Interface Modification for Energy Level Alignment and Charge Extraction in CsPbI3 Perovskite Solar Cells

In perovskite solar cells (PSCs) energy level alignment and charge extraction at the interfaces are the essential factors directly affecting the device performance. In this work, we present a modified interface between all-inorganic CsPbI3 perovskite and its hole-selective contact (spiro-OMeTAD), realized by the dipole molecule trioctylphosphine oxide (TOPO), to align the energy levels. On a passivated perovskite film, with n-octylammonium iodide (OAI), we created an upward surface band-bending at the interface by TOPO treatment. This improved interface by the dipole molecule induces a better energy level alignment and enhances the charge extraction of holes from the perovskite layer to the hole transport material. Consequently, a Voc of 1.2 V and a high-power conversion efficiency (PCE) of over 19% were achieved for inorganic CsPbI3 perovskite solar cells. Further, to demonstrate the effect of the TOPO dipole molecule, we present a layer-by-layer charge extraction study by a transient surface photovoltage (trSPV) technique accomplished by a charge transport simulation.

All chemicals are used as received.

FTO substrates cleaning
Patterned FTO substrates (TEC 15, Yingkou company, with dimensions 2.5 cm 2.5 cm) were × numbered on the back side (glass side).These numbers were designated to every device (e.g.Z1, Z2, Z3, etc.).Substrates were cleaned with 2% Mucosal solution with a very fine brush to clean the FTO surface, then washed with distilled water to remove soap contents.Afterwards, cleaned with acetone and isopropanol for 15 min by sonication.After drying with a nitrogen gun, the substrates were placed in a UV-ozone cleaner for 15 minutes right before the titanium oxide layer deposition.

Solution preparation
 2% Mucasol solution was made by mixing 20 mL mucosal in 1000 ml distilled water. TiO 2 solution was prepared by adding 150 µL TIAP into 15 mL of ethanol. 1.0 M PbI 2 solution was prepared by dissolving 1.17 g of PbI 2 solution in 2.422 ml DMF solvent.The mixture was stirred at 80 °C for 2 hours to get the PbI 2 solution. To get 0.60 M CsPbI 3 solution, 1.987 ml PbI 2 solution was added in 0.4370 g CsI salt and stirred for 10 min until it completely dissolved. To make 1:1:1 (atomic ratio) CsI: PbI 2 : DMAI solution, 2.542 ml above CsPbI 3 solution was added in DMAI and stirred for 10 min until a clear, yellowish perovskite solution was formed. 45 mM MACl solution was made by dissolving 57.7 mg MACl salt in 19 mL IPA and was stirred for two hours until it completely dissolved. OAI solution was prepared by dissolving 3 mg of OAI in 1mL IPA and stirring for 1 hour. 20 mM TOPO solution was made by dissolving 15 mg TOPO in 1.940 ml toluene and other concentrations were made by using the dilution formula that is M 1 V 1 = M 2 V 2 . A 36 mM solution of spiro-OMeTAD was prepared by dissolving 200 mg spiro-OMeTAD in 2.2 ml chlorobenzene with 87.78 μL tBP, 51.11 μl LiTFSI with a stock solution of 520 mg/ml in acetonitrile, and 22.22 μl FK209 with a stock solution of 375 mg/ml in acetonitrile.

Device Fabrication
TiO 2 compact layer TiO 2 compact layer was deposited by spray pyrolysis with oxygen as the carrier gas.16 substrates were placed on a hot plate fitted inside a fume hood.One edge of each substrate is covered by around 5 mm using a cover glass to keep the conductive FTO side exposed for low contact resistance.Then the substrates were heated up to 450 °C and were kept at this temperature for 15 min before and 30 min after the spray of the precursor solution.The whole solution was transferred into a spray nozzle and sprayed at roughly 20 cm away from the substrates with an inclination angle of 45 degrees, with at least 20 seconds of delay between each spraying cycle.Afterward, substrates were left to cool down to room temperature and then put in an ozone chamber for 15 minutes before perovskite film deposition.

