Indoor photovoltaic energy harvesting based on semiconducting π-conjugated polymers and oligomeric materials toward future IoT applications

With the advancement of artificial intelligence computing systems that can collect, analyze, and utilize metadata from our activities and surrounding environments, establishing self-powered electronic systems/networks supported by energy harvesters is strongly desired. With the lowering of power consumption in contemporary IoT electronics such as wireless sensors, indoor organic photovoltaic devices (iOPVs), which can be driven under ambient indoor light, have recently attracted significant interest as self-sustainable eco-friendly power sources. iOPVs based on organic semiconductors have unique advantages, such as light weight, flexibility, solution processability, and feasibility of low-temperature mass production. Additionally, the spectral tunability and high optical absorptivity of organic semiconductors make iOPVs more effective as energy harvesters in indoor lighting environments. With recent intensive research effort, iOPVs have realized the delivery of high power conversion efficiencies exceeding 25% with output power densities of several tens to a hundred μW cm−2, which are sufficient to drive various low-power electronics compatible with the IoT. This review article focuses on recent progress in iOPVs based on π-conjugated polymers and oligomeric materials and outlines their fundamental principles and characterization techniques. With recent advancements in the Internet of Things (IoT), indoor organic photovoltaic devices (iOPVs) have attracted increasing attention because of their potential utility as self-sustainable, eco-friendly power sources. This review highlights emerging iOPV technologies based on π-conjugated polymers and oligomeric materials and outlines their fundamental principles and characterization techniques.


Introduction
Electronic devices have become indispensable tools for modern life and society. With the emergence of novel lowpower electronics such as wireless sensors and portable/ wearable devices [1][2][3], the Internet of Things (IoT) is rapidly developing and spreading. In this circumstance, combined with artificial intelligence (AI) computing systems that enable metadata analysis, the exponential growth of off-grid smart sensor networks for collecting big data is desirable. In the next decade, it is expected that billions to trillions of wireless sensors will be installed to support our lives, industry, commerce, logistics, and infrastructure.
Currently, most IoT devices are powered by batteries, which require periodic replacement and maintenance owing to their relatively short lifespans as power sources. Considering advanced next-generation IoT ecosystems, powering a huge number of IoT devices solely from batteries would not be practically sustainable from environmental, resource, safety, and cost perspectives [4]. Energy harvesting technologies that collect unused energy from sources such as ambient light (photovoltaic harvesting) [5][6][7], heat (thermoelectric harvesting) [8,9], and mechanical vibration/movement (piezoelectric and triboelectric harvesting) [10] have recently attracted significant interest as key elements for solving these problems and constructing self-sustainable IoT ecosystems.
Photovoltaic devices (PVs) are widely used as solar cells in outdoor applications [11]. PVs are also capable of generating power (even though relatively low power) by harvesting artificial indoor light. Although sunlight is not available in all locations and at all times, ambient indoor lighting is always available to supply energy for operating low-power IoT devices. Unlike robust inorganic siliconbased PVs, organic PVs (OPVs) based on organic semiconductors have various inherent advantages such as light weight, flexibility, solution processability, and costeffective large-area manufacturing capabilities [12][13][14][15][16]. Moreover, OPVs can convert weak indoor light into electricity more efficiently than silicon-based PVs because of their spectral tunability and higher optical absorptivity, as well as the lower leakage currents of organic semiconductors. These features endow OPVs with great potential for application in versatile distributed power sources that can be implemented anywhere with any shape and size compatible with the IoT. The power conversion efficiencies (PCEs) of OPVs as solar cells have recently reached over 18% (for single-junction cells) under 1-sun conditions (AM 1.5 G, 100 mW cm −2 ) [11, 17,18]. However, it should be noted that high-performance OPVs under outdoor 1-sun conditions do not necessarily deliver high performance under dim-light indoor conditions. In other words, developing indoor OPVs (iOPVs) requires design guidelines different from those of conventional solar cells from the perspective of both materials chemistry and device physics [19][20][21][22][23][24][25][26][27].
The critical differences between indoor and outdoor PV systems are the emission spectra and power densities of the light sources available to them. In contrast to outdoor sunlight, for which the spectrum is widely distributed from the ultraviolet (UV) to the infrared (IR) region over the range of 280-3000 nm, the emission spectra of indoor lighting, as represented by fluorescent lamps (FLs) and white light-emitting diodes (LEDs), are typically limited to the visible range of 400-750 nm (Fig. 1a). Therefore, to achieve superior energy harvesting functionality, organic semiconductors with optical responses (absorptions) well matched spectrally with indoor light sources should be more favorable for iOPVs than the broad photoabsorption spectra preferred for outdoor solar cells. Moreover, the incident power densities of these indoor light sources (typically 0.1-1 mW cm −2 ) are approximately two to three orders of magnitude lower than that of sunlight (AM 1.5 G, 100 mW cm −2 ), so that the charge carrier densities are significantly reduced in iOPVs. In this case, Shockley-Read-Hall (SRH) recombination (also called trap-assisted recombination) via carrier trapping can be pronounced, which causes significant voltage loss. Over the past decade, several studies have been conducted to enhance the performance of iOPVs by considering these design guidelines. To date, PCEs of state-of-theart iOPVs exceed 25% under white LED illumination [28,29], delivering high output power densities of up to 110 μW cm −2 (at 1200 lx), which are sufficiently high to drive certain low-power electronic devices (Fig. 1b).
In this review, we introduce the basic principles of iOPVs and discuss their characterization methods along with key strategies for achieving superior indoor photovoltaic performance. Subsequently, we outline recent advances in iOPVs from the perspectives of materials chemistry and device engineering.

