Operational stability, low light performance, and long-lived transients in mixed-halide perovskite solar cells with a monolayer-based hole extraction layer

Due to their tunable bandgap and low manufacturing cost, metal halide perovskite solar cells are attractive for ambient/indoor light-harvesting applications. In this work, we evaluate p-i-n perovskite solar cells fabricated with MeO-2PACz, a molecular monolayer hole extraction layer, as potential candidates for ambient light-harvesting applications. Two triple-cation mixed halide lead perovskite absorbers are compared, one with high bromide content (Br/I ratio 1:2, bandgap 1.72 eV, 16.1% power conversion e�ciency) and one with low bromide content (Br/I ratio 1:11, bandgap 1.57 eV, 19.1% power conversion e�ciency). Both materials demonstrated good stability while operating under simulated sunlight at the maximum power point for 100 h, a cumulative light dose comparable to over two years of ambient use. After 100 h operation, however, the measured device e�ciency fell temporarily due to a transient loss of output current before returning to the nominal level after a long recovery period in the dark. These transient losses were more apparent in the wide bandgap device under strong light and were likely caused by light-induced halide segregation. After recovery, both devices retained good performance under a wide range of light intensities. Under a simulated ambient light source (835 lx white LED), the power conversion e�ciency of the wide bandgap device reached 30.4%.


Introduction
Solar cells based on metal halide perovskite materials have demonstrated exceptional e ciency (25.7%), approaching that of crystalline silicon.[1] The perovskite absorber layers have many favorable electrooptic properties, including high defect tolerance, [2] long charge carrier lifetimes and diffusion lengths, [3,4] and high absorption coe cients.[5] One particularly useful feature is the ability to tune the bandgap by substituting different ions in the perovskite crystal structure.[6,7] The bandgap of these mixed composition perovskites can be optimized for use in tandem cells, [8][9][10][11] or cells designed to operate e ciently under arti cial light sources.[12] In combination with the low manufacturing cost, the high potential e ciency of wide-bandgap perovskite solar cells under arti cial light makes them highly promising for powering small, low-power devices such as remote sensors, RFID tags, and Bluetooth Low Energy (BLE) beacons.[13] The bandgap of metal halide perovskites can be adjusted over a wide range by varying the ratio of halide ions, normally iodide and bromide, occupying the X-site of the ABX 3 crystal structure.[14,15] The ratio of the A-site cations, typically a mix of cesium (Cs + ), formamidinium (FA + ), and methylammonium (MA + ), is optimized to increase the stability and/or improve the electrical properties of the perovskite.[16,17] Increasing the bandgap via the halide ratio has limitations, however.As the Br/I ratio is increased, stability and performance can be compromised by the tendency for the I − and Br − ions to separate into distinct phases when the perovskite is illuminated.[18,19] This transient diffusion phenomenon -the halide distribution returns to a homogeneous mixture in the dark -is known as halide segregation.[20][21][22][23] In this work, we focused on a p-i-n or "inverted" device structure employing a carbazole-based selfassembled monolayer (SAM) as a hole extraction layer.Record e ciencies for both single junction and tandem solar cells have been recently achieved by coating the transparent oxide substrates with these molecular monolayers.[24][25][26] It has also been reported that perovskite layers deposited on SAM-treated substrates are resilient toward halide segregation.[24] We wanted to explore the stability and performance of the monolayer-based device stack with both wide and narrow gap absorbers.To this end, we prepared devices using mixed-halide perovskites with different bromide-to-iodide ratios.One material has a narrow bandgap optimal for natural sunlight, the other has a wide bandgap suitable for ambient light-harvesting or tandem applications.Operational stability was assessed by tracking the maximum output power for 100 h under AM1.5G simulated sunlight, a total light ux equivalent to over two years of operation under typical indoor lighting conditions.While the e ciency of both the narrow and wide gap devices remained above the initial value throughout the test, signi cant transient losses were observed after the test was completed.These long-lived transients, which act to suppress the responsivity of the device under strong light, were more apparent when the wide-bandgap perovskite with high bromide content was used as the absorber layer.The initial performance could be recovered after storing the devices in the dark.After recovery, both devices retained good performance under low light conditions.Finally, the wide bandgap device was con rmed to operate e ciently under simulated ambient light.In addition to quantifying the stability and e ciency of the SAM-based device structure for ambient light harvesting, our results provide insights into the in uence of perovskite bandgap and composition on the operational stability, low light performance, and transient behavior of monolayer-based p-i-n perovskite solar cells.

