Device Structure of the TPU-SC
The fabricated TPU-SC consists of an n–i–p CsPbBr3/GaAs heterojunction. The device structure is shown in Fig. 1a. For the WGS, we grew a CsPbBr3 layer (shown in green), and for the NGS, we used a 1-inch p-doped GaAs substrate (001) (shown in black). Zinc oxide (ZnO) was used as the electron transport layer (ETL). We denote the CsPbBr3/GaAs heterointerface as HI-I, and the ZnO/CsPbBr3 heterointerface as HI-II. The band diagram of the ZnO/CsPbBr3/p-GaAs structure was simulated using the COMSOL Multiphysics® software, and the result is shown in Fig. 1b. A TPU process is expected to occur at HI-I after the photogenerated electrons in the CB of GaAs have accumulated at the interface states. Results of the TPU process occurring at HI-I are discussed in the following sections.
The Ideal Energy Conversion Efficiency
Figure 2 shows two maps of the theoretical conversion efficiency as functions of the NGS and WGS bandgaps for two different band-offset configurations. Here, the band-offset configuration is expressed by the ratio of the CB and VB discontinuities (ΔECB:ΔEVB). For the prediction of the efficiency limit under 1-sun illumination, we used the detailed balance approach1,10,17 and the AM 1.5G spectrum. Figure 2a shows the data for ΔECB:ΔEVB = 3:2. This value of ΔECB:ΔEVB is nearly equal to that of an Al0.3Ga0.7As/GaAs heterojunction15. The highest conversion efficiency is 46.4%, which is achieved by using NGS and WGS bandgap energies of Eg, NGS ≈ 1.60 eV and Eg, WGS ≈ 3.10 eV, respectively. Therefore, GaAs (Eg = 1.42 eV) is a reasonable good choice for the NGS. However, the optimal WGS bandgap is very large, and thus AlxGa1-xAs is not the best choice for the WGS of such a high-efficiency TPU-SC (the bandgap energy of AlxGa1-xAs is only about 2.20 eV even for x = 0.9). Note that Al0.3Ga0.7As (Eg = 1.81 eV) was used in the previously reported TPU-SCs. Furthermore, although a larger Al molar fraction x provides a larger bandgap, AlxGa1-xAs switches from a direct bandgap semiconductor to an indirect bandgap semiconductor when x ≥ 0.4541.
The conversion efficiency map for ΔECB:ΔEVB = 7:1 is shown in Fig. 2b. Since the electron affinity (χ) and the Eg values of CsPbBr3 and GaAs are known, we can estimate the barrier heights for both the CB and the VB: The χ values of CsPbBr3 and GaAs are 3.30 eV and 4.07 eV, respectively42,43. Therefore, the ratio of ΔECB to ΔEVB becomes 7:1 (ΔECB = 0.77 eV and ΔEVB = 0.11 eV). Figure 2b indicates that the maximum conversion efficiency is 48.5% for Eg, NGS ≈ 1.50 eV and Eg, WGS ≈ 2.60 eV. This efficiency maximum is 2.1% higher than that for ΔECB:ΔEVB = 3:2. Furthermore, we find that CsPbBr3 is suitable for the WGS of such a TPU-SC, because CsPbBr3 perovskite has an Eg of 2.33 eV. In comparison to the Al0.3Ga0.7As/GaAs-based TPU-SC, the CsPbBr3/GaAs-based TPU-SCs provide a higher maximum conversion efficiency in addition to the exceptional optoelectronic properties of CsPbBr3 and GaAs.