Deposition of control perovskite films
After the ozone treatment, substrates were transferred into a glove box filled with nitrogen (O 2 <0.1 ppm, H 2 O <0.1 ppm).The substrates were placed on a hot plate at 70 °C for 5 min before perovskite deposition.80-100 µl perovskite solution was added and spin-coated quickly at 3000 rpm for 30 seconds.Then 350 µl MACl solution was dropped on the top and spin-coated for another 35 seconds.The wet films were then annealed in a dry air box with a relative humidity (RH) of ~ 1% for 1 min at 210 °C.Afterwards, substrates were transferred back to a nitrogen-filled glove box, where 100 µl OAI solution was dropped on the top and spin coated at 5000 rpm for 30 seconds, followed by annealing at 100 °C for 5 min.

Perovskite films with TOPO treatment
After the control perovskite films cooled down to room temperature, 200 µl TOPO solution at varied concentrations was deposited by spin coating at 5000 rpm for 30 seconds.No annealing is needed for this step.

Hole Transport Layer (HTM) deposition
100 µl spiro-OMeTAD solution was deposited by spin coating at 3500 rpm for 30 seconds.No annealing is needed for this step.Afterward, all the samples were transferred into a dry air box (RH ~ 0.1%) for oxygen soaking.

Deposition of metal contact
Gold was evaporated using a thermal evaporator under a vacuum of approximately 1*10 -6 pa.The deposition rate was programmed at 0.02 Å/s for the first 1 nm, 0.1-0.2Å/s for the following 5 nm, and then 0.5 Å/s until 20 nm and then 1Å/s for the rest of the deposition.Overall, it takes around 25 min for the deposition of 100 nm of gold.The active area of the device was 0.18 cm 2 defined by the shallow mask.

Solar cell characterization
The light source was provided by an Oriel LCS-100 class ABB solar simulator (1Sun, AM1.5G, 100 mWcm -2 ) installed inside a nitrogen-filled glovebox.Before the light J-V measurement, the light intensity was calibrated with a silicon reference cell (Fraunhofer ISE).A Keithley power meter (2400 SMU) was used for the bias application to solar cells for the J-V scans, programmed by LabView.The bias was applied to scan from 1.25 V to -0.1 V back and forth, with a scan rate of 200 mV/s and a step size of 0.02 V.

EQE measurements
EQE spectra were recorded with the TracQ-Basic software, connected to the light source (Oriel Instruments QEPVSI-b system integrated with a Newport 300 W xenon arc lamp) with an optical fiber.The spectrum of the light source was calibrated with a Si reference cell with a known spectral response before the measurement.The monochromatic light was provided by a Newport Cornerstone 260 monochromator with a chopping frequency of 78 Hz.

Absolute Photoluminescence Spectroscopy (PL)
A 445 nm CW laser (Insaneware) was used as the excitation source for the PL measurements with an optical fiber connected to an integrating sphere where samples were loaded.Samples for the PL measurements were encapsulated with a cover glass before being taken out of a nitrogen-filled glovebox.

Time-resolved photoluminescence spectroscopy (TRPL)
TRPL signals were acquired with a TCSPC system (Berger & Lahr) after excitation with a mode-locked Ti: sapphire oscillator (Coherent Chameleon) that provides a pulse-picked and frequency-doubled output, with nominal pulse durations ~ 100 fs and fluence of ~ 30 nJ/cm² at a wavelength of 470 nm.

X-ray and Ultraviolet photoelectron spectroscopy (XPS and UPS)
Ultraviolet photoelectron spectroscopy (UPS) was conducted using a monochromated helium discharge lamp (HIS 13 FOCUS GmbH, photon energy of 21.22 eV) in an ultrahigh vacuum system (base pressure of 1 × 10 -9 mbar).With a monochromator, the visible light was eliminated and UV flux was significantly reduced (attenuation by a factor of ca. 100 folds as compared to that of the standard helium lamp).X-ray photoelectron spectroscopy (XPS) was performed using a standard Mg Kα radiation (1253.6 eV, anode power of 20 W) generated from a twin anode X-ray source.All spectra were recorded at room temperature and normal emission using a hemispherical electron analyzer (SPECSPhoibos 100).The illumination experiments were conducted using a white halogen lamp (Solux MR16 4700K, 50 W, daylight rendering) during UPS measurements with an intensity of ca. 100 mW/cm 2 .The secondary electrons cutoff (SECO) spectra were conducted at a negative bias of 10 V.