Device fabrication and test conditions
For the low bromide content (low-Br) material we selected our Cs 0.05 FA 0.80 MA 0.15 PbI 2.75 Br 0.25 reference composition.[27] The ratio of bromide to iodide is 1:11 (8.3% of the total halide content is bromide), and the bandgap of 1.57 eV (787 nm) is optimized for single junction performance under AM1.5G sunlight.For the high bromide content (high-Br) formulation we used Cs 0.05 FA 0.80 MA 0.15 PbI 2 Br.The bromide to iodide ratio is 1:2 (33% bromide), resulting in a wider bandgap of 1.72 eV (722 nm).This wide-bandgap absorber is expected to perform well under ambient lighting such as white LEDs.The composition of the A-site is kept unchanged.The hole extracting monolayer was made with a carbazole-based SAM molecule, [2-(3,6-dimethoxy-9H-carbazol-9-yl)ethyl]phosphonic acid (MeO-2PACz, molecular structure shown in Fig. 1a), which was applied to indium tin oxide (ITO)/glass substrates by spin coating from 1 mM ethanol solution.[28] Perovskite solar cells were fabricated on the SAM-treated ITO glass substrates based on the processes elaborated in our previous reports [29][30][31] and outlined in the supporting information.Each device had an active (masked) area of 0.1 cm 2 .The complete device structure, illustrated in Fig. 1b, is ITO/MeO-2PACz/perovskite/C 60 /BCP/Ag.Cross-sectional SEM images (Fig. S1) con rm the perovskite layers in both devices to be about 600 nm thick, with comparable grain size and morphology.While the devices were exposed brie y to air to apply solder to the electrode contacts, they were otherwise kept under an inert atmosphere for the duration of the work.All tests and measurements were performed at ambient temperature.

Device performance
Stabilized output and current-voltage (J-V) curves for representative low-Br and high-Br devices are shown in Fig. 1c.Under AM1.5G, the power conversion e ciency (PCE) for the low-Br device was 17.6%, while that of the high-Br device was 15.1%.The lower PCE of the high-Br device under AM1.5G is expected considering the wider bandgap, as less light energy is absorbed.The difference in bandgap leads to the high-Br device having a lower observed short circuit current (J sc ) and higher open circuit voltage (V oc ) compared to the low-Br device.Fig. 1d shows stabilized output and current-voltage (J-V) curves for the same devices after 100 h operation at the maximum power point under AM1.5G.The characteristics of both cells improved: The stabilized PCE of the low-Br device reached 19.1%, while that of the high-Br device increased to 16.1%.While the short circuit currents decreased slightly, the open circuit voltages increased to 1.08 V and 1.17 V, respectively.The ll factors (FF) also increased.The FF of the low-Br device remained slightly higher than that of the high-Br counterpart, at 0.79 and 0.76, respectively.
The output characteristics before and after operation for 100 h are compiled in Table 1.

Operational stability
In this section, we present the evolution of the device properties during 100 h operation under AM1.5G.The maximum power point tracking (MPPT) algorithm which kept the devices at the maximum output power, p m , was paused for brief intervals every 20 minutes to record J-V curves.The progression of J sc , V oc , and FF was obtained from the J-V curves, while the PCE is taken directly from the MPPT data.The results are compiled in Fig. 2, while representative J-V scans at 1, 10, and 100 h are shown in Fig. S2.V oc and FF increased over the 100-h test interval for both the low-Br and high-Br devices, while J sc fell slightly.
The increase in voltage and ll factor offsets the decrease in output current such that the overall e ciency of both cells after 100 h MPPT remained higher than the initial value.The PCE increased and reached a maximum value at about 25 h.This "burn-in" behavior has been previously noted as a characteristic of p-i-n perovskite solar cells, though the exact cause is not known.[32] Given the improvements in output voltage over the burn-in period and the noted sensitivity of p-i-n monolayer devices to modi cation of the perovskite/C 60 interface, [33,34] this interface may become naturally passivated as the device ages over the initial hours of operation.In sharp contrast, considerable degradation was observed when the perovskite materials were tested in n-i-p devices with spiro-OMeTAD hole transport layers.These results con rm the superior operational stability of the monolayer-based p-i-n device stack for both the wide and narrow bandgap perovskite absorber layers.The 100 h MPPT data for the spiro-OMeTAD-based devices is given in Fig. S3.