Equivalent circuits and photovoltaic parameters
The performance of an iOPV is generally evaluated by PCE, that is, PCE (%) = P out /P in × 100, where P out (mW cm −2 ) is the output power density of the device and P in Fig. 1 a Comparison of emission spectra for various light sources: sunlight (AM 1.5 G), halogen lamp, fluorescent lamp (FL) and lightemitting diode (LED). b Power consumption by various IoT electronic devices and related energy harvesting capabilities of iOPVs [35] (mW cm −2 ) is the incident light intensity (irradiance). P out can be expressed as P out = V oc × J sc × FF, where V oc (V) is the open-circuit voltage, J sc (mA cm −2 ) is the short-circuit current density, and FF (%) is the fill factor. These parameters are extracted from the current density-voltage (J-V) curves measured under different P in conditions. As schematically shown in Fig. 2a, the equivalent circuit model for iOPVs consists of a parallel-connected current source, diode, shunt resistance (R sh ), and an additional series resistance (R s ). While R s is mainly attributed to the resistance of the bulk active layer and the interfacial contact resistance between the active layer and electrodes, R sh originates from the leakage currents in the cell. Under dark conditions, the J-V characteristics of iOPVs represent the exponential response of a diode with a high J in the forward bias and low J in the reverse bias. Light illumination on iOPVs generates a photocurrent (J ph ); therefore, their J-V curves upon illumination are ideally the superposition of the dark and photocurrent characteristics (Fig. 2b). Accordingly, the J-V characteristics of an ideal device can be expressed using the Shockley diode equation, as follows (Eq. 1) [30][31][32][33][34]: where J 0 is the reverse saturation current density, q is the elementary charge, n is the ideality factor of the diode, k is the Boltzmann constant, and T is the absolute temperature. When n = 1 (i.e., an ideal diode), carrier recombination in the depletion region is zero (or negligible), and the diffusion current is dominant. In contrast, when n = 2, carrier recombination dominates the dark current. By solving Eq. 1 under short-circuit (V = 0) and opencircuit (J = 0) conditions, J sc and V oc are given by Eqs. 2, 3, respectively. Assuming an ideal device, i.e., J 0 and R s are sufficiently small and J sc and R sh are sufficiently high, J sc and V oc can be expressed by the simple formulas shown on the right-hand side of each equation.
Based on Eq. 1, the key parameters R s , R sh , n, J 0 , and J ph can be extracted from the measured J-V curves (Fig. 2). R s and R sh are readily determined from the inverse slopes (∂V/ ∂J) of the forward and reverse J-V curves, respectively, whereas n and J 0 can be roughly estimated from the slope and intercept, respectively, of the linear portion of the dark ln(|J|)-V curve in the forward bias. Finally, J ph can be deduced using the extracted parameters (R s , R sh , n, and J 0 ), according to Eq. 2. The accuracy of J-V curve fitting and analysis can be verified from the integrity of the J ph and J sc values. For actual iOPVs, the incident light intensity (irradiance) has a significant effect on these parameters. Unlike the situation for sunlight, iOPVs exhibit peculiar behavior, especially under dim-light illumination (low P in ) conditions; therefore, an accurate and detailed analysis is needed, as is discussed later. V oc attenuation under dim-light indoor conditions is a critical issue that causes PCE reductions in iOPVs [29,35]. According to Eq. 3, V oc can be maintained high by sufficiently reducing J 0 and increasing R sh . However, under dimlight conditions, the resulting photocurrent (J ph ) is inevitably small because of the low carrier density. In this case, the Fig. 2 a Equivalent circuits for iOPVs under dark (left) and light illumination (right) conditions. b Relationship between key parameters (R s , R sh , n, J 0 , and J ph ) and J-V curves of iOPVs with a different load (R s ) measured under dark (left) and light illumination (right) conditions. J d = current density flowing through the diode, J sh = current density flowing through the shunt resistance, J ph = photocurrent density generated from photodiode, J 0 = reverse saturation current density, n = ideality factor, k = Boltzmann constant, q = elementary charge, T = absolute temperature, R s = series resistance, R sh = shunt resistance contributions of the dark and leakage currents (J 0 and J sh ) become non-negligible and cause significant V oc attenuation. Figure 3 shows the typical V oc -J ph relationships simulated for iOPVs with different J 0 values. A steep reduction in V oc with decreasing J ph is universally observed for all J 0 conditions (10 −7 -10 1 mA cm −2 ). In the high-J ph regime satisfying V oc « J ph R sh , V oc exhibits a quasi-linear dependence on ln(J ph ) and satisfies the relationship V oc = (nkT/q)ln(J ph /J 0 ), as in the case of conventional solar cells. However, in the low-J ph regime with V oc » J ph R sh , the degree of V oc attenuation becomes more pronounced and follows the relationship V oc = (nkT/q) ln{1 + (J ph /J 0 )(1 − (V oc /J ph R sh )}. The inflection points of the V oc attenuation curves are found to be susceptible to J 0 , suggesting that a device with a lower J 0 can maintain a higher V oc , even under dimly lit conditions. Thus, modulating these parameters can effectively suppress V oc attenuation under indoor conditions, leading to high PCEs over a wide illuminance range for iOPVs. It can be concluded that a higher V oc under AM 1.5 G and a smaller voltage loss under indoor conditions are crucial for iOPVs.
The parasitic resistances (R s and R sh ) also have a large impact on the resulting FFs of iOPVs [36][37][38]. It is evident that the largest possible R sh (ideally ∞ ) and the smallest possible R s (ideally 0) are preferred to enhance the FF and thereby maximize P out . Figure 4 shows the simulated J-V characteristics of iOPVs with different R s and R sh (R s /R sh ratios). Under indoor conditions, R s and R sh show effects that are quite different from those found under outdoor conditions. Evidently, under high illuminance (P in = 100 mW cm −2 , Fig. 4a), the photovoltaic characteristics are mainly governed by R s , and the impact of R sh is very limited. In contrast, under low illuminance (P in = 1 mW cm −2 , Fig. 4b), R sh becomes more crucial for indoor photovoltaic performance, and R s no longer has a substantial impact on the overall J-V characteristics. This is because the trap-assisted recombination associated with R sh becomes dominant in the carrier recombination processes at low carrier densities. Therefore, increasing R sh by suppressing trapassisted recombination is vital for achieving high FFs and thereby high PCEs for iOPVs. These basic design principles are quite different from the guidelines established for conventional solar cells.

