Both materials demonstrated roomtemperature photoluminescence (PL), as shown in Figure S1 (Supporting Information). The PL peak for Cs 3 Sb 2 Cl x I 9-x was centered at 1.96 ± 0.01 eV, which is consistent with the optical bandgap. The PL for BiOI, on the other hand, was centered at 1.77 ± 0.01 eV. We are currently investigating the origin of this Stokes shift and going into this in detail is beyond the scope of the present work. We investigated the photostability of the materials by repeatedly measuring the PL spectrum over the course of 30 min illumination with a pulsed 3.1 eV laser with a fluence of 14 µJ cm −2 pulse −1 . No shift in the PL peak occurred in either material ( Figure S1, Supporting Information), showing the materials to be photostable under the excitation conditions used.
We constructed devices from both materials using the architectures that have yielded the highest efficiencies under 1-sun illumination (Figure 2a,b, with device stacks shown in Figure S2, Supporting Information). [45,56] The external quantum efficiency (EQE) spectra of both devices demonstrate significantly greater overlap with the FL and WLED spectra than with the AM 1.5G spectrum (Figure 2c). This points to the considerably greater potential of these materials for IPV applications than for outdoor solar harvesting.
The current density-voltage (J-V) characteristics of BiOI and Cs 3 Sb 2 Cl x I 9-x devices were measured under AM 1.5G, FL (1000 lux), and WLED (1000 lux) illumination (Figure 2d-f; see Table 1 for the corresponding performance parameters). As shown in Figure 2e, devices based on BiOI and Cs 3 Sb 2 Cl x I 9-x have a power conversion efficiency (PCE) of 0.9% and 1.2% under AM 1.5G illumination respectively, which is in agreement with the median efficiency of these materials in prior reports. [45,56,58] However, under FL and WLED illumination, the PCE values increased to the range of 4-5%, peaking at 4.4% for BiOI and 4.9% for Cs 3 Sb 2 Cl x I 9-x under FL lighting (Figure 2d,f).
These PCEs have already entered the range of efficiencies reported for commercial a-Si:H IPV (4.4-9.2%). [15,[21][22][23] This confirms the appeal of these lead-free absorbers for IPV applications. Importantly, their efficiency values under indoor light sources correspond to a fourfold increase with respect to their PCE under 1-sun illumination, reflecting the closer match of their absorption spectra with indoor light sources.
The IPV performance of both BiOI and Cs 3 Sb 2 Cl x I 9-x devices without encapsulation was found to be stable over a 5 month period, during which they were kept in a N 2 -filled environment that contained >1000 ppm oxygen and >500 ppm H 2 O (stability curves shown in Figures S3 and S4 (Supporting Information); the distribution of the photovoltaic parameters is provided in Figure S5, Supporting Information). These conditions are similar to those that might be expected for encapsulated devices stored in ambient air. [59] In fact, the devices improved in performance after 5 months of storage in this environment, in which the PCE of Cs 3 Sb 2 Cl x I 9-x more than doubled, increasing from 2.3% to 4.9% under FL illumination, and from 1.9% to 4.4% under WLED illumination. In the case of BiOI, the PCE increased from 2.4% to 4.0% under FL illumination, and remained stable at ≈3.4% under WLED illumination before and after 5 months of storage. In particular, the PCE improvement of the BiOI device under FL lighting can be primarily traced to an improved J sc -which increased from 46 to 61 µA cm −2 -and fill factor-which increased from 32% to 45% (see Figure S3, Supporting Information). In contrast, the Cs 3 Sb 2 I x Cl 9-x device manifested an appreciable improvement mainly in the J sc and V oc , which increased from 50 to 82 µA cm −2 and from 0.41 to 0.49 V, respectively (see Figure S4, Supporting Information). The mechanism through which the improvement occurs is unclear, but, based on related findings  in the literature, possible mechanisms include: 1) oxygen passivation of layers or interfaces in the device stacks, or 2) improvement in the work function of charge transport layers, leading to improved band alignment. [60][61][62] While further work is needed to elucidate the exact mechanism, these results show that the materials exhibit promising stability for commercial applications, as well as very promising performance for powering IoT devices.