EQE Spectra
The room-temperature EQE spectra of our CsPbBr3/GaAs-based TPU-SC are shown in Fig. 3. The blue curve represents the EQE spectrum measured under single-color excitation conditions, which means that carrier generation occurs mainly through interband excitation. The EQE signal gradually increases as the excitation wavelength decreases. At wavelengths longer than ~ 900 nm, there is no EQE signal, because these photons cannot excite any semiconductor in this device. The EQE signal shows a sharp onset when the excitation wavelength is about 900 nm, where GaAs starts to absorb the photons. The signal reaches a plateau at about 870 nm, which corresponds to the GaAs bandgap. As a result of the built-in electric field, the photogenerated electrons in the CB of GaAs (generated by the photons in the wavelength range from approximately 530 to 870 nm) drift toward HI-I, while the photogenerated holes drift to the rear electrode (the holes are not influenced by HI-I). The electrons are trapped at the interface states of HI-I and accumulate there. These accumulated electrons partially recombine with the holes that reach HI-I by diffusion. We described this scenario in previous publications10–13. A part of the electrons at HI-I can overcome the energy barrier by thermal excitation (evidence for thermionic emission is provided in Influence of the Temperature). At an excitation wavelength of about 530 nm, the EQE signal abruptly increases, because the corresponding photon energy is high enough to induce interband transitions in both CsPbBr3 and GaAs. As the wavelength becomes shorter, the photon energy becomes higher than the CsPbBr3 bandgap energy and the photons are stronger absorbed by the CsPbBr3 layer (the number of photons that reach the GaAs layer becomes smaller). Since the holes generated in the CsPbBr3 VB have a longer path to the rear electrode, which implies a larger recombination probability, the EQE should decrease if the excitation wavelength is reduced further.
The dark red curve in Fig. 3 represents the EQE spectrum measured under two-color excitation conditions. The comparison with the data for single-color excitation reveals that the EQE in the wavelength range between the CsPbBr3 bandgap and the GaAs bandgap (~ 530–870 nm) decreases when the device is additionally irradiated with the IR photons, and in the wavelength range below 530 nm (the CsPbBr3 bandgap), the IR photons lead to an enhancement of the EQE signal. In contrast, our previous work on TPU-SCs with a single heterointerface showed an IR-induced EQE enhancement in the range between the WGS and NGS, and no enhancement in the short-wavelength range. In particular, in an Al0.3Ga0.7As/GaAs-based TPU-SC, there is no IR-induced EQE enhancement when the wavelength of the photons for interband excitation is shorter than 680 nm (the Al0.3Ga0.7As bandgap). This result is attributed to the fact that high-energy photons are strongly absorbed by the WGS,10–13 which causes a reduction of the number of photons that reach the NGS layer, leading to a reduction of the electron density at the heterointerface. As a result, the IR-induced photocurrent enhancement gradually disappears as the wavelength of the photons for interband excitation becomes shorter; if there are no photogenerated electrons in the NGS CB, the IR laser beam has no effect, because the photogenerated electrons in the WGS CB do not accumulate at the energy barrier10–12.
To clarify the behavior of our CsPbBr3/GaAs-based TPU-SC, we consider our previous results on a double-heterointerface TPU-SC with an additional Al0.7Ga0.3As (~ 2.50 eV) layer grown on the Al0.3Ga0.7As/GaAs (1.79 eV/1.42 eV) structure: this device exhibited an IR-induced EQE improvement in the short-wavelength range13. Hence, the IR-induced EQE enhancement observed in Fig. 3 at wavelengths below 530 nm is likely due to the interface states of HI-II. To support this interpretation, we estimated the fraction of 500-nm light absorbed in each layer of the CsPbBr3/GaAs-based TPU-SC (the calculation details are explained in Section S4 of the Supplementary Information). Our estimation reveals that photons with such short wavelengths do not induce electron accumulation at HI-I. Furthermore, in this measurement, non-linear phenomena such as multi-photon absorption in CsPbBr3 are negligible, because we used a tungsten-halogen lamp and a CW laser for the excitation, i.e., light sources with a relatively low energy density. Therefore, we believe that the IR-induced EQE enhancement in the short-wavelength regime is caused by electron accumulation at HI-II.
Conversely, it has been found that the EQE signal in the wavelength region between the bandgaps of CsPbBr3 and GaAs reduces with the IR irradiation. The change in EQE with the IR irradiation strongly depends on the excitation power density. We discuss the details in Excitation Power-Dependence of JSC.
Figure 4 shows the temperature dependence of the EQE spectrum for single-color excitation (the wavelength of the photons for interband excitation is shown on the x-axis, and the temperature is shown on the y-axis). At temperatures below 180 K, the CsPbBr3/GaAs-based TPU-SC hardly produces photocurrent due to the CB discontinuity at HI-I: The difference between the electron affinities χ of CsPbBr3 and GaAs provides a barrier height of 0.77 eV at HI-I. Since the thermal activation energy at 297 K is only 0.026 eV, the electrons accumulated at the heterointerface can hardly overcome the barrier at much lower temperatures. In this measurement, the temperature governs the electronic excitation at HI-I, because 1319-nm (sub-bandgap) photons were not used. Consequently, Fig. 4 confirms the thermal activation of electrons accumulated at HI-I, which plays an important role in the photocurrent generation when sub-bandgap photons are absent44. Figure 4 also reveals a clear shift of the absorption edge of GaAs: it shows a blueshift with decreasing temperature due to the well-known effect responsible for the temperature-dependent changes in the bandgaps of III–V semiconductors45,46. The absorption edge of CsPbBr3, on the other hand, is almost insensitive to the temperature within the measured range.