Evaluation of iOPV efficiency
The PCEs of iOPVs are strongly dependent on the type of indoor light source (i.e., emission spectra and illuminance). To determine their PCEs accurately, we need to select a suitable light source and evaluate its power density (P in ). The 1-sun condition (P in = 100 mW cm −2 ) with a simulated AM 1.5 G spectrum (Fig. 1a) is the global standard used for characterizing outdoor PVs. However, neither indoor light sources nor a standard protocol have been specified for evaluating iOPVs. This is because, unlike the sun, we use a variety of light sources, including LEDs and FLs, in indoor environments. As shown in Fig. 1a, there are significant differences in the emission spectra of artificial indoor light sources. Accordingly, establishing and standardizing appropriate evaluation methodologies for iOPVs are current challenges [39,40].
In general, indoor light is measured as photometric illuminance L (lx), which is typically in the range of 100-1000 lx. Theoretically, L and P in of an arbitrary light source can be calculated by integrating the spectral irradiance F(λ) (mW cm −2 nm −1 ) using Eqs. 4 and 5: where K m is a constant representing the maximum luminous efficiency (K m = 683 lm W −1 ) and V(λ) denotes the spectral luminous efficiency function for human photopic vision. In practice, L can be measured using an illuminance spectrophotometer. Figure 5a depicts the typical irradiance-illuminance (P in -L) relationships of various indoor light sources. It should be noted that different light sources afford different P in values, even at the same L, because of the different matching levels between the light-source emission spectra and human eye photopic sensitivity. Under a low L of 200 lx with warm-white LED illumination (3000 K), which corresponds to living environments, P in is~60 μW cm −2 , whereas P in increases to~300 μW cm −2 under a high L of 1000 lx. However, these incident P in values are less than 0.1% and 1%, respectively, of outdoor 1-sun illumination. Note also that, even if L is the same, if the lightsource color temperature is different, P in differs substantially. Therefore, to evaluate iOPV efficiency, not only the illuminance of the incident light but also its spectral irradiance (or P in ) and color temperature should be clarified as a prerequisite. According to recent research, it is recommended that iOPV efficiencies should be evaluated by using a white LED as standard indoor lighting by specifying its color temperature and spectral irradiance. What is the limit of the iOPV efficiency? The maximum PCEs of iOPVs can be assessed based on the Shockley-Queisser (SQ) limits [41] for various light sources, with the assumption that photon energies above the bandgap (E g ) of the material are all absorbed and charge carriers are fully extracted. Theoretical simulations reveal that the optimal E g of iOPVs for white LED illumination is 1.7-1.9 eV (ca. 650-730 nm) [39,40,42,43]. As shown in Fig. 5b, for AM 1.5 G illumination, it is well known that the SQ-limited PCE is 33%, as the photovoltaic materials have an E g of 1.1-1.3 eV. According to the best research-cell efficiencies chart provided by the National Renewable Energy Laboratory (NREL) [11], crystalline silicon (c-Si) solar cells with E g of~1.1 eV have achieved significantly high PCEs of up to 27.6%, which approach the SQ limit for 1-sun illumination. However, unlike these outdoor conditions, photovoltaic materials with a much wider E g are more suitable for indoor conditions, leading to ideal SQ-limited PCEs as high as~57%. This is primarily because the emission spectra of indoor light sources are relatively narrow compared to sunlight, so the irradiated photons can be more effectively utilized by organic semiconductors while suppressing thermalization and transparency losses. The best PCEs of iOPVs reported thus far are~30% under illumination with a warm-white LED (3000 K) [29]. There is still room for improvement in terms of both the materials and devices for future IoT applications.

Recent advances in iOPVs
To achieve high-efficiency iOPVs, the design of appropriate organic semiconductor materials that play the core role in . Reproduced from reference [40] with permission from Royal Society of Chemistry the active layer is of primary importance. Although iOPVs are a class of recently emerging energy-harvesting devices, we can take advantage of material technologies for organic solar cells that have been accumulated over the last two decades [12][13][14][15][16][17][18]. However, as described in Section "Basic principles and characterizations for iOPVs", the design guidelines for iOPVs differ from those for conventional solar cells in several respects. In particular, spectral matching with indoor light sources and suppression of V oc reduction under low illuminance are key factors. For indoor applications, the optimal E g for iOPV materials is comparatively large (1.7-1.9 eV), as only photons in the visible region need to be harvested.

Polymer donor-fullerene acceptor binary systems
The initial iOPVs reported in the 2010s were based on polymer:fullerene BHJ systems, in which a classical semiconducting polymer, regioregular poly(3-hexylthiophene) (P3HT, Fig. 6), was used in combination with the soluble fullerene derivatives PC 61 BM and PC 71 BM (Fig. 7) [38,49,53]. However, their indoor photovoltaic performance was unattractive because of their low V oc (<0.5 V) and low PCEs (<9%). To enhance the V oc of iOPVs, an indene-C 60 bisadduct (ICBA) with a shallower lowest unoccupied molecular orbital (LUMO) energy level (E LUMO = −3.74 eV) was subsequently applied as an acceptor instead of PC 61 BM (E LUMO = −3.90 eV) [53,56,58]. The larger energy offset between the highest occupied molecular orbital (HOMO) of P3HT and the LUMO of ICBA resulted in a V oc higher than that of the P3HT:PCBM system. Consequently, the P3HT:ICBAbased iOPVs achieved an improved PCE of 13% with a higher V oc of 0.73 V under LED illumination (500 lx).