Transient effects
The devices showed good stability during the 100-h operational test under continuous AM1.5G, with only minor changes to the J-V curves and minimal net loss of PCE.We were therefore surprised to see a sharp drop in output current when, following the completion of the test, the devices were subsequently left under ambient room light for a short period (<1 h).This loss was transient, however, and the nominal performance was recovered after the device was left in the dark for a longer period (500 h).These transient effects were more pronounced for the high-Br device than the low-Br device.The details of the transient effect for the high-Br device are explored in Fig. 3.
As illustrated by the J-V curves in Fig. 3a, an acute performance drop develops after the MPPT test is completed and the device was left for an hour or so in the dark.J sc fell from 18.3 mA cm −2 to about 12 mA cm −2 , before returning to 18.7 mA cm −2 after a 500-h recovery period in the dark.We noted, however, that the magnitude of the transient loss was greatly reduced when the device was probed with lowintensity light.To help characterize this interesting previously unremarked phenomenon, it is instructive to examine J sc normalized to the incident light power.This quantity is known as the responsivity of the device, (j is the irradiance).The responsivity is shown plotted against light intensity in Fig. 3b.Although the responsivity remained essentially constant over the measured range of light intensities in the initial and recovered states, in the transient state the responsivity falls with increasing light intensity from about 1/100 sun.The responsivity below 1/100 was unchanged from the initial and recovered values, however, so the impact of the transient effects under ambient light is expected to be minimal.
External quantum e ciency (EQE) spectra, shown in Fig. 3c, were taken to help clarify the transient changes.All three of the measured spectra -taken for the initial, transit, and recovered states -show the same overall EQE of 80-85%.The measured EQE is not lower during the transient regime because the intensity of the monochromatic light source used for this experiment corresponds to only about 1/100 sun -the threshold below which, as noted above, the transient loss of responsivity ceases to be signi cant.Although the spectral shape and intensity are unchanged, the onset region temporarily developed a low-energy shoulder during the transient regime which dissipated after recovery.This is more clearly seen in Fig. 3d, where the differential of the EQE signal is plotted against photon energy.Following the procedures developed by U. Rau et al., the bandgap is estimated from the peak centers determined from Gaussian curve tting.[35] As listed in Table 2, the bandgap of the high-Br material shifts from 1.717 to 1.699 eV before returning to 1.717 eV, and the peak width increases from 83 to 110 meV before returning to 85 meV.Interpolating the bandgap and bromide content of the two perovskite materials, the shift in the bandgap of the high-Br device suggests a reduction in bromide content from 0.33% to 0.30%.
In other words, the regions of the perovskite layer where light is most strongly absorbed become temporarily iodide rich.This is consistent with the previously reported segregation of the halide ions into Br-rich and I-rich subdomains under strong light.[22,23,[36][37][38] Light-induced lattice expansion at the illuminated perovskite surface seems the likely driving force.[39][40][41][42] It's noteworthy though that in the present instance the detrimental in uence of the halide segregation on device performance is exhibited after exposure to AM1.5G rather than during it.
Signi cant structural changes within the perovskite layer are deduced from the X-ray diffraction (XRD) measurements shown in Fig. S4.A high-Br perovskite thin lm was prepared on a MeO-2PACz coated ITO substrate and exposed to AM1.5G for 100 h.(Mirroring the MPPT experiments, the perovskite was exposed from the ITO side.)A signi cant shift of all the perovskite diffraction peaks to lower 2q values was observed following light exposure, accompanied by an overall broadening.Following the previous report, we interpret this as a light-or temperature-induced heterogenous expansion of the perovskite lattice.[43] The diffraction patterns were unchanged for at least several hours after illumination ceased, and began to show clear signs of recovery after 24 h in the dark.It seems likely that the drop in current responsivity in the high-Br device and its gradual recovery in the dark is related to this lattice expansion, which also coincides with the halide segregation induced shift in the perovskite bandgap determined from the EQE measurements.While lattice expansion and phase segregation are probably both contributing factors, more research is needed to determine the role of the p-i-n device stack, and in particular the in uence of the MeO-2PACz monolayer, in driving or suppressing the transient changes in responsivity.
While similar changes in device responsivity and output current were observed for the low-Br device (Fig. S5) the loss of output current was much smaller and there was no corresponding shift in the onset of the EQE spectrum.These transients were similarly recovered after 500 h in the dark.The contrasting behavior for the high-Br and low-Br devices establishes further evidence that the transient effect is related to halide segregation.We also brie y checked the transient behavior of the high-Br perovskite absorber in a p-i-n device where the MeO-2PACz monolayer HEL was replaced by PTAA.The operational stability is lower than the MeO-2PACz devices, as shown in Fig. S6.This agrees with an earlier study.[44] While a similar shift in bandgap was observed from the EQE data after 100 h MPPT, there was no transient loss of device current following the MPPT experiment.Instead, the device performance further degraded in the days following the MPPT experiment, with no indication of recovery.
Impedance measurements tell a different story, however.Irreversible changes in both devices were observed in the complex impedance spectra recorded under AM1.5G (Fig. S7 and Appendix A).The parallel resistance [45] falls over the course of the MPPT experiment and does not return to the initial state even after 500 h recovery.The effective shunt resistance remains high, however, and the impact of the change on the measured device performance (as expressed by the J-V curves and MPPT under AM1.5G) should be negligible.