Kelvin probe measurement for Work Function
The measurement of WF was performed by the non-contact and non-destructive Kelvin probe method, in which the sample and probe form a parallel plate capacitor. 1

Kelvin Probe Force Microscopy (KPFM)
KPFM was recorded in an nitrogen-filled glovebox on a Bruker MultiMode mircroscope.Pt-Ir-coated cantilever tips (Bruker SCM-PIT, f 0 = 75 kHz, k = 2.8 N/m) with a tip radius of 25 nm were calibrated with respect to freshly cleaved HOPG.Determination of the CPD values succeeded via Gaussian fitting of the histograms extracted from the (2 x 2) μm 2 images.

Transient surface photovoltage (tr-SPV) measurements
Charge extraction in the time range of 5 ns up to 0.5 s was studied by non-contact SPV measurements exited by 5 ns above bandgap laser (1.8 eV).We used fluences of 0.072 µJ, which corresponds to a carrier concentration of 3×10 15 cm -3 close to 1 sun operation conditions.Detailed SPV setup description is given in our previous study 1,2 .Contour plots were recorded with a tunable laser in the range 0.6-3 eV using fluences of 72 µJ to ensure a good signal-tonoise ratio.

Simulation of charge extraction and recombination
Eq. S1-6 describe the simulation model for charge separation, trapping, and recombination where n and p are the concentration of photo-induced electrons and holes.The constants K e and K h correspond to electron and hole injection rates from perovskite to HTM.The constant K eTiO corresponds to electron injection rates from perovskite to ETM (TiO 2 ).Similarly, K eb and K hb are reinjection rates of electron and hole to perovskite from HTM, which effectively include back tunneling/thermionic emission, and diffusion of the free carriers, as well as the drift of the free carriers due to the presence of the space charge.is radiative recombination   constant.N t and σ are concentrations and capture a cross-section of defects responsible for SRH non-radiative recombination.and characterizes the carriers' lifetime in HTM     and ETM.The system of the equations cannot be solved analytically; therefore, we used the Adams backward differentiation formula (BDF) solving algorithm.We used the Levenberg-Marquardt method to fit constants with minimal deviation from experimental SPV results.SPV data were extrapolated logarithmically for better fitting results.The results of the fit are given in Fig. S15 and summarized in Table S7.
Due to , , and assuming a uniform distribution of charges in   ≫     ≫   perovskite, ETM, and HTM layers; so SPV signal can be simplified in the form:

Scanning electron microscope (SEM)
The SEM images were recorded with the Hitachi S-4100 at an acceleration voltage of 5 kV.

X-ray diffraction (XRD)
X-ray powder diffractometer Bruker D8 Advance in Bragg-Brentano geometry with Cu Kα as the target and LYNXEYE as the detector was used for the XRD measurement, at a voltage of 20 kV and current of 5 mA, The samples were scanned from 5° to 70° with a step size of 0.01°.

UV-vis spectroscopy
Perkin Elmer LAMBDA 1050 UV/VIS spectrometer was used for the transmittance measurements of samples.Samples were encapsulated with a cover glass before being taken out of a nitrogen-filled glovebox for the UV-Vis measurements.

Ageing of Solar Cells (long-term stability measurement)
Solar cells were aged in a custom-built High-throughput Ageing Setup. 3 A light-cycling experiment according to ISOS-LC-1I 4 was performed with cycles of 12 h illumination phase + + (Eq.S7) - -    0 followed by 12 h of dark phase.During the illumination phase, special electronics were used to MPP-track all cells individually.A perturb and observe algorithm 5 with a delay time of 1 s and a voltage step-width of 0.01 V was applied to track the MPP.PCE MPP values were taken every 2 min for all cells automatically.During the dark phase, cells were fully shaded with an automatic shutter system and disconnected from the MPP trackers.Additionally, JV-scans, with a scan speed of 90 mV/s, were performed on every cell after 11 hours of the light phase of a cycle.During the dark phase, the shutter was shortly opened to perform JV-scans on selected pixels after 11.5 hours of darkness.Devices were always kept at 25 °C with the help of actively controlled Peltier elements.Solar cells' active areas were touching a heating pad for direct thermal coupling.Aging was performed under a continuous flow of nitrogen in a closed box, no additional encapsulation was used.Sunlight with 1 sun intensity was provided by a metal-halide lamp using a H6 filter.
A UV-blocking foil was used to block UV light with wavelengths below 380 nm.Fig. S24 shows the spectrum of the light source in comparison to AM1.5G.The light intensity was actively controlled with the help of a silicon irradiation-sensor which was calibrated using a KG3 silicon reference cell from Fraunhofer ISE.