BiOI and Cs 3 Sb 2 Cl x I 9-x Indoor Photovoltaics for Powering Thin-Film Transistor Electronics
In order to assess their significance for IoT applications, we tested the ability of mm-scale BiOI and Cs 3 Sb 2 Cl x I 9-x devices to power printed thin-film transistor (TFT) circuits. Circuits based on solution-processed TFT technologies (e.g., featuring organic semiconductors, [63,64] amorphous metal-oxide semiconductors, [65,66] and semiconducting carbon-nanotube networks [67,68] ) are being intensively investigated with the aim of realizing flexible, easy-to-fabricate, and low-cost electronics for smart-sensor systems, wearable electronics, and the IoT ecosystem. [69][70][71] In particular, as a proof-of-concept, we investigated the ability of BiOI and Cs 3 Sb 2 Cl x I 9-x devices to power inverters (see experimental arrangement in Figure 3a) made with printed carbon nanotube (CNT) TFTs. [68] An inverter (also known as NOT gate) is an essential component of digital circuits, where it carries out the logic inversion of the digital signal applied at its input (i.e., it provides a "low" output for a "high" input and vice versa).
The measured voltage transfer characteristics (VTCs) of one of our inverters powered by a BiOI or a Cs 3 Sb 2 Cl x I 9-x device under FL lighting (500-1000 lux) are shown in Figure 3b,c. All measured VTCs (for all illumination levels) exhibited the expected near-step-like transition from "high" to "low" as the input signal is swept from "low" to "high." This indicates that the power supplied by the PIM devices under typical indoor illumination levels is sufficient to operate the printed TFT inverters. In fact, the power dissipated by the inverters is entirely  supplied by the PIM devices, from which the TFT channel current is drawn. In particular, voltage gain values (quantifying the steepness of the VTCs) greater than 7 V V −1 (see Figure 3b,c) are obtained at all illumination levels, denoting adequate inverter performance for robust digital signal processing. Moreover, the VTCs exhibit only minor shifts as the illumination level is varied, which can be traced to the weak dependence of the V oc on light intensity (see Figure 4). Importantly, this implies that our IPV devices could provide sufficient power for printed TFT electronics while such electronics is moved about in an indoor environment, which opens up attractive prospects for autonomous wearable electronics. These results demonstrate for the first time the capability of PIM IPVs to power printed TFT circuits. Crucially, we achieved this with a direct connection of one of the BiOI or Cs 3 Sb 2 Cl x I 9-x devices to a TFT circuit-i.e., without connecting several such devices in series in order to boost the overall voltage supplied, and without connecting them in parallel in order to provide sufficient current. This aspect relates to the capability of BiOI and Cs 3 Sb 2 Cl x I 9-x devices to provide a V oc in the range that keeps the power dissipation of the printed TFT electronics to a minimum (i.e., devices with a larger V oc would need to provide larger power for the same circuits to be able to function). It is also noteworthy that we achieved this by using BiOI and Cs 3 Sb 2 Cl x I 9-x devices with a particularly small area (7.25 mm 2 ), which is highly attractive for the realization of compact wearable electronic devices and smart-sensor nodes with minimal footprint (e.g., "smart dust"). In particular, the inverters employed in this proof-of-concept demonstration require a maximum current of 600 pA each to operate. Therefore, our IPV devices could supply sufficient power (while operating near V oc ) for electronics of complexity equivalent to ≈7000 transistors (e.g., 25 8-bit full-adders). This is a level of complexity sufficient to address real-world applications. Consequently, our proof-of-concept demonstration provides highly promising indications of the viability of the technological convergence between lead-free perovskite-inspired IPV and printed TFT electronics towards the realization of autonomous devices for emerging application areas.