Light intensity dependence
Light intensity dependence tests and performance evaluation under ambient light were performed on cells previously stressed with 100 h operation at maximum power under AM1.5G, after full recovery from the transient effects described above.During these tests, the cells absorb a photon ux equivalent to over two years of continuous operation under ambient light.The results presented here are therefore indicative of how these devices, provided su cient protection from the atmosphere, might perform after several years of operation in a typical Internet of Things application.
To arrive at a clearer picture of how the performance of these stressed cells change under different light intensities, simulated AM1.5G radiation was attenuated using neutral density lters in combination with a variable aperture.(Calibration methods are given in the supporting information.)By keeping the spectral composition constant, direct comparison of the PCE data is possible across a range of light intensitysomething that would not have been possible if, for example, we had switched to a LED source for the low light measurements.Fig. 4a shows the PCE of the low-Br and high-Br devices as the AM1.5G source is attenuated from 1 sun to less than 1/1000 suns.The PCE is calculated from the average of forward and reverse J-V curves recorded at each light intensity.(Full datasets are given in Appendix B of the supporting information.)The Shockley-Queisser (SQ) limit for the two materials is also indicated.While the output from both devices is lower under low light, the SQ limit also falls as the light intensity is reduced.At any given light intensity, both devices perform equally strongly vs. their respective theoretical limits.This can be clearly seen in the plot of the cell output normalized to the SQ limit shown in Fig. S8.

Device characteristics under simulated ambient light
Having established that both materials retain the same performance relative to their respective SQ limits under attenuated AM1.5G, we could con rm the stable output of the widegap high-Br composition under simulated ambient indoor light.A 50 W white LED adjusted to an irradiance of 250 μW cm −2illuminance 835 lx, roughly the brightness of a normal o ce environment -was used to simulate ambient light conditions.As shown in Fig. the spectral intensity is mostly restricted to between 430-700 nm.As shown in Fig. S10, the absence of any signi cant infrared component imparts an e ciency advantage to absorbers with wide bandgaps in the region of 1.7-1.9eV, including the high-Br perovskite studied in this work.
Fig. 4b shows the maximum power output of the high-Br device under the white LED.The cell was kept in the dark overnight before the measurements started.The average output of the high-Br cell during 10 h of operation was 76.2 μW cm −2 , corresponding to a PCE of 30.4%.This e ciency, while promising, might be increased by changing the halide composition to slightly increase the bandgap of the perovskite absorber, [46] or by reducing charge carrier recombination at the perovskite/ETL interface using 2D capping layers. [34]

Conclusion
The performance and stability of p-i-n perovskite solar cells using MeO-2PACz as a hole extracting monolayer were examined with a focus on ambient light-harvesting applications.During this work, we also clari ed the in uence of the perovskite bromide-to-iodide ratio on device performance.Two different mixed halide perovskite absorbers were trialed, a low-Br material (Br/I 1:11) with a bandgap of 1.57 eV and a high-Br formulation (Br/I 1:2) with a bandgap of 1.72 eV well-suited for operation under arti cial light sources such as LEDs.We found that the devices maintain high PCE during 100 h operation at maximum power under AM1.5G.Transient losses were observed after the MPPT tests were completed, however.The transient changes are attributed to halide segregation and dissipated after the devices were stored in the dark for 500 h.Fortunately, given the intended application in ambient light harvesting, the effect of the transient changes on the device performance in low light was found to be negligible.After recovery, the performance of both materials closely tracked their respective SQ limits over a wide range of incident light intensities from less than 1/1000 sun to 1 sun.Under low-intensity white LED illumination (835 lx) the wide bandgap, high-Br perovskite material achieved a stable PCE of 30.4%.Overall, we conclude that the monolayer-based p-i-n device structure with wide bandgap perovskite absorber layers shows great promise for long-term deployment in ambient-light harvesting applications from the standpoint of both e ciency and operational stability.

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download.

Figures
Figures

Figure 2 Electrical
Figure 2

Figure 3 Properties
Figure 3

Table 1 .
Devices characteristics before and after 100 h operation under AM1.5G.
a Calculated from the J-V scans b Stabilized measurement after at least 300 s maximum power point tracking

Table 2 .
Gaussian peak fitting results for the derivative of the external quantum efficiency, d(EQE)/dE.