Supplementary data
Dipole formation at the interface between TOPO and perovskite Scheme S1.Representation of TOPO and the relative dipole formed at the interface with perovskite.The dipole is defined as pointing to the positive side.S1): The molecules OAI and TOPO are being used as passivating layers for the perovskite surface separately.In our work, we have used OAI and TOPO together in our champion device.

Inorganic nature of CsPbI 3
We first confirmed the crystal phase structure of our bare perovskite films via XRD.As shown in Fig. S1a, the sample presents dominating diffraction peaks at 14.35° and 28.86°, which corresponded to the (100) and ( 220) facets of γ-CsPbI 3 (Pbnm space group) crystallites 18 .The two broad peaks at 9.73° and 21.57° are likely to be (100) and ( 111) facets of δ-CsPbI 3 (yellow phase) with amorphous features 18,19 .This could be caused by the slow penetration of moisture into the dome filled with nitrogen during the XRD data measurement.The diffraction peak of DMAPbI 3 at 2theta of around 12° as reported in reference 20,21 was not observed here, indicating that the DMA + cation mostly vanished after the annealing.
We then conducted XPS measurement to further examine if any trace amount of organic moieties were left from DMAI or MACl in the pristine CsPbI 3 perovskite.Fig. S1b shows no signals for N 1s peak at the surface (< 10 nm) of CsPbI 3 thin films.
Together with the XRD data that revealed the absence of DMAPbI 3 in the bulk of CsPbI 3 thin films, we proved the chemical nature of CsPbI 3 inorganic perovskite.

Optical bandgap of CsPbI 3
An optical bandgap of around 1.70 eV is observed from the Tauc plot and the EQE spectra, which is a little higher than the reported bandgap for γ-CsPbI 3 , i.e. 1.69 eV.This is likely due to slow degradation in samples during the storage and transportation as reported in reference 18 . a)

How to measure penetration depth for perovskite thin film:
The penetration depth can the perovskite film can be calculated by the following relation; Whereas, the absorption coefficient (α) can be defined as, The absorbance (A) of the light in a material has a logarithmic relationship to the transmittance (T).
= - 10  Eq.S10 In the experiment, absorbance (A) was calculated by transmittance (T) measurements of the thin films (measured by UV-Vis spectroscopy) as mentioned in Figure S2a while thickness (d) of the film was measured by cross-sectional SEM of the perovskite film i.e. ~300 nm.The photon energy at every wavelength can be described as; The penetration depth at every wavelength can be measured by plotting penetration depth vs. hv as shown in Fig S2 d.
The Tauc plot method is given by Equation S12, in which, is plotted against energy (ℎ ) 2 and the linear segment is fitted to calculate the band gap (E g )at the x-axis. 22,23ℎ) Eq. S12