It has been reported that the bandgap energy of lead bromide perovskites (XPbBr3, with X = FA+, MA+, or Cs+) decreases with decreasing temperature, in contrast to the behavior of III–V semiconductors. The increase of the XPbBr3 bandgap energy with increasing temperature is attributed to changes in the crystal phase. In general, XPbBr3 perovskites can exist in various crystal phases including tetragonal and cubic phases. In the case of CsPbBr3, the orthorhombic phase can be observed easily at room temperature, because of its good stability, and as the temperature increases, phase changes occur. Mannino et al. reported that the Eg of CsPbBr3 has a small dependence on the temperature in the range 200–300 K38. This phenomenon is caused by the thermal expansion of the orthorhombic phase. At about 380 K (107°C), the CsPbBr3 bandgap changes abruptly due to the phase transition from the orthorhombic phase to the tetragonal phase (several investigations confirmed this phase transition already at 361 K21,39,40). This is in contrast to FAPbBr3 and MAPbBr3, where a tetragonal–cubic phase change occurs in the temperature range 200–300 K, and results in an Eg change of CsPbBr3 of approximately 0.02 eV in the range 200–380 K38. Within the temperature range used for Fig. 4, CsPbBr3 is in the orthorhombic phase. As a result, the absorption edge is almost constant 2.33 eV (the Eg change is within ~ 0.008 eV).
JV Characteristics
The JV characteristics under different illumination conditions are shown in Fig. 5; the black, blue and dark red points represent the JV characteristics in the dark, under single-color excitation conditions (PInter = 101 mW/cm2), and under two-color excitation conditions (PInter = 101 mW/cm2 and PIntra = 86 mW/cm2), respectively. Because the ZnO and CsPbBr3 layers are transparent for 784-nm photons, these photons are absorbed in the p-GaAs substrate and generate electrons and holes in the GaAs CB and VB, respectively. The photogenerated electrons drift toward HI-I, where they can accumulate at the interface states. The photogenerated holes drift to the rear electrode10,11,13. While thermal excitation can in principle excite the electrons at HI-I to the CB of CsPbBr3, thermal activation hardly occurs since the energy difference between the CB minima of CsPbBr3 and GaAs is 0.77 eV, which is much higher than the thermal energy at room temperature (0.026 eV). This is the reason for the small magnitude of the detected photocurrent in the case of single-color excitation.
For two-color excitation, a slight photocurrent and photovoltage enhancement due to the additional IR photons can be confirmed. The IR-induced gain in the short-circuit current (ΔJSC) and the IR-induced gain in the open-circuit voltage (ΔVOC) are 0.011 µA/cm2 and 0.025 V respectively. Although this data at first seems inconsistent with the EQE results, both results are actually in agreement. To understand this, we need to consider that a tungsten-halogen lamp was employed for the EQE measurement, which only provided 6.0×10–4 mW/cm2 at about 780 nm (a photon flux of 2.4×1016 m–2s–1). As shown later, intraband transitions hardly occur as long as PInter is below 20 mW/cm2 (a photon flux of 8.0×1020 m–2s–1). Therefore, the power density of the monochromatic beam for interband excitation used in the EQE measurement was too low. For the JV measurements, we employed a CW solid-state laser operating at 784 nm, which provided a photon flux of 4.0×1021 m–2s–1 for interband excitation.
The behavior of the open-circuit voltage (VOC) is interpreted as follows: The VOC enhancement in the case of additional illumination with sub-bandgap photons (i.e., in the case of optically-induced intraband transitions) occurs because the 784-nm photons induce interband transitions only in the GaAs layer. When the electrons accumulated at HI-I are excited by the 1319-nm photons and reach the CB of CsPbBr3, the electron quasi-Fermi level shifts to a higher level and thereby increases the VOC. Thus, the observed positive ΔVOC evidences an adiabatic intraband excitation process at HI-I10,44. In Excitation Power-Dependence of the Open-Circuit Voltage, we show how ΔVOC depends on PInter and PIntra.