The search for high-performance semiconducting polymers has continuously improved the performance of fullerene-based iOPVs. In 2015, Mori et al. reported the indoor photovoltaic properties of a PTB7-Th:PC 71 BM system [48]. Under LED illumination (890 lx), the PTB7-Th:PC 71 BM-based iOPV exhibited a higher PCE (11.6%) than the c-Si cell (9.6%) because of its higher V oc (0.62 V) than that of the c-Si cell (0.43 V). The higher V oc and wellmatched absorption spectrum are advantages of iOPVs over Si-based cells for indoor energy harvesting. Subsequent studies by other research groups further improved the PCEs of PTB7-Th:PC 71 BM-based iOPVs to~16% [53,65].
In 2016, Lee et al. compared the indoor photovoltaic properties of P3HT, PTB7, and PCDTBT ( Fig. 6) as polymer donors in combination with PCBM [49]. The PCDTBT:PC 71 BM-based iOPV exhibited the best performance among them, affording a high PCE of 16.6% with a high V oc of 0.72 V under FL illumination (300 lx). Although the PTB7:PC 71 BM-based device showed better performance than the PCDTBT:PC 71 BM-based device under AM 1.5 G, the indoor PCE tended to be lower (14.6%) because of its narrower E g . Yin et al. subsequently reported comparably high indoor PCEs of up to 18.7% with the same PCDTBT:PC 71 BM system under a white LED (300 lx) [55,57]. Inspired by the success of PCDTBT, since 2019, various benzothiadiazolecontaining wide-bandgap donor polymers, such as PPDT2FBT [59,64], PDTBTBz-2F [61], and P3TEA [66], have been tested in combination with PCBM for iOPVs, and they have exhibited high PCEs under LED or FL illumination (Table 1). In particular, the PDTBTBz-2F:PC 71 BM-based iOPVs achieved a notably high PCE of 23.1% with V oc of 0.82 V, generating P out as high as 65 μW cm −2 under LED illumination (1000 lx) [61], which is the best indoor photovoltaic performance among reported fullerene-based iOPVs. Importantly, the PDTBTBz-2F:PC 71 BM-based device afforded a PCE as low as 6.9% under 1-sun conditions, suggesting that materials unsuitable for outdoor solar cells may be more suitable for indoor energy harvesting. It is also worth mentioning that a relatively thick photoactive layer can be more suitable for iOPVs [50,57,59,66,67]. For the PPDT2FBT:PC 71 BM-based iOPVs, even devices with thick BHJ photoactive layers (390-870 nm) demonstrated excellent indoor photovoltaic performance with PCEs of 16.0-12.5% [59]. According to Eq. 2, J sc is inversely proportional to 1 + R s /R sh . Hence, under dim-light indoor conditions, R s /R sh is sufficiently small, and J sc becomes tolerant to the active layer thickness.
Among the renowned donor polymers developed for outdoor solar cells, PBDB-T and its halogenated derivatives ( Fig. 6) have demonstrated the highest PCEs, especially when combined with NFAs under 1-sun conditions. Solar cells employing PBDB-TF (alias PM6) as the donor delivered record-high PCEs of up to 18% [19]. In 2019, Hou and co-workers reported an impressive indoor photovoltaic performance for the PBDB-TF:PC 71 BM system in 1 cm 2 cells, achieving a high PCE of 18.1% with V oc of 0.78 V under a 2700 K LED (1000 lx) [62]. By replacing the benzodithiophene dione (BDD) subunit in PBDB-TF with a quinoxaline derivative, Lee and co-workers developed new donor polymers, WF3 and WF3F [63]. While the E g values of these two polymers are similar in the range of 1.73-1.78 eV, WF3F has a deeper HOMO can be significantly altered by a slight difference in the substituents. More recently, PTQ10 [94], which has a reputation as a donor polymer for outdoor solar cells, was reported to exhibit good performance in iOPVs [67]. The PTQ10:PC 61 BM-based iOPVs delivered a high PCE approaching 20% with a V oc of 0.79 V under LED (500 lx). The lower synthetic complexity of

Small-molecule donor-fullerene acceptor binary systems
Another category of iOPV systems is based on a combination of SM donors and fullerene acceptors. Unlike polymer-based systems, this material class has some inherent advantages, such as well-defined molecular structures, monodispersity, high purity, negligible batch-to-batch variations, and energy-level tunability. Nevertheless, the development of SM donors for iOPVs has lagged behind that for their polymer counterparts. Excellent indoor photovoltaic characteristics have been achieved for some SM donor-fullerene acceptor BHJ systems, as described in this section.
In  Acceptor Light source Ref.
Polymer donor-fullerene acceptor binary system L illuminance of incident light, P in input power density (irradiance), PCE power conversion efficiency, V oc open-circuit voltage, J sc short-circuit current density, FF fill factor, P out output power density In 2019, Yasuda and co-workers reported promising indoor photovoltaic functionality for the new SM donors, BDT-1T-ID and BDT-2T-ID (Fig. 6) [60,96], which possess ideal E g s of 1.7-1.8 eV and deep-lying HOMOs of −5.23 and −5.13 eV, respectively. The BDT-1T-ID:PNPand BDT-2T-ID:PNP-based devices demonstrated higher PCEs of 12.4% and 16.0% (with V oc s of 0.84 and 0.75 V, respectively) under LED (200 lx) compared to those measured under 1-sun illumination (PCE = 3.6% and 5.8%, respectively). Furthermore, the BDT-2T-ID:PNPbased device demonstrated stable PCEs over a wide illuminance range of 200-10000 lx, retaining high values exceeding 16%. These almost illuminance-independent PCE characteristics make SM-based iOPVs more attractive and reliable for practical applications. Indeed, by adopting the BDT-2T-ID:PNP system, six seriesconnected flexible iOPV modules were successfully produced, which are capable of generating a high output power over 100 μW and high V oc of 4.2 V even under 200 lx dim-light conditions (see Section "iOPV modules for practical applications" for details). This achievement opens a new avenue for preparing dispersive selfsustainable electric power sources.