Film morphology
We noticed barely any difference in the grain size and film thickness of CsPbI 3 before and after TOPO treatment in Fig. S3 .Yet, the surface morphology changed slightly due to the coverage  (Eq.S13)  0, = •∫  ()•ɸ  ()• where J 0, rad is the radiative thermal equilibrium recombination current density in the dark at room temperature (at 2.5 10 -22 Am -2 ), q is the elementary charge, EQE (E) is the EQE spectrum as a function of photon energy (E), is the black body spectrum, given in the following   equation: (Eq.S14) Where h is the Plank constant, c is the speed of light in vacuum, and k B T is the thermal energy at room temperature.
QFLS is calculated from the following equation: (Eq.S15)  =     ln (    0, + 1 ) +     ln () Where J G is the photogenerated current density, taken as J sc measured from J-V curves.Here we observed that our TRPL data in Fig. 1b, Fig. S7 b,d, and Fig. S8 showed a biexponential decay instead of a mono-exponential one.It indicates that the bimolecular electronhole recombination is dominating in the PL decay of our samples while the trap-related monomolecular recombination lifetime is rather long, up to a microsecond. 24,25,26Thus, we fitted the spectra with a bi-exponential equation given in the following: (Eq.S16) Table S2.Fitted parameters of TRPL spectra in Fig. 1b and Fig. S7 b,d Comparing the τ 1 extracted from Fig. S7 b and d, the value is dropped by approximately one order of magnitude, indicating the τ 1 related decay happened much faster in the samples with ETM.Meanwhile, the weight (A 1 ) of the first exponential decay over the whole PL decay curve increased in Fig. 1b, indicating that τ 1 related decay contributes to a large portion of the PL decay.We believe the difference comes from the presence of the electron contact layer in samples presented in Fig. S7 d.For neat perovskite films without ETM or HTM, the PL decay comes from either the trap-related monomolecular recombination or the electron-hole-related bimolecular recombination.In our trPL data, the τ 1 reflects the decay due to the electron-holerelated bimolecular recombination while the τ 2 reports the decay coming from the trap-related monomolecular recombination.The long monomolecular lifetime of up to 3.5 μs extracted from our work agrees with the literature. 24,25,26th the ETM adjacent layer, the photo-generated electrons and holes in the perovskite film can experience bimolecular recombination and trap-inducted monomolecular recombination, and additionally, electrons be extracted at the interface.Each of the three paths will lead to a decay in the PL signal.It is likely that for perovskite with ETM, τ 1 relates to the electron extraction reduced PL decay because we did not observe such a fast decay in neat perovskite films.τ 2 might connect to the trap-related monomolecular recombination or the electron-hole bimolecular recombination.If it reflected the decay due to monomolecular recombination, then the value is reduced by nearly two orders of magnitude compared to the neat perovskite samples.This would mean a large increase in trap densities because of the introduction of the ETM adjacent layer.However, the contour plots of the transient SPV in Fig. S13 reflect no formation of trap states below the bandgap of CsPbI 3 perovskite deposited on TiO 2 -ETM.Thus, we believe that τ 2 is not likely related to the trap-related monomolecular recombination.Here we assign it to the decay process caused by electron-hole bimolecular recombination.The value falls in the similar range of the τ 1 in neat perovskite samples.

Work function shift
We further confirmed the shift in WF in our samples using the Kelvin probe.Fig. S12a shows the average WF of these samples given with an error bar.We observe that bare CsPbI 3 film has an average WF of 4.35 eV, reduced to 4.18 eV for control samples, and further reduced to 3.9 eV after TOPO treatment.The energetic level scheme is presented in Fig. S12b, with values summarised in Table S5.

Table S6. Additional constants for simulations
Where N, P, C b , σ e , σ h , v e , and v h are photo-induced electron concentration, photo-induced hole concentration, radiative recombination constant, electron capture-cross section of trap, hole capture cross-section of trap, thermal velocity of electron and hole, respectively.ε per , ε ETM, and ε HTM are dielectric constants.τ HTM and τ ETM are lifetimes in HTM and ETM.The concentration of free carriers (N, P) was calculated according to the fluence of 0.010 μJ/cm 2 .Constants C b , v e , v h , and ε per , were adapted from our previous study 2 .Dielectric constants ε ETM and ε HTM were adapted from literature 31,32 .Constants τ HTM , τ ETM , K eETL , σ e , and σ h are fitted directly and are in agreement with previous reports 2,32,33 .

N (P), cm -3
C b , cm 3 s -1 K eETL , 10       Long-term stability data-II (Figure S21 a-c) An analysis based on the distribution of maximum PCE.From the overall 24 solar cells (six on each substrate), we selected 18 out of them following the protocol given below.We refer to this method as the "PCE distribution filter".a) Distribution of maximum PCE for the control and TOPO-treated samples.Maximum PCE was taken from the first illumination cycle.Counts PCE (%) TOPO-maxPCE 24 solar cells b) We selected the solar cells that present an efficiency within the top 75% that covers pixels from each substrate.For the control, it includes six pixels from sample Z11, five pixels from sample Z16, three from Z8, and four from Z10.For the TOPO-treated samples, it includes five pixels from Z7, four pixels from Z2, four pixels from Z6 and five pixels from Z13.Overall, 18 pixels out of 24 pixels are selected.Meanwhile, all the selected solar cells possess an efficiency of over 10%.The stability data with cycled illumination for the first round are given in the following.For the control samples, we have four pixels from sample Z11, four from Z16, three from Z8, and four from Z10.For the TOPO-treated samples, we have four pixels from Z7, three from Z2, four from Z6, and four from Z13.So, we have 15 pixels for the stability data for each condition.The results are as follows.The superior stability in topo-treated samples becomes dominant in the second-round experiment.This conclusion we reached based on the "PCE distribution filter" agrees well with what we discovered with the "best PCE+stability filter".

