Temperature-dependent hysteresis effects in perovskite-based solar cells

Organo-lead-halide perovskite (OHP) based solar cells were reported to achieve energy-to-electricity power conversion efficiency (PCE) as high as 19.3%, which combined with reported methods for low-cost, exibility, and large-area solar cell fabrication such as ultra-sonic spray-coating and printing technology, makes OHP cell technology amenable to scaling up to production levels. The potential toxic effects of Pb have been discussed. Alternatives or solutions have been proposed, such as: Pb-free perovskite based solar cells, encapsulation, and Pb recycling. Methylammonium (MA) lead iodide (CH3NH3PbI3), the most commonly employed material in halide perovskite solar cells, was reported to have both a high absorption coefficient (direct bandgap of 1.55 eV) and high mobilities for electrons (7.5 cm V 1 s ) and holes (12.5–66 cm V 1 s ), resulting in long carrier diffusion lengths (100 nm to 1 mm). Although the amount of incorporated Cl is still under debate, mixed methylammonium-lead halide MAPbI3 xClx is another type of halide perovskite reported with an even higher charge-carrier mobility ( 33 cm V 1 s ), resulting in carrier diffusion lengths of up to 3 mm. Despite all the superb properties, perovskite solar cells suffer from a strong hysteresis in current– voltage (I–V) measurements typically conducted under AM1.5G illumination. Hysteresis in these cells is strongly inuenced

Organo-lead-halide perovskite (OHP) based solar cells were reported to achieve energy-to-electricity power conversion efficiency (PCE) as high as $19.3%, [1][2][3][4][5][6][7][8] which combined with reported methods for low-cost, exibility, and large-area solar cell fabrication such as ultra-sonic spray-coating 9 and printing technology, 10 makes OHP cell technology amenable to scaling up to production levels. 2,5The potential toxic effects of Pb have been discussed.][13][14] Methylammonium (MA) lead iodide (CH 3 NH 3 PbI 3 ), the most commonly employed material in halide perovskite solar cells, was reported to have both a high absorption coefficient (direct bandgap of $1.55 eV) and high mobilities for electrons (7.5 cm 2 V À1 s À1 ) and holes (12.5-66 cm 2 V À1 s À1 ), resulting in long carrier diffusion lengths (100 nm to 1 mm). 15Although the amount of incorporated Cl is still under debate, 16 mixed methylammonium-lead halide MAPbI 3Àx Cl x is another type of halide perovskite reported with an even higher charge-carrier mobility ($33 cm 2 V À1 s À1 ), resulting in carrier diffusion lengths of up to 3 mm. 179][20] Hysteresis in these cells is strongly inuenced by the perovskite grain size and the structure of the underlying TiO 2 layer, which is used as an electron transport layer (ETL).Efforts now are concentrated on generating perovskite lms with larger grain sizes to minimize the hysteresis. 4,19,21,22In addition, the replacement of MA + with cations that have a larger radius, such as formamidinium (NH 2 CH]NH 2 + , FA + ) 23 and 5ammoniumvaleric acid, 10 have been reported to help reduce the hysteresis effects.The planar heterojunction architecture employing a compact-TiO 2 (c-TiO 2 ) layer is of particular interest due to its simple cell conguration.9][20] Hysteresis effects in perovskite solar cells are generally associated with a high capacitance (of the order of mF cm À2 ) compared to Si cells (of the order of mF cm À2 ) 19,[23][24][25][26] or other types of solar cells (as described further in the ESI †).0]27 In this work, we performed a series of staircase voltage sweep measurements [18][19][20] at three different temperatures (250 K, 300 K, and 360 K) to quantify the photocurrent transient behavior in a complete perovskite cell composed of FTO/c-TiO 2 /perovskite/spiro-MeOTAD/Au where FTO stands for uorine doped tin oxide.The hole-collection took place through the spiro-MeOTAD hole transport layer (HTL).Our photocurrent data suggest that multiple processes are responsible for the complex transient behavior.9][30][31] Thus, device performance characterization of perovskite cells under various conditions is of paramount importance.Moreover, it can provide fundamental understanding of the charging-discharging processes in perovskite cells when studied in a controlled environment (temperature, humidity, and the gas environment). 31Temperature-dependent studies of steadystate I-V curves provide trends regarding the open-circuit voltage (V oc ), short-circuit current (I sc ), and ll factor (FF) electrical parameters as a function of temperature, shedding light on the physical processes taking place within the layers of perovskite solar cell.These results highlight the importance of establishing a protocol for the precise and artifactfree evaluation of perovskite solar cells.