Following a design strategy aimed at HOMO deepening, the same group subsequently developed new SM donors, 1DTP-ID and 2DTP-ID, by incorporating electronwithdrawing π-fused lactam (DTP) subunits [35]. 1DTP-ID and 2DTP-ID have E g s of 1.65 and 1.70 eV and E HOMO s of −5.25 and −5.35 eV, respectively, which are well suited for iOPVs. Under a warm-white LED (2900 K, 200 lx), the 1DTP-ID:PNP-and 2DTP-ID:PNP-based iOPVs achieved PCEs as high as 19.3% and 17.8%, with V oc s of 0.67 and 0.72 V, respectively. Additionally, important insight into the color-temperature dependence of the LED sources was obtained with this iOPV system (Fig. 8). Artificial white LED power spectra typically comprise sharp blue emission from GaN/InGaN and broad yellow emission from phosphors (Ce:YAG). Cool-white LED light with a high color temperature (~7000 K) has a relatively high ratio of blue light intensity, whereas warm-white LED light with a low color temperature (~3000 K) is dominated by long wavelength emissions (Fig. 8a). Such differences in the spectral irradiance characteristics of the LED sources significantly affect spectral matching with the organic photoactive layers in iOPVs, which causes variations in the resulting PCEs and P out values. As shown by the incident photon-to-electron conversion efficiency (IPCE) spectra (Fig. 8b), the photoactive layers exhibited the most efficient photoresponses in the range of 500-700 nm. Therefore, the corresponding devices showed a strong tendency to exhibit higher PCEs L illuminance of incident light, P in input power density (irradiance), PCE power conversion efficiency, V oc open-circuit voltage, J sc short-circuit current density, FF fill factor, P out output power density under warm-white LED illumination ( Fig. 8d and Table 1), where spectral matching was more favorable. As mentioned earlier, to properly evaluate iOPVs, it is crucial to clarify the color temperature, emission spectrum, and irradiance of the indoor light source. To avoid overestimating and underestimating the intrinsic indoor photovoltaic performance, it is preferable to evaluate iOPVs using a standard natural-white LED with a color temperature of approximately 5000 K, which corresponds to common office and store lighting.

Polymer donor-NFA binary systems
Over the past few years, the performance of fullerene-based iOPVs has steadily improved, with PCEs exceeding 20%, and practical iOPV modules have been successfully developed. However, fullerene-based iOPVs have limited potential for further enhancement of their PCEs because the limited energy-level tunability and visible absorptivity of fullerene acceptors make it difficult to boost V oc and J sc . In contrast, NFAs [44][45][46][47] offer opportunities to improve V oc and J sc owing to their adjustable (designable) energy levels, optical E g , and absorptivity. Since 2019, many research groups have reported the indoor photovoltaic properties of polymer donor-NFA binary systems [28,29,62,64,66,, the performance data for which are summarized in Table 2. ITIC [97] and its derivatives (Fig. 7), which are based on the electron-donating dithieno[2,3-d:2′,3′-d′]-s-indaceno[1,2b:5,6-b′]thiophene (DTIDT) π-core end-capped with electron-accepting 3-dicyanomethylene-1-indanone groups, are renowned NFAs commonly used in organic solar cells. Owing to the steric hindrance of the side chains, excessive self-aggregation of the seven-ring-fused donor moieties can be suppressed to form suitable nanosegregated BHJ structures. In 2019, Hou and co-workers reported the indoor photovoltaic performance of representative NFAs, IT-4F and ITCC, in combination with PBDB-TF as a polymer donor [62]. . This result is mainly attributable to the better spectral matching and higher V oc of the PBDB-TF:ITCC system. As shown in Fig. 9, ITCC has a wider optical bandgap (E g = 1.67 eV) along with shallower LUMO level (E LUMO = −3.76 eV) compared to IT-4F (E g = 1.50 eV and E LUMO = −4.14 eV), thus leading to both better spectral matching with warm-white LED lighting and increase of V oc in iOPVs. The same group specifically developed a new wider-bandgap NFA for iOPVs, named IO-4Cl (Fig. 7), by employing 5,6-dichloro-indene-1,3-dione as the end-capped group [28]. By combining IO-4Cl (E g = 1.80 eV) with PBDB-TF (E g = 1.82 eV), the BHJ system exhibited optimal absorption that matched the emission spectrum of a warmwhite LED. Under AM 1.5 G conditions, the PBDB-TF:IO-4Cl-based device afforded a PCE of 9.8% with an exceptionally high V oc of 1.24 V along with a small energy loss (E loss = E g − qV oc ) of <0.6 eV. Reflecting this feature, high V oc s of 1.03-1.10 V were maintained under LED illumination (2700 K, 200-1000 lx), thereby resulting in notably high PCEs of 22.2-26.1%. Moreover, the PBDB-TF:IO-4Clbased iOPV maintained its initial PCE for 1000 h under continuous LED illumination, demonstrating great stability for practical applications.
Although the narrow-bandgap IT-4F would not be an optimal NFA for iOPVs, as described above, reasonably high PCEs can be obtained by combining it with more appropriate donor polymers. In 2019, Son and co-workers developed a new donor polymer, PBDB-TSCl (Fig. 6), by modifying PBDB-TF to better match IT-4F [72,73]. PBDB-TSCl has a slightly deeper E HOMO (−5.61 eV) than PBDB-TF, resulting in an increase in V oc . The PBDB-TSCl:IT-4F-based iOPV exhibited a higher PCE (21.5%) than the corresponding PBDB-TF:IT-4F-based device (15.6%) under 500 lx FL illumination. Additionally, chlorinated PBDB-TSCl presented higher morphological stability than fluorinated PBDB-TF under thermal stress; after heating at 100°C for 34 h, the PBDB-TSCl-based device lost only 5% of its initial PCE, while the PCE of the Reproduced from reference [35] with permission from Royal Society of Chemistry PBDB-TF-based device was reduced by 25%. At approximately the same time, Facchetti and co-workers reported a new medium-bandgap donor polymer, TPD-3F, for iOPVs [74]. Incorporating an electron-accepting thieno [3,4-c]pyrrole-4,6-dione (TPD) unit renders TPD-3F with a deeper E HOMO of −5.62 eV and wider E g of 1.90 eV compared to those of PBDB-TF. Consequently, the TPD-3F:IT-4F-based iOPV demonstrated a PCE as high as 26.2% under FL illumination.