Excitation Power-Dependence of JSC
Figure 6 shows the dependence of the short-circuit current (JSC) on PInter for single-color excitation. The 784-nm photons excite the GaAs substrate and intraband transitions are not induced, because sub-bandgap photons are absent. The photogenerated electrons in the CB of GaAs drift toward HI-I, where electrons accumulate due to the CB discontinuity. To overcome the barrier, an energy is required that is nearly 30 times the thermal energy at room temperature. Therefore, the temperature is important for the JSC under single-color excitation conditions. The effect of the temperature is discussed in Influence of the Temperature.
To discuss the data in Fig. 6, we fitted the data to a single power law, \({\text{J}}_{\text{SC}}\text{ }\text{∝}\text{ }{\text{P}}_{\text{Inter}}^{\text{ n}}\). The estimated power exponent is n = 0.67, which significantly deviates from unity. The reason for this sublinear power dependence is the weak built-in electric field and the recombination with holes: The electrons generated in the GaAs CB drift toward HI-I due to a weak electric field. As the electrons accumulate at HI-I, the electric field becomes weaker, which increases the electron–hole recombination rate. Therefore, the rate of thermal activation at HI-I decreases as PInter increases, leading to the observed sublinear relationship.
Figure 7a shows the photocurrent gain under short-circuit conditions (\(\text{∆}{\text{J}}_{\text{SC}}\text{ =}{\text{ }\text{J}}_{\text{SC, with infrared}}\text{ – }{\text{J}}_{\text{SC, without infrared}}\)) as a function of the power density of the 784-nm laser beam in the case of PIntra = 440 mW/cm2. The ΔJSC increases with PInter, because stronger interband excitation provides a higher electron density at HI-I (the 784-nm photons induce interband transitions in GaAs, and the photogenerated electrons are transported to HI-I). The data can be well fitted using a single power law, ΔJSC\(\text{ }\text{∝}\text{ }{\text{P}}_{\text{Inter}}^{\text{ n}}\), with n = 0.83. Ideally, ΔJSC should be proportional to the electron density at HI-I. Furthermore, the recombination in ZnO and CsPbBr3 (as well as that at HI-II) should be negligible, because the 784-nm photons induce interband transitions only in the GaAs layer, and thus there are no photogenerated holes in ZnO and CsPbBr3. The sublinear characteristic, therefore, originates from a recombination within GaAs or at HI-I. While stronger interband excitation leads to a higher electron density at HI-I, this increased electron density reduces the built-in electric field, facilitating electron–hole recombination within GaAs. Hence, the electron density at HI-I sublinearly increases with PInter. This behavior was also observed in our previous reports11,47.
Figure 7b clarifies the dependence of ΔJSC on PIntra in the case of PInter = 450 mW/cm2. This data cannot be fitted to ΔJSC\(\text{ }\text{∝}\text{ }{\text{P}}_{\text{Intra}}^{\text{ n}}\), because the data shows a gradual change of the power exponent, i.e., the value of n decreases as PIntra is increased. In the low excitation regime, we find n = 0.26, which is the maximum value in the measured range of power densities. The minimum value is n = 0.11, observed at PIntra ~ 450 mW/cm2. This reduction of n can be interpreted as a reduction of the carrier separation efficiency at HI-I. Since the 784-nm photons induce interband transitions only in the GaAs layer, recombination in ZnO and CsPbBr3 (including that at HI-II) can be ignored. Therefore, the observed reduction of the carrier separation efficiency occurs at HI-I. The origin of the reduction is a lower built-in electric field at higher electron densities in the CBs of ZnO and CsPbBr3 (and at HI-II), because the electrons that are optically excited to the CB of CsPbBr3 hardly drift to HI-II if the built-in field is weak. This means that the functionality of HI-I is limited by the electron densities in ZnO and CsPbBr313.