3 Fig
Fig. S1.a) XRD spectra of CsPbI 3 film deposited on TiO 2 -covered FTO.b).XPS of N 1s coreshell spectrum of as-obtained CsPbI 3 sample after annealing in a dry air box.
Photon energy (eV) of the TOPO layer deposited at high concentrations (particularly 20 mM).The film roughness was characterized by atomic force microscopy (AFM), with the topography images presented in Fig.S4.It shows a slight increase in film roughness, which is likely caused by the slight changes in surface morphology during the dripping process of solutions for the additional layers.

5 Fig
Fig. S12.a) Work function (Փ) of bare CsPbI 3 (black square), control CsPbI 3 (red sphere), and w/TOPO (blue triangle) with error bars (standard deviation, in unit of eV), measured by Kelvin probe technique.b) energy level diagram of these samples, taking the average value plotted in a).

FigureFig. S14 .
FigureS13shows SPV signals as a function of photon energy and time.Whereas SPV amplitude is presented as a color code in the right corner.The SPV signals are measured with light excitation by photon energy in the range of 0.7-3 eV.The excitation of the trap by light leads to charge separation.The signal below the band gap corresponds to the activation energy of the trap.SPV signal is induced by the diffusion of the excited free carrier out of a trap.The activation energy is the lowest photon energy that we observe in the SPV signals.A similar approach has been used, in previous studies28,29,30 to find the activation energy.

Fig. S16 .
Fig.S16.a) J-V curve of the champion devices for CsPbI 3 with and without TOPO treatment, measured at 1 Sun AM1.5G at room temperature inside the nitrogen-filled glovebox at the scan rate of 200 mV/s for both forward (dash line) and reverse (solid line) scans.b) EQE spectra of champion devices with the integrated J sc from EQE spectra overplotted on the right y-axis.c) Absorption onset of CsPbI 3 solar cells extracted from derivated EQE spectra.

Fig. S17 .
Fig. S17.Box charts of a) PCE, b) J sc , c) V oc, and d) FF of more than 40 solar cells ("shunted/dead" solar cells data are removed and not included).The data (shaded area) shows that there is an improvement in PCE and other parameters for TOPO-treated samples.

Fig. S18 .
Fig. S18.Distribution analysis of a) V oc b) PCE and c) FF of more than 70 solar cells for control and TOPO ("shunted/dead" solar cells data are removed and not included).Champion V oc devices the second round, three pixels were taken out of 18 pixels because of their low efficiency.

Fig. S22 .
Fig. S22.Device Stability: MPP tracking for control and TOPO-treated samples for 700 hoursTo further, compare the stability, we have conducted constant illumination aging testing for up to 700 hours.Fig.S22shows that both types of devices undergo burn-in initially and then get stabilized.Compared to the control, TOPO-treated devices tolerate the efficiency for a long time.

Fig. S24 .
Fig. S24.Spectrum of the lamp of the High-throughput Ageing Setup used to age solar cells in comparison to AM1.5G.In this work, a UV filter was used (blue curve). .

Table S4 .
Work function ( ), ΔE, and VBM of the following samples extracted from UPS

Table S5 .
The work function ( ) of the following samples was measured by the Kelvin Φ probe.

Table S7 Main fitting constants of simulation
. N non-rad -concentration of non-radiative recombination defects (both surface or bulk), K h -hole injection rate, K e -electron injection rate, K hb -hole reinjection rate, and K eb -electron reinjection rate.STD is the average standard deviation of the fit from the experimental SPV signal.

Table S8 .
Photovoltaic parameters summary for champion devices of CsPbI 3 perovskite." and "fw" indicate the reverse scan (bias sweeps from V oc to J sc ) and forward scan (bias sweeps from J sc to V oc .),respectively.