In 2020, Yan and co-workers reported high-efficiency iOPVs using PBDB-TF (alias PM6) as the donor in combination with a new NFA, Y6-O [29]. Y6-O was designed by replacing the two alkyl groups of Y6 with alkoxy groups and exhibits upshifted HOMO and LUMO levels as well as blue-shifted absorption compared to Y6 [98]. The PM6:Y6-O-based iOPVs incorporating PDI-NO as an electron transport layer (ETL) achieved record-high PCEs of over 30% with a high V oc of 0.83 V and a FF of 76% under LED illumination (3000 K, 1200 lx). As shown in Fig. 10, the choice of ETL is critically important for device performance. Both the PM6:Y6-O-based devices with PDI-NO and PFN as the ETL showed very similar performance under 1-sun (AM 1.5 G) conditions (Fig. 10b). However, under indoor conditions, the PDI-NO-containing device exhibited better performance than the PFN-containing device over a wide illuminance range (170-1650 lx). This difference can be explained by the work function (WF) shift at the ETL/Al interface. Upon Al modification with PDI-NO (E HOMO = −6.21 eV), the WF downshifted from 4.22 to 3.97 eV, while it shifted from 4.22 to 4.12 eV with PFN (E HOMO = −5.61 eV). Although both ETLs can reduce the WF of Al to form ohmic contacts with the acceptor and improve electron extraction, the energy difference (Δh) between the E HOMO of the ETL and the Fermi level (E F ) of Al, which represents the hole-blocking barrier, is clearly different for the respective devices (2.31 eV for PDI-NO and 1.49 eV for PFN). Compared with the PFN-containing device, the PDI-NO-containing device showed a lower leakage current and higher R sh (similar R s ) because of its good hole-blocking ability (large Δh), thus leading to higher V oc and FF, even under dim-light conditions (Fig. 10c). In this way, not only optimization of the donor-acceptor combinations but also interfacial engineering is of vital importance for improving iOPV performance.
In seeking to enlarge the optical E g and upshift the E LUMO of Y6, in 2021, the same group developed the medium-bandgap NFAs, LBT-DF and LBT-SCl (Fig. 7), by replacing the fused thieno [3,2-b]thiophene subunit of Y6 with benzothiophene [86]. The E g s of LBT-DF and LBT-SCl are 1.64 and 1.62 eV, respectively, and therefore, their absorption spectra, which cover 500-750 nm, are more compatible with indoor lighting than that of Y6. iOPVs based on S2:LBT-DF and S2:LBT-SCl blends achieved high PCEs of 23.0-25.7% and 23.6-27.3%, respectively, under a 2600 K LED with an illuminance range of 500-2000 lx. The same group also developed FCC-Cl and its analogs based on a fluorenedicyclopentathiophene (FDCT) π-core as new medium-bandgap NFAs for iOPVs [85,92]. FCC-Cl with an E g of 1.71 eV was combined with two widely used donor polymers (PBDB-TF and D18), and both BHJ systems showed rather high PCEs of 25-30% under a 2600 K LED at 100-2000 lx.
Various new NFAs, which are distinct from the fusedring ITIC and Y series NFAs, have recently been developed and utilized for iOPVs. In 2019, Liu and co-worker reported that polymer acceptors (PBN series, Fig. 7) featuring boron-nitrogen coordination (B ← N) bonds exhibited excellent indoor photovoltaic properties in combination with a wide-bandgap polymer, CD1, as the donor [75,83,89]. Benefitting from the medium optical bandgap of PBN-10 (E g = 1.95 eV) and good energy-level alignment, all-polymer iOPVs based on the CD1:PBN-10 system exhibited high PCEs of 26.2% and 21.7% with a markedly high V oc of 1.14 V under 1000 lx FL and LED illumination, respectively [75] (Table 2). To date, this is the highest V oc value reported for iOPVs under dim-light conditions, indicating the high potential of all-polymer iOPVs. Meanwhile, for the CD1:ITIC-based device, a large E LUMO offset between CD1 and ITIC resulted in a much lower V oc of 0.77 V and thereby lower PCE of 15.4%.
Another choice for NAFs is a perylene diimide (PDI)based material. In 2019, Welch and co-workers designed tPDI 2 N-EH (Fig. 7) as a PDI-based NFA and compared its performance with those of PC 61 BM, ITIC-M, and IT-4F in iOPVs [64]. tPDI 2 N-EH has a relatively wide E g (2.22 eV) with photoabsorption in the range of 400-600 nm, which is complementary to the photoabsorption of the donor polymer PPDT2FBT. The PPDT2FBT:tPDI 2 N-EH-based device exhibited a low PCE of 6.5%, despite its high V oc of over 1 V under 1-sun illumination. However, its PCE increased to 9.0% under LED (5600 K, 300 lx), which is higher than those of the iOPVs using ITIC-M and IT-4F as NFAs. It is Fig. 9 Energy-level diagram for representative nonfullerene acceptors. E g = optical bandgap Indoor photovoltaic energy harvesting based on semiconducting π-conjugated polymers and oligomeric. . . also noteworthy that these iOPVs can be manufactured by slot-die coating using nonhalogenated solvents (o-xylene/ diphenyl ether). Yan and co-workers introduced another PDI-based wide-bandgap NFA, FTTB-PDI4, for iOPVs [29,66]. The P3TEA:FTTB-PDI4 system, which features a near-zero energy offset for the LUMOs, enabled small voltage loss and thereby high V oc . Consequently, the P3TEA:FTTB-PDI4-based iOPVs achieved remarkable PCEs of 23.7-24.2% with high V oc s of 0.94-1.0 V over a wide range of illuminance (170-1650 lx) under a 3000 K LED [66]. Additionally, even with thick active layers (~200 nm), the devices maintained high PCEs of over 21%.