3][34] Sample preparation details are described in the ESI.† Sample characterization by X-ray diffraction (XRD) and scanning electron microscopy (SEM) is shown in Fig. 1.SEM reveals complete surface coverage and the formation of interconnected crystal domains.The determination of grain sizes was difficult based on those images.A lm thickness of $380 nm was determined by a prolometer.XRD data showed the characteristic peaks at 14.1 , 28.5 , and 43.3 corresponding to the (110), (220), and (330) planes in the perovskite structure. 3The as-prepared sample was loaded into a variable temperature probe station chamber (Lake shore, CRX-6.5K)coupled with a pumping station (HiCube 80, Pfeiffer).The pressure inside the probe station was observed to vary between $2 Â 10 À6 and 4 Â 10 À5 Torr depending on the working temperature range (between 250 K and 360 K).Electrical signals from the probes contacting the FTO and Au electrodes were recorded using a semiconductor characterization system (4200-SCS, Keithley).The low-noise level (<1 pA) of our photocurrent data allowed us to unambiguously detect the different charge transport processes.It was difficult to couple a solar simulator with the probe station.Thus, light illumination on the perovskite samples was conducted using a tungsten halogen lamp.The relative light intensity of $9% compared to the standard AM1.5G was roughly estimated based on measurements on a reference Si solar cell.This light intensity was strong enough to induce hysteresis effects (photocurrent in the mA range) in perovskite solar cells during the current-voltage step (I-V step ) measurements.The perovskite cells were measured under vacuum conditions at three different temperatures (250 K, 300 K, and 360 K).The measurements under vacuum resulted in V step curves with high stability.No inuence of the pre-illumination step was observed in our studies. 20,30It is most likely that the photodoping effect is inactive in our measurements because our samples were measured under vacuum, i.e., the inuence of reactive gases such as O 2 and/or H 2 O is minimal. 20,30,31he I-V step measurements were conducted at three different temperatures (250 K, 300 K, and 360 K) with V step sweeping from À1 V to +1 V and from +1 V down to À1 V in steps of AE0.1 V (Fig. S3-S5 †).To make a convention for this work, we set the voltage sweep starting from the negative to the positive voltage to be the forward scan direction, and the voltage starting from positive to negative to be the reverse scan direction.In the I-V quadrant where the solar cell outputs electrical power (Fig. 2), the photocurrent response shows transient behavior at all temperatures.Every time the voltage is increased or decreased by 0.1 V, a sharp initial (at t 0 ) jump in photocurrent is followed by the subsequent rise or decay.20]27 The capacitive effect is enhanced under illumination conditions (photocurrent, Fig. 2) and shows a minimal inuence under dark conditions (dark current, Fig. 2).It has been previously reported that lead halide perovskites (e.g., prepared from PbCl 2 and MAI precursors) show a giant dielectric constant (GDC) phenomenon with a low frequency dielectric constant of the order of 3 $ 1000 in dark conditions. 25More interestingly, under the illumination of one sun, 3 was reported to increase by a factor of $1000.The origin of GDC and 3 enhancement upon light illumination is still unclear at the moment, 25 but it has been proposed that it may originate from the ferroelectric response induced by the photogenerated carriers leading to the structural rearrangement of methylammonium ions (MA + ). 20,25,27As a consequence of this structural phase transition upon light illumination, charge transport is affected, and therefore enhanced photocurrent transients are produced within the perovskite lm. 20,25,35It is also found that at a lower temperature (250 K, Fig. 2a and b) and room temperature (300 K, Fig. 2c and d), the photocurrent transient period appears to be longer compared to that of a higher temperature (360 K, Fig. 2e and f).This may be related to the effects of capacitance at selective-contacts 25,36 and will be discussed in more detail later.Additional transient behavior, which was generally observed above the open circuit voltage, 20 including its effects on the photocurrent corresponding to an applied bias of 0.7 V at 360 K (Fig. 2e) will not be discussed because it is beyond the scope of this work.