Figure 8 shows ΔJSC as functions of PInter (x-axis) and PIntra (y-axis). The data scatters around zero and mostly shows negative values when PInter is below ~ 20 mW/cm2. When PInter exceeds 20 mW/cm2, an increase in PInter also leads to a remarkable increase in ΔJSC. Note that, in the region below 20 mW/cm2, the data is independent of the intensity of the IR light. We interpret this behavior as follows: Since the electron density in the GaAs CB is mainly determined by the intensity of the photons used to induce interband transitions, the electron density remains low as long as PInter is low. In addition, we used a p-GaAs substrate for the NGS layer (there is no i-GaAs layer), which makes it more difficult to increase the electron density compared to an Al0.3Ga0.7As/GaAs-based TPU-SC. Because the possibility of upconversion at HI-I is determined by the electron density at the initial energy levels for the intraband transitions, there is almost no ΔJSC in the case of weak interband excitation, even at high values of PIntra.
The result of the negative ΔJSC when PInter is below ~ 20 mW/cm2, shown in Fig. 8, agrees with the EQE results in Fig. 3. Since the photons used for interband excitation in our EQE measurement were generated by a tungsten-halogen lamp, the photon density was below the threshold. The power density of the monochromatic 784-nm beam generated by the tungsten-halogen lamp and the monochromator was approximately 6.0×10–4 mW/cm2, which is equivalent to 0.005 suns. In other words, the used power density was too small to produce a significant density of electrons at HI-I. According to Fig. 8, PInter needs to be higher than ~ 20 mW/cm2 to let a significant amount of electrons accumulate at HI-I. This result points out that electron accumulation at the heterointerface is a prerequisite for adiabatic intraband transitions caused by sub-bandgap photons. It is well known that non-radiative recombination plays a significant role due to the presence of interface states48,49. For two-color excitation conditions, the IR light is able to pump out carriers passivating the interface states50, which additionally induces non-radiative recombination at the heterointerface, and, thereby, the IR-induced EQE reduction occurs.
Excitation Power-Dependence of the Open-Circuit Voltage
The excitation power dependence of the VOC for single-color excitation is provided in Fig. 9. The VOC increases with PInter, because the electron density in the CB of GaAs increases. Although the accumulated electrons are thermally excited to the CB of CsPbBr3 as shown in Fig. 6, the VOC under single-color excitation conditions is determined by the quasi-Fermi-level splitting in GaAs, because in this case, the excitation at HI-I is a thermionic emission process.
Figure 10a shows the IR-induced gain in the open-circuit voltage (ΔVOC) as a function of PInter in the case of PIntra = 228 mW/cm2, and Fig. 10b shows the ΔVOC as a function of PIntra in the case of PInter = 260 mW/cm2. The positive ΔVOC values originate from the increase of the electron density in the CB of CsPbBr3 caused by the additional illumination with the 1319-nm photons (optically induced intraband transitions):
Figure 10a shows an increase in ΔVOC with increasing PInter. We consider that the carrier extraction at HI-I is an adiabatic optical process10,44. Therefore, the VOC increases. Compared to the thermionic emission and tunneling processes at HI-I, the intraband transitions induced by the 1319-nm photons provide an increased electron population in the CB of CsPbBr3. This induces a split between electron quasi-Fermi levels of CsPbBr3 and GaAs. The positive ΔVOC values in Fig. 10 evidence the presence of an adiabatic TPU process at HI-I. Since stronger interband excitation causes a higher electron density at HI-I, the ΔVOC increases with PInter.
Similarly, Fig. 10b shows an increase of ΔVOC with increasing PIntra, in agreement with the ΔJSC data shown in Fig. 7. As PIntra increases, more electrons accumulated at HI-I are upconverted to the CB of CsPbBr3, which induces a widening of the quasi-Fermi-level splitting.
Figure 11 shows ΔVOC as functions of PInter and PIntra. Overall, this map shows an increase in ΔVOC as both PInter and PIntra increase. Furthermore, a feature identical to that observed in Fig. 8 can be seen: when PInter is below ~ 20 mW/cm2, the ΔVOC values are scattered around zero. The increase with either PInter or PIntra only appears for sufficiently high PInter values. This feature indicates that the mechanism responsible for the IR-induced enhancement of VOC is identical to that for the IR-induced enhancement of JSC; the adiabatic upconversion of electrons contributes to both photocurrent and photovoltage. Furthermore, also the threshold characteristic in Fig. 11 is due to the requirement of a sufficiently high electron density at HI-I.