In 2018, So and co-workers demonstrated the first effective ternary BHJ system for iOPVs [55]. By introducing the donor polymer PDTSTPD as the third component into the binary PCDTBT:PC 71 BM system, the PCEs increased from 16.0% (binary) to 18.1% (ternary) under LED illumination (3000 K, 300 lx). Blending of (PDTSTPD) can passivate shallow traps near the band edges of the BHJ and improve the charge transport properties, leading to an improvement in the PCE. Similarly, Lee and co-workers introduced a PDI-based SM acceptor (EP-PDI) as a third component into a binary PTB7:PC 71 BM system [69]. The optimal PTB7:PC 71 BM:EP-PDI-based device showed an improved PCE of 15.4% under LED illumination at 500 lx compared to the binary PTB7:PC 71 BM-based device (PCE = 8.9%).
Ko and co-workers conducted a comparative study of D1:D2:A and D:A1:A2 ternary blends consisting of PBDB-T:PTB7-Th:PC 71 BM and PBDB-T:PC 71 BM:ITIC-Th [71]. illumination. c Illuminance dependence of photovoltaic parameters (V oc , FF, and PCE) for the PM6:Y6-O-based devices with PFN or PDI-NO interlayers. Reproduced from reference [29] with permission from Elsevier Although both systems were universally effective, the D:A1:A2 system incorporating an NFA (ITIC-Th) benefited from less charge recombination and higher PCEs. Consequently, the optimized ternary PBDB-T:PC 71 BM:ITIC-Thbased iOPVs showed better performance, with PCEs of 25.6% and 26.4% under LED illumination at 200 and 1000 lx, respectively. In this system, the third NFA additive, ITIC-Th, self-organized at the PBDB-T:PC 71 BM interface to form effective cascade ternary junctions, resulting in less charge recombination. Moreover, the ternary PBDB-T:PC 71 BM:ITIC-Th-based iOPVs showed superior thermal durability and retained~90% of their initial PCEs after heating at 60°C for 100 h. This feature would be crucial as iOPVs move forward toward practical applications.
Yang and co-workers newly synthesized a non-fused 2,2'-bithiophene-based NFA, Y-Th2, and introduced it into a PBDB-TF:Y6 system to construct ternary iOPVs with cascade energy-level alignment [76]. Owing to its wider optical bandgap (E g = 1.74 eV), Y-Th2 complemented the absorption spectra of PM6 and Y6 in the relatively short wavelength region with high absorptivity. Consequently, impressive indoor photovoltaic performance with PCEs of up to 22.3% was obtained for the ternary PM6:Y6:Y-Th2based iOPV under a solar LED (1000 lx).
In addition, several iOPVs featuring quaternary BHJ systems have been reported. In 2019, Ko and co-workers successfully demonstrated quaternary iOPVs composed of two polymer donors (PBDB-T and PTB7-Th) and two SM acceptors (PC 71 BM and ITIC-Th) [70]. Compared to their corresponding binary counterparts (PBDB-T:PC 71 BM and PTB7-Th:ITIC-Th), the quaternary BHJ system benefitted from broadened spectral response, enhanced charge transport, and improved morphological stability. Consequently, the quaternary PBDB-T:PTB7-Th:PC 71 BM:ITIC-Th-based iOPVs achieved an improved PCE of 14.3% under LED illumination (1000 lx). This quaternary system also presented a semitransparent feature without impairing the high PCEs. More recently, to fabricate high-performance quaternary iOPVs, the same group introduced sequential deposition based on consecutive spin-coating of two binary donor/acceptor blends [93]. To form a desirable quaternary BHJ layer, the first blend solution (PTB7-Th:ITIC-Th) was spin-coated onto a ZnO-coated indium-tin-oxide (ITO) substrate, and then the second blend solution (PM6:PC 71 BM) was promptly spincoated onto the first blend layer. The sequentially processed quaternary iOPV exhibited superior performance with a PCE of 25.0% under LED (3000 K, 1000 lx) compared to the binary PM6:PC 71 BM-based device (PCE = 19.9%).

iOPV modules for practical applications
As outlined in the above sections, the performance of iOPVs has drastically improved in a short period of time and is now reaching practicality as a promising energy harvesting technology. The inherent flexibility, light weight, and applicability to a variety of solution-based manufacturing techniques (such as spin coating, slot-die coating, inkjet printing, and R2R printing) [99] make iOPVs attractive for indoor applications. Development of largearea devices and modules is essential for practical use of the iOPVs. This section briefly introduces the research achievements for state-of-the-art iOPV modules composed of series-connected unit cells, the performance data for which are summarized in Table 3.
In 2016, Lee et al. successfully fabricated iOPV modules (eight cells in series) based on PCDTBT:PC 71 BM with a total active area of 100 cm 2 and achieved a module PCE of 11.2% (with V oc = 4.87 V, I sc = 314 μA, and FF = 61%), generating a maximum power (P out ) of 938 μW under a 300 lx FL illumination (Fig. 11a) [49]. In 2020, a module PCE exceeding 20% was achieved by Facchetti and coworkers [74]. The iOPV modules (five cells in series) based on TPD-3F:IT-4F fabricated by blade coating exhibited an outstanding PCE of 21.8% (with V oc = 3.21 V, I sc = 361 μA, and FF = 71%) under 1000 lx FL (Fig. 11b). The present modules also performed well as solar cells under 1-sun (AM 1.5 G) conditions, yielding a PCE of 10.4% (V oc = 4.45 V, I sc = 78.7 mA, and FF = 61%). The potential for dual use in indoor and outdoor environments is highly attractive and offers great practicality and additional value.
iOPV modules with high mechanical flexibility and bendability are ideal energy harvesters because such devices can be installed anywhere, including curved walls and narrow crevices. High-performance flexible iOPV modules were first demonstrated by Yasuda and co-workers (Fig. 11c, d) [35,60]. They fabricated BDT-2T-ID:PNPbased iOPV modules (six cells in series) with a total active area of 9.5 cm 2 on a rigid glass substrate and on a flexible polyethylene naphthalene (PEN) substrate, and these modules exhibited high PCEs of 20.2% and 18.4%, respectively, delivering P out as high as 111 and 101 μW under LED illumination at 200 lx. The same group also developed similar flexible iOPV modules based on 1DTP-ID:PNP, achieving a comparatively high module PCE of 17.0% (V oc = 4.05 V, I sc = 35.1 μA, and FF = 67%) [35]. In these six series-connected modules, the average value of V oc for each sub cell was consistent with the V oc of a single test cell. Therefore, upscaling and modularization without lowering V oc and FF enable a high P out and PCE under indoor conditions.

