Effect of surface modification of nickel oxide hole-transport layer via self-assembled monolayers in perovskite solar cells

Self-assembled monolayers of para-substituted phenylphosphonic acids were used to modify the surface of nickel oxide layer, which served as the hole-transport layer in the fabrication of inverted perovskite solar cells. The monolayer was installed to modulate the work function of NiO x based on their electron-withdrawing or electron-donating substituent, while the perovskite film morphology and quality were not significantly altered due to analogous hydrophilic nature of the modified surfaces. The modification impacted the device performance such that the best-performing device was observed with electron-withdrawing cyano-substituted phosphonic acid modification, with a maximum power conversion efficiency of 18.45% obtained. The effect of modification on parameters such as open-circuit voltage, short-circuit current, and field factor are discussed.


INTRODUCTION
Since the last decade, the field of perovskite solar cell (PSC) has been at the forefront of the research of photovoltaics. For the first time in 2009, the perovskite material was used in a solar cell device by Miyasaka et al., yielding an efficiency of 3.8%. [1] Since then, there have been numerous attempts to improve the efficiency and the stability of PSCs, with the toxicity of the lead-based perovskite material as more recent concern. [2][3][4][5][6] For enhancing device efficiency, various approaches have been considered, for instance, changing deposition condition for improved morphology or changing active layer composition for more light absorption or reduced defect sites. [5] Besides active layer engineering, other approaches This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