Influence of the Temperature
In this section, the effect of the temperature on the photocurrent and photovoltage of the CsPbBr3/GaAs-based TPU-SC is discussed. The temperature dependence of JSC for single-color excitation with PInter = 428 mW/cm2 is shown in the Arrhenius plot in Fig. 12. In general, the JSC increases with temperature, in agreement with the temperature dependence of the EQE in Fig. 4. This increase evidences thermionic emission at least at one of the two heterointerfaces in this device, since a higher temperature implies a higher thermal energy. Hence, at higher temperatures, the photogenerated electrons in the CB of GaAs can more easily overcome the band discontinuity by thermal excitation and then are collected at the front electrode.
The dashed line in Fig. 12 represents the result of fitting the data to the Arrhenius equation,
$${\text{J}}_{\text{SC}}\text{ = A}\text{exp}\left(\text{–}\frac{{\text{E}}_{\text{A}}}{{\text{k}}_{\text{B}}}\text{∙}\frac{\text{1}}{\text{T}}\right) \text{,} \text{(1)}$$
where A is a fitting parameter, kB is the Boltzmann constant, T is the absolute temperature, and EA is the thermal activation energy. Regarding EA, the observed IR-induced photocurrent gain (Figs. 7a and b) would suggest an EA that corresponds to the CB discontinuity at HI-I. However, the fit resulted in an EA of 0.108 eV, which is significantly smaller than the estimated CB discontinuity of 0.770 eV. We ascribe the obtained value of EA to an average energy level of occupied interface states at HI-II:
As indicated in the band diagram in Fig. 1b, the CB discontinuity at HI-II has a barrier height of ~ 1.00 eV (\({\text{χ}}_{\text{ZnO}}\) is much larger than \({\text{χ}}_{\text{CsPbB}{\text{r}}_{\text{3}}}\)). In this case, the electrons in the CB of CsPbBr3 should not need thermal energy to reach ZnO. Therefore, we consider the following process: First, the photogenerated electrons in the CB of GaAs accumulate at the interface states of HI-I. Then, the electrons are thermally excited to the CB of CsPbBr3 and transported to the front electrode. A fraction of these electrons relaxes from the CB to the interface states of HI-II and are trapped. These electrons need to be thermally activated to reach the CB of ZnO in order to be collected at the front electrode. This means that these electrons are first thermally excited at HI-I, and then are again thermally excited at HI-II. As a result, the physical meaning of the obtained value of EA may be the energy difference between the CB minimum of ZnO and the average level of occupied interface states at HI-II.
As shown in Figs. 8 and 11, ΔJSC and ΔVOC increase with both PInter and PIntra (if PInter > 20 mW/cm2). In Fig. 13, we provide additional evidence for the presence of optically induced intraband transitions at HI-I. Figure 13 shows two data sets of ΔVOC as a function of ΔJSC for PInter = 450 mW/cm2: The dark red points represent the ΔJSC–ΔVOC data acquired at room temperature as a function of PIntra (the data showing the IR-induced gain in JSC and VOC). This dataset indicates an increase in both ΔJSC and ΔVOC with increasing PIntra. The blue points represent the ΔJSC–ΔVOC data for PIntra = 0 as a function of the temperature (the data showing the temperature-induced changes in JSC and VOC). This dataset exhibits a clearly different feature, i.e., the ΔVOC decreases as ΔJSC increases. These two data sets allow us to compare the influence of optically induced intraband transitions (sub-bandgap photons) and thermal activation (thermal energy) on the SC performance.
In Fig. 12, we demonstrated thermal activation, which provides more additional photocurrent (ΔJSC) as the temperature increases. However, the advantage gained by this process is not identical to the advantage gained by the photocurrent increase through optically induced intraband transitions. To understand the difference, the ΔVOC in the case of thermal activation needs to be considered. It is well-known that the VOC has a strong inverse dependence on the dark saturation current, which increases with temperature51,52. Therefore, the VOC decreases with increasing temperature and the temperature-induced change in ΔVOC becomes more negative with increasing temperature (Fig. 13; blue data), in contrast to the ΔVOC due to the IR-induced intraband transitions (Fig. 13; red data). Because IR-induced intraband transitions are optical processes, the ΔVOC is positive (the quasi-Fermi-level splitting increases as discussed in Excitation Power-Dependence of the Open-Circuit Voltage). Consequently, the two ΔJSC–ΔVOC data sets have a different origin for the observed ΔJSC, and the positive ΔVOC values obtained using sub-bandgap photons confirms the presence of adiabatic optical excitation at HI-I10,44.