Summary and outlook
In this review, we have summarized recent progress in emerging iOPV technologies. Full-fledged iOPV research began a few years ago, and many challenges must be overcome for practical use. Although much of the knowledge accumulated for outdoor organic solar cells can be used as a reference for developing iOPVs, they require different design strategies in terms of both materials chemistry and device physics. This difference primarily originates from the different emission spectra and intensities of the light sources used (i.e., sunlight vs. indoor LED lighting). Unlike solar cells, because only visible light ranging from 400 to 750 nm is available in iOPVs, their photoactive layers based on organic semiconductors should have wider optical bandgaps. Theoretical simulations predict that the optimal bandgap for iOPVs under white LED illumination is 1.7-1.9 eV and the SQ-limited maximum PCE is as high as~57% [40]. Therefore, the key to improving iOPV performance is to design and apply appropriate organic semiconductors that enable good spectral matching with white LEDs as an incident light source.
However, the current problem is that the light sources are not standardized; therefore, establishing appropriate evaluation methodologies for iOPVs is an urgent task. It is crucial to specify the color temperature, emission spectrum, and irradiance of an indoor light source for iOPV measurements. In general, warm-white LED light with a color temperature of approximately 3000 K tends to afford a higher PCE owing to its superior spectral matching with common iOPVs. Because the PCEs and output power densities vary depending on the color temperature of the light source, it is recommended that iOPVs should be evaluated with a standard natural-white LED with a color temperature of approximately 5000 K, thus enabling fair comparisons of the results reported by different groups. A effective active area, L illuminance of incident light, P in input power density (irradiance), PCE power conversion efficiency, V oc open-circuit voltage, I sc short-circuit current, FF fill factor, P out output power.  [60]. Reproduced from references [49,60] with permission from AIP Publishing, Elsevier, and American Chemical Society Additionally, many previous iOPV studies have reported results obtained only at 1000 lx or even higher, which typically leads to higher PCEs than under lower illuminance; however, this cannot be representative of indoor energy harvesters. Hence, measurements under lower illuminance (200-500 lx), in addition to 1000 lx, are recommended to obtain a better illustration of the actual indoor photovoltaic performance. From the device design perspective, the most important point is how to suppress V oc reduction under low illuminance. A key strategy with which to overcome this challenge is to maximize the R sh of the device as much as possible. As discussed in Section "Polymer donor-NFA binary systems", introduction of a hole-blocking/electrontransporting layer is a viable and effective approach to increasing R sh by suppressing carrier leakage, thereby enhancing V oc and PCE under dim-light conditions [29]. Additionally, the use of a thicker BHJ active layer can reduce the leakage current (J 0 ) and increase the R sh of iOPVs [50,59,66,67,85]. However, considering the recombination losses of carriers and excitons, long propagation paths in these thick active layers may not be a favorable strategy in some cases. In summary, highefficiency iOPVs can be achieved by (i) optimizing the bandgaps of organic semiconductors (donors and acceptors) to match the indoor light emission spectrum, (ii) constructing BHJ systems with the highest possible V oc (with low voltage loss), and (iii) designing materials and devices with sufficiently large R sh and minimized leakage current.
Currently, the PCEs of state-of-the-art iOPVs have surpassed 30% under indoor conditions, resulting in power outputs of approximately 25, 63, and 110 μW at 290, 700, and 1200 lx, respectively [29]. These efficiencies seem to be sufficiently high for practical application of iOPVs, but from a scientific perspective, there is still room for improvement up to the SQ limit, as mentioned above. Thus, we expect that synergistic advances in basic physical research on exciton/carrier dynamics under low illuminance and state-of-the-art materials and device engineering will further improve the PCEs of iOPVs.
Another critical factor for practical application is the stabilities and durabilities of the devices. Ultimately, to replace the battery market for IoT ecosystems, highly stable iOPVs that can maintain over 80% of their initial efficiencies in indoor environments for several years are desired. Owing to the absence of UV light and the relatively low light intensity, iOPVs have shown better stability under indoor LED lighting than under 1-sun illumination [28], which can overcome the shortcomings related to the photostabilities of organic materials. However, the stabilities of iOPVs under elevated temperature, humidity, and oxygen conditions have rarely been studied; thus, additional systematic investigations are necessary to ensure comprehensive device stability for practical application. We envision that, compared to the harsh outdoor conditions with drastic fluctuations in solar irradiation, heating, and weather, operation under relatively mild indoor conditions will offer a promising route for sufficiently prolonged device durability for iOPVs.
As the next target of iOPV research, upscaling, modularization, and integration of devices should be highlighted to find real industrial markets. Nevertheless, to date, only a few achievements have been reported for practical iOPV modules (especially flexible modules) [35,49,52,60,74,89]. A process shift from spin coating to innovative printing techniques compatible with large-scale, low-cost manufacturing is necessary to produce practical iOPV modules. The unique features of iOPVs based on organic semiconductors can increase the freedom of design, shapes, colors, and transparency of actual devices, thereby facilitating their integration into real IoT applications. With rapid advances in both materials and devices, iOPVs will realize their full potential for commercialization in the near future.