Loss Analyses of BiOI and Cs 3 Sb 2 Cl x I 9-x Photovoltaics
A systematic investigation of the photoconversion mechanism of BiOI and Cs 3 Sb 2 Cl x I 9-x under low-light conditions was conducted in order to understand the main limiting factors that need to be overcome for future efficiency improvements to become possible. We characterized the optical power dependence of the open-circuit voltage (V oc ), fill factor (FF), and short-circuit current density (J sc ) over a power range of 10-200 µW cm −2 , which is relevant to indoor light sources. Varying the incident power density changes the quasi-Fermi level splitting, and studying how this affects the photovoltaic parameters gives insights into the dominant recombination type (trap-assisted recombination or band-to-band recombination). In these measurements, a monochromatic light source (λ = 525 nm) of varying intensity was used to ensure uniform illumination throughout the active layer. As shown in Figure 4a, the V oc of BiOI devices gave a slope of ≈kT/q, while a slope of 2kT/q was obtained from Cs 3 Sb 2 Cl x I 9-x devices (where T is the device temperature, k the Boltzmann's constant, and q the elementary charge). A slope equal to 2kT/q (≈0.050 V at room temperature) in the V oc versus light intensity plot has been associated with defect-assisted (Shockley-Read-Hall, SRH) recombination involving mid-gap states in the absorber. [72][73][74][75] Therefore, the observed light-intensity dependence of the V oc of Cs 3 Sb 2 Cl x I 9-x devices suggests that they are predominantly limited by nonradiative recombination. This is consistent with the low photoluminescence quantum efficiency (PLQE) of Cs 3 Sb 2 Cl x I 9-x , which we measured to be ≈0.1%, demonstrating recombination to be dominated by non-radiative processes. This may be related to the smaller grain size of Cs 3 Sb 2 Cl x I 9-x films, and possibly the presence of deep recombination centers. The discontinuous morphology of the Cs 3 Sb 2 Cl x I 9-x layers could also contribute to nonradiative recombination in devices (Figure 1c). With regard to BiOI, while a slope equal of kT/q in the V oc versus light intensity plot has been customarily attributed to band-to-band recombination, this is not consistent with our previous results, in which we showed the fluence-independence of the decay in photoluminescence (from time-correlated single photon counting measurements) and ground-state bleach (from transient absorption spectroscopy measurements). [45,58,74] Furthermore, we measured the PLQE of BiOI, which we found to be <0.1%. These results suggest that recombination in BiOI is dominated by nonradiative processes. In fact, a slope of kT/q in the V oc versus light intensity plot of BiOI devices could also originate from SRH recombination through shallow or mid-gap defect states in the presence of a high majority carrier concentration. [74,75] This could arise in the BiOI devices due to the space-charge region that could form next to the NiO x hole transport layer (illustrated in Figure S6, Supporting Information). We previously found, from systematic X-ray and ultraviolet photoemission spectroscopy measurements, that BiOI exhibits downwards band-bending next to the interface with NiO x , owing to the lower work function of NiO x compared to BiOI. [45] Such downwards band-bending could cause an accumulation of electrons next to the NiO x /BiOI interface under illumination, leading to the formation of a spacecharge region (refer to Figure S6, Supporting Information).
The trend in FF with light intensity may provide further insights into the recombination mechanism. For Cs 3 Sb 2 Cl x I 9-x , the FF increases with increasing light intensity (Figure 4c). This may be due to SRH recombination. The population of mobile carriers increases with increasing light intensity, while the density of traps remains the same. The FF trend for Cs 3 Sb 2 Cl x I 9-x in Figure 4c therefore seems to be consistent with its measured V oc dependence on light intensity. Another possibility, however, is that the FF increases with light intensity due to a shunting effect, which is also suggested by the imperfections in the Cs 3 Sb 2 Cl x I 9 layer we observed via SEM (see Figure 1c). On the other hand, BiOI exhibits a reduction in FF with increasing light intensity. This agrees with our previous intensity-dependent FF measurements made at higher light intensities (20-100 mW cm −2 ) and is consistent with the effects of a space-charge region in BiOI next to the interface with NiO x . [45] At lower light intensities, there is a smaller population of photogenerated carriers, hence a smaller space-charge region forms, and a larger fraction of carriers can be extracted. This indicates that BiOI devices made with the current device structure are more suitable for lower light-intensity applications.