The photocurrent signals are modelled by using exponential functions.We observe that none of our raw data can be tted with a mono-exponential function.Multi-exponential functions, eqn (1) and (2), with three terms (i ¼ 0, 1, 2) are needed to reproduce the raw data, i.e., giving a reasonably low c 2 .The transient photovoltage decay technique is widely used to describe the physical processes of charge dynamics in solar cells.However, different descriptions for the transient signal are reported in the literature.For some perovskite solar cells, a mono-exponential function 37,38 was used, whereas for others, a bi-exponential function 23,39 was employed.This shows the complexity of the system under analysis, because the chargingdischarging characteristics in the cell can be strongly inuenced by the cell architecture and/or preparation conditions. 23onsidering the previously discussed capacitance factors 27 in perovskite solar cells, semi-logarithmic plots of time-dependent photocurrent signal as a response to the applied voltage step (Fig. 2) are expected to generate a straight line aer subtracting the photocurrent under steady-state conditions (t / N). 20 where s i is the time constant of the system.Thus, as shown in the eqn ( 1) and ( 2 The semi-logarithmic plots show at least two regimes with characteristic time decays.In the rst time interval (t < $1.8 s for 250 K and 300 K and t < $1 s for 360 K, Fig. 3), aer applying the voltage step, several fast charging-discharging processes (more than one because of the non-linearity observed in the semi-logarithmic plot) are inferred and designated as s fast .A quantitative description of this s fast (not a single time constant) is not possible from our experimental data because of the convolution of multi-exponential terms, eqn (1) and (2).However, considering the slowest process(es) in the rst time interval, s fast is observed to decay faster at a high temperature (360 K) contributing less to hysteresis effects.A more elaborate equivalent circuit with RC components has been proposed by Dualeh et al. 39 and our data could be associated with (i) selective-contact resistance(s) and capacitance(s) within the FTO/c-TiO 2 /perovskite/spiro-MeOTAD HTL/Au layers or (ii) due to the conduction of ion(s) within the perovskite lm 18,20,39 and/or HTM (e.g., Li + from Li-TFSI dopant 39,40 ) inuenced by temperature.TiO 2 is a semiconductor material that manifests a sizedependent capacitive effect in this nanostructured form, showing enhanced chemical capacitance (C m ); a smaller C m is expected for our c-TiO 2 , but may still provide some inuences on the charging-discharging processes. 26,29,36In the second time interval (t > $1.8 s for 250 K and 300 K and t > $1 s for 360 K, Fig. 3), a linear regime prevails with the slowest decay time of the system designated as s slow .23,37 Least-square tting method was applied on the second time interval of the experimental data to extract the s slow values.In the particular cases of applied voltage steps shown in Fig. 3, s slow ¼ 5.15 AE 0.02 s (Fig. 3a) [and 3.25 AE 0.01 s (Fig. 3b)] at 250 K, 7.49 AE 0.03 s (Fig. 3c) [and 3.23 AE 0.03 s (Fig. 3d)] at 300 K, and 5.6 AE 0.2 s (Fig. 3e) [and 2.01 AE 0.04 s (Fig. 3f)] at 360 K were extracted from the tted slopes in forward [and reverse bias].A summary of the extracted s slow values for all solar cell working bias conditions (between 0 V and V oc ) are shown in Fig. 4 and tabulated in the ESI (Table S1 †).9][20] It has been shown that the transient time period is strongly inuenced by (i) the perovskite crystal size (e.g.21]26 Our SEM images (Fig. 1a) reveal that our perovskite lms are composed of interconnected crystallites with smaller and larger grains fully covering the underlying c-TiO 2 layer.The determination of grain sizes is difficult based on the SEM images.No drastic variations in s slow were observed when the sample was held at three different temperatures of 250 K, 300 K, and 360 K.A slightly faster s slow of $2.5 s is observed when the sample is held at 360 K and the voltage is swept in the reverse direction (Fig. 4c).s slow is proportional to the perovskite capacitance, and therefore to the dielectric constant of the perovskite lm.2][43] For example, in the particular case of CH 3 NH 3 PbI 3 crystals, a steep discontinuity of 3(T) at the orthorhombic / tetragonal phase transition (161.4K) and a smooth 3(T) transition in the tetragonal / cubic phase (330.4K) were reported. 