Finally, the trends in J sc with light intensity give insight into the dominant recombination mechanisms at short-circuit conditions. Both Cs 3 Sb 2 Cl x I 9-x and BiOI devices exhibit a linear trend (Figure 4b), which is consistent with our previous observations at higher light intensities (20-100 mW cm −2 ), and is suggestive of defect-assisted recombination through mid-gap states. [39,56,76] To corroborate our analysis of the intensity-dependent performance measurements, we performed optical loss analyses. For both devices, the largest loss in current is due to unabsorbed photons. This is slightly larger for BiOI (31-38%) than for Cs 3 Sb 2 Cl x I 9-x (29% under 1 sun; ≈34% under indoor lighting; Table 2), which is in part due to BiOI having a lower absorption coefficient than Cs 3 Sb 2 Cl x I 9-x (Figure 1d). Improved light management to reduce optical losses (e.g., nanophotonic structures at the back electrode) would improve performance. Beyond unabsorbed photons, the next largest loss in current is due to uncollected carriers. This is larger in BiOI, where this loss process accounts for 20% of the current under all illumination conditions ( Table 2). This is consistent with the intensitydependent analyses above and also the optical loss analyses in our previous work, in which carriers were found to be lost due to sub-optimal band alignment in the device stack. [45] In Cs 3 Sb 2 Cl x I 9-x , uncollected carriers account for 16% of the  Table 2), which may arise from nonradiative recombination.
All things considered, the loss analyses show that one of the ways to improve the performance of Cs 3 Sb 2 Cl x I 9-x IPVs is to optimize the morphology in order to achieve a continuous film and larger grains. For BiOI, on the other hand, IPV performance could be improved by developing higher work function hole transport layers that prevent downwards band bending while maintaining dense, a/b-axis oriented platelets. The next question is what efficiencies these materials could achieve if all parasitic optical losses and limitations due to sub-optimal morphology or band positions could be eliminated. This is discussed in the next section.

Future IPV Potential of Lead-Free Perovskite-Inspired Materials
Similar to BiOI and Cs 3 Sb 2 Cl x I 9-x , there is a wide range of PIMs that have demonstrated modest photovoltaic performances under 1-sun conditions (typical PCEs of ≈1-2%) but have bandgaps that are more suited for indoor light sources. [44,46] To provide direction for the future exploration of the broader family of PIMs for indoor light harvesting, we calculated the theoretical efficiency limits (see Note S1, Supporting Information, for details on methodology) for BiOI and Cs 3 Sb 2 Cl x I 9-x , as well as a range of commonly investigated PIMs: rudorffites (Ag-Bi-I), vacancy-ordered double perovskites (Cs 2 TiBr 6 ), binary halides (BiI 3 , InI), and defect-ordered perovskites (A 3 B 2 X 9 absorbers). In this section, we first discuss the future potential of Cs 3 Sb 2 Cl x I 9-x and BiOI, before identifying other classes of PIMs that hold significant potential for IPV, which are worth future investigation.
The efficiency limits were calculated for FL and WLED sources with power density values of 3.4 and 3.2 W m −2 respectively (which correspond to the 1000 lux illuminance of the sources used in this work). A theoretical maximum efficiency of ≈62% can be reached under both FL and WLED lighting with absorbers possessing a bandgap of 1.9 eV (Figure 5) based on the detailed balance approach taken by Shockley and Queisser. [77] While the radiative limit (RL) gives an indication of the magnitude of the IPV efficiency that can be reached, it does not account for materials properties such as the absorption coefficient, and degree of non-radiative recombination at open circuit voltage conditions, which depend on the direct/indirect nature of the bandgap. A more realistic efficiency limit can be calculated by taking into account the absorptance, thickness of the absorber, and nature of the bandgap (direct vs indirect), along the lines pursued by Zunger et al. in their definition of the spectroscopically limited maximum efficiency (SLME). [78] It is noteworthy that, while the SLME limit has been applied to calculate the 1-sun optically limited efficiency of materials, to the best of our knowledge, a similar concept has not been applied to assess the potential of materials for IPV applications. Therefore, here we introduce the IPV-equivalent of the SLME, which accounts for the spectra of the indoor light sources, and we refer to it as the indoor SLME (i-SLME in short form; see Note S1, Supporting Information, for details). While SLME values are typically calculated using the absorptance derived from computed absorption coefficient values, [78] in this work we calculated i-SLME values based on the experimentally measured absorption coefficients reported in the literature, which would be more reflective of the maximum IPV efficiencies that could be ultimately reached. For all absorbers, a thickness of 500 nm was taken in the i-SLME calculations.