42Although, the above description provides an idea of the phase transitions and expected relative 3(T) values as a function of temperature, a direct comparison with our perovskite material is not valid because our study was conducted on a thin lm ($380 nm) that differs signicantly from the single crystal.In addition, the presence of Cl in our lms might have had a signicant inuence on the structural phases and transition temperatures. 16,23,44,45Based on the XRD results (Fig. 1b), our perovskite lm exhibits an orthorhombic structure 3 measured at room temperature and no large variations in the dielectric constants at temperatures of 250 K and 360 K can be inferred. 43Note that the s slow values differ when the voltage is swept in the forward or reverse direction, suggesting a strong dependence on the applied voltage history (remnant polarization), which is consistent with ferroelectricity-inuenced behavior. 27g. 3 Semi-logarithmic plot of the photocurrent responses as a function of time after applying a voltage step at t ¼ 0 (a, c and e) from 0.3 V / 0.4 V (forward) and (b, d and f) from 0.5 V / 0.4 V (reverse) at (a and b) 250 K, (c and d) 300 K, and (e and f) 360 K.The spectra show that multiple processes, designated as s fast and s slow , take place during the charging-discharging within a perovskite solar cell.The degree of hysteresis depends on the waiting period to measure the photocurrent after applying a voltage, i.e., the delay time (t d ).
Transient photocurrent response, as a consequence of capacitive effects in perovskite solar cells, has a strong inuence on the shape of the I-V curves.Based on the time-dependent results shown in Fig. 2, we can deduce the I-V curves (Fig. 5) with the different delay times (t d ), as indicated in Fig. 3a.The I sc , V oc , and FF extracted from the I-V curves are summarized in the ESI (Table S2 †).The I-V curve shapes are signicantly inuenced by the voltage sweep direction and t d .Overall, low photocurrent and high photocurrent are measured in the forward and reverse directions, respectively, when t d is small.Upon increasing t d , a rise and decay in the photocurrent are observed in the forward and reverse scan directions, respectively, eventually reaching the steady-state photocurrent conditions, I(t d / N).In the particular case of the cell measured at 300 K, a hump is observed at $0.5 V in the reverse direction (Fig. 5d) when the voltage is swept too fast (t d $ 0-40 ms).This is an artifact of the solar cell measurement conditions because the photocurrent has not reached the steady-state value. 24,46The delay time in Fig. 4 provides a lower bound for the solar cell photocurrent to attain its steady-state values.Considering that a perovskite solar cell is modelled with RC circuits, 25,26,34,36,39 the time constant (s ¼ RC) characterizes the capacitor charging to $63% between the initial and nal values (forward) and discharging down to $37% of the initial value (reverse).To reach a photocurrent level that is within 1% variation from the steady state, it takes a delay time of 5s slow , which is on the order of $30 s. [18][19][20] This is impractically long and new procedures are needed.At each temperature, comparing only the I-V curves corresponding to the steady-state conditions (t d / N) in forward and reverse scans, we notice that none of the curves match each other exactly.A large discrepancy in the I sc values between the forward and reverse I-V curves is noticed at low (250 K, Fig. 5a  and b) and high temperatures (360 K, Fig. 5e and f) while the I sc values are closely matched at 300 K 19,20 (Fig. 5c and d and Table S2 †).For example, at 250 K, the steady-state I-V curves in the forward and reverse directions provide I sc (V oc and FF) values of 5.8 mA (0.84 V and 0.24) and 14.9 mA (0.90 V and 0.18), respectively.Meanwhile, at 300 K, the forward and reverse I sc (V oc and FF) values are 15.1 mA (0.77 V and 0.35) and 14.4 mA (0.77 V and 0.49), respectively.Finally, at 360 K, the forward and reverse I sc (V oc and FF) values are 14.2 mA (0.54 V and 0.51) and 18.3 mA (0.54 V and 0.49), respectively.Excess I À or MA + ions are proposed to be present as interstitial defects induced by photoexcitation and/or during sample preparation conditions.Under reverse bias with illumination, the negatively charged interstitial anions (I À ) migrate to the c-TiO 2 layer and cations (MA + ) towards the spiro-MeOTAD HTL.The ionic concentrations induce energy barriers for electrons and holes at their respective selective-contacts.Forward bias conditions drive these interstitial ions in the opposite direction.The difference in the redistribution of ions within the perovskite cell when operated under forward and reverse bias conditions can explain why the steady-state I-V curves measured in the forward and reverse scan directions (Fig. 5) do not match exactly.Accurate photovoltaic parameters (V oc , I sc , and FF) can only be analyzed if the I-V measurements are carried out under steady-state conditions. 20nterestingly, V oc decreases as the cell temperature is increased (Fig. S12 †).In addition, the series resistance is signicantly enhanced when the cell is held at 250 K.These trends may have the same origin as previously described in solid-state dye sensitized solar cells (ssDSCs) 28,40 and more recently highlighted by the strong inuences of the selective contacts on the performance of perovskite solar cells. 36,47Based on a systematic study varying the combinations of TiO 2 electron transport layer (E), perovskite (P), and spiro-MeOTAD hole transport layer (H) architectures (E/P/H, E/P, P/H, and P), impedance spectroscopy measurements by Juarez-Perez et al. showed that the photovoltaic parameters (I sc , V oc , and FF) are strongly affected by the hole selective contact. 36Thus, the observed decrease in I sc in our perovskite cell at low temperatures is possibly due to the decrease in the hole mobility of the organic spiro-MeOTAD HTM. 36,47At the same time, the observed increase in FF at high temperatures (360 K) is consistent with the reduced series resistance due to the higher diffusion current. 36,47In parallel, the charge-recombination rate is expected to increase with temperature, inducing an overall decrease in the charge density within the cell, which reduces V oc . 36,47Based on the quantitative analysis shown in Fig. 5, the major inuence of temperature is on s fast processes (rather than on the s slow process), which may be associated with selective-contacts.Further investigation is needed to determine unambiguously the origin of s fast processes.

Conclusions
In summary, staircase voltage sweep measurements at 250 K, 300 K, and 360 K conducted on a perovskite solar cell reveal a complex time-dependent photocurrent transient signal.Our photocurrent data suggest multiple charging-discharging processes take place within the perovskite cell.Semi-logarithmic plots of the photocurrent responses reveal a linear regime showing the slowest transient process (well-dened mono-exponential trend) with a time constant (s slow ) of the order of seconds.This process was interpreted to have originated from the polarization response of the perovskite layer.Additional studies are needed to describe the convoluted s fast processes (multi-exponential terms), which had a stronger inuence on the temperature.I-V curves under steady-state conditions were composed from the transient photocurrent data.The hysteresis effect was smaller at 360 K and higher at 300 K and 250 K. On the basis of our study, in order to compare the results from different laboratories, it is essential to establish a protocol for extracting hysteresis-free I-V curves on perovskite solar cells corresponding to steady-state conditions.The extrapolation method used in this work to extract the steadystate photocurrents is suggested as a possible method.
), any eventual straight line in the ln Iðt 0 Þ À IðNÞ IðtÞ À IðNÞ (forward scan) and ln IðtÞ À IðNÞ Iðt 0 Þ (reverse scan) versus time plots will yield slopes with +1/s and À1/s values, respectively.The representative spectra are displayed in Fig. 3 for the cases of 0.3 V / 0.4 V (forward) and 0.5 V / 0.4 V (reverse) of applied voltage steps at three different temperatures.The complete set of individual I-V step analyses obtained at different temperatures is described in the ESI (Fig. S6-S11 †).

Fig. 2
Fig. 2 Photocurrent and dark current as a function of applied voltage using a staircase function generator.The voltage sweep from zero to positive voltage and from positive to zero are denoted as forward (a, c and e) and reverse (b, d and f), respectively.The 8 s long step delayed photocurrent responses are shown for three different temperatures: 250 K (a and b); 300 K (c and d); 360 K (e and f).The steep jumps observed in the current signals after each step-voltage are signatures of the capacitive effects in the device.

Fig. 4
Fig. 4 Extracted time-constants corresponding to the slowest process (s slow ) observed in the semi-logarithmic plot of the transient part of the photocurrent response upon applied voltage.

Fig. 5
Fig. 5 Photocurrent versus voltage curves composed from the time-dependent data shown in Fig. 2 considering the different delay times (t d ).The steady-state photocurrent values, I(t d / N), were calculated from the semi-logarithmic plots.