Although BiOI and Cs 3 Sb 2 Cl x I 9-x have very similar RL efficiency values ( Figure 5), BiOI has a shallower absorption onset and a lower absorption coefficient than Cs 3 Sb 2 Cl x I 9-x (≈10 4 vs ≈10 5 cm −1 , respectively; Figure 1d). As a result, in the i-SLME limit, the J sc and V OC of BiOI are lower ( Figure 5; Tables S2 and S3 and Figure S7, Supporting Information). Even if we take the extreme case where the fraction of radiative recombination (f r ) is 1 (i.e., we have a direct bandgap), the maximum efficiency would be only 44% under FL and WLED  illumination. Films thicker than 500 nm would increase the proportion of absorbed photons but could lead to increased series resistance. Thus, although BiOI has achieved a similar PCE under indoor lighting as Cs 3 Sb 2 Cl x I 9-x , we would expect that Cs 3 Sb 2 Cl x I 9-x could ultimately achieve higher IPV performance with optimized morphology and device structure (see Section 2.3). Nevertheless, BiOI has the important advantages of being stable in air without encapsulation, and processable using a scalable t-CVD method. [45] These advantages will be important for commercial applications, and future development of BiOI for IPV is warranted.
Our analysis also illustrates the importance of being able to control the polymorphism in the Cs 3 Sb 2 I 9 system. Figure 5 shows that the 0D structure has i-SLMEs that are approximately 5% lower than the 2D structure. This is primarily due to the lower V OC of the 0D compound. The 0D polymorph has an indirect bandgap, whereas the 2D structure has a bandgap that is nearly direct. [47,50,56] This results in the 0D compound having a smaller f r value, which lead to a lower calculated V OC value (refer to Tables S1-S3, Supporting Information). The layered Cs 3 Sb 2 Cl x I 9-x compound is therefore a promising material for further investigations for IPV and has some of the highest predicted i-SLMEs (50-54%; see Figure 5).
In order to benchmark these findings, it is useful to compare them against a-Si:H, which is the workhorse compound for IPV. While the i-SLMEs of a-Si:H are at 47-50% ( Figure 5; Tables S2 and S3, Supporting Information), a-Si:H has an IPV efficiency that has long saturated to values in the range of 4.4-9.2%. [15,[20][21][22] BiOI and Cs 3 Sb 2 Cl x I 9-x provide IPV efficiencies that are already in the same range as a-Si:H (see Section 2.1). Considering that this is the first report on BiOI and Cs 3 Sb 2 Cl x I 9-x for indoor light harvesting, and that we have already identified strategies that could enable future increases in their IPV efficiencies (see Section 2.3), our i-SLME calculations confirm the potential of BiOI and Cs 3 Sb 2 Cl x I 9-x to surpass the IPV performance of a-Si:H.
Beyond Cs 3 Sb 2 Cl x I 9-x and BiOI, our analysis in Figure 5 shows that there is a wide range of other PIMs that are highly promising for IPV and have significant potential to exceed the performance of a-Si:H IPV. These include the Ag a Bi b I x (x = a + 3b) rudorffites (e.g., AgBiI 4 and Ag 2 BiI 5 ) and Sb 2 S 3 , which have i-SLMEs in the region of 50-60%. Both classes of materials have shown very promising efficiencies to date under 1-sun illumination. [79][80][81] We note that Sb 2 S 3 is limited by the formation of self-trapped excitons, which leads to unavoidable losses at room temperature and limits the open-circuit voltage. Nevertheless, Sb 2 S 3 solar cells have achieved 6% PCE under 1-sun illumination, with an estimated 1-sun efficiency limit of 16%. [82,83] Given their close match with the indoor light spectrum, Sb 2 S 3 devices are very likely to achieve IPV efficiencies well above 10%. Therefore, coupled with its stable performance in ambient air, Sb 2 S 3 is worth future investigations for IPV. [83] IPV efficiencies higher than a-Si:H could also be expected from Rb 3 Sb 2 I 9 , which, just like BiOI and Cs 3 Sb 2 Cl x I 9-x , is a layered compound with high mobility planes that allow photogenerated charges to be efficiently extracted. [32,50] MA 3 Sb 2 I 9 is another potential material. Although the i-SLMEs of the 0D compound are low, recent work has shown that the 2D phase can be stabilized, leading to a near-direct bandgap (which allows higher absorption coefficients) and a lower bandgap value of 2.1 eV (which is closer to the optimum for IPV). [84] InI is another material with high i-SLMEs of 53%. However, InI devices have demonstrated relatively poor 1-sun photovoltaic efficiencies of ≈0.4%. [85] Furthermore, InI has been reported to oxidize easily due to the relative instability of In + compared to In 3+ , and this will likely limit its potential applications in both outdoor and indoor photovoltaics. This limitation is not present in the materials discussed earlier based on Bi 3+ and Sb 3+ , which are in their most stable oxidation states. [83,86,87]

Conclusion
In this work, we explored, for the first time, the capability and potential of lead-free perovskite-inspired absorbers for indoor photovoltaics. We examined in detail two exemplar PIMs, BiOI and Cs 3 Sb 2 Cl x I 9-x . These materials are photostable and have bandgaps close to 1.9 eV, which is too wide for efficiently harvesting solar radiation in single-junction devices, but is optimal for indoor light harvesting. Whereas the PCEs were only approximately 1% under 1-sun illumination, the same devices reached PCEs of 4-5% under FL and WLED lighting, with peak values of 4.4% for BiOI devices and 4.9% for Cs 3 Sb 2 Cl x I 9-x devices under FL lighting. This is already comparable to the IPV performance of industry-standard a-Si:H. We showed that our mm-scale PIM IPVs had V OC values that were sufficient for powering printed thin-film-transistor electronics, demonstrating the potential of such IPVs to power indoor IoT devices. Critically, our loss analyses showed that our PIM IPVs can still achieve further increases in efficiency. Moreover, the efficiency limits calculated based on their absorption coefficients show that these materials have significant potential for exceeding the performance of industry-standard a-Si:H IPV. More broadly, we calculated the spectroscopically-limited maximum efficiency of the most common PIMs explored in the wider field that have bandgaps between 1.45 and 2.3 eV. From this analysis, we identified that the silver-bismuth-iodide rudorffites and Sb 2 S 3 , and possibly also Rb 3 Sb 2 I 9 and MA 3 Sb 2 I 9 (particularly the 2D polymorph), are also very promising for IPV applications. Overall, this work encourages researchers to regard IPV as a key target area in the future exploration of perovskite-inspired materials. In particular, the attractive optoelectronic properties and lowtoxicity profile of many wide-bandgap PIMs provide a formidable opportunity for the development of nontoxic, stable and efficient IPV that can sustainably power the rapidly growing IoT device ecosystem.

Experimental Section
Chemicals were used as purchased without additional purification. CsI sources at 1000 lux had an irradiance of 340 and 320 µW cm −2 , respectively. The accuracy of the light meter and power meter is the primary determinant of the uncertainty in the measured PCE values, which amounts to approximately ±6-7%, as can be determined from a straightforward uncertainty analysis based on the device testing conditions and the specifications of the power and light meters. Consequently, the PCE values are reported herein with 1 decimal place.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. The raw data can be found from https://doi.org/ 10.14469/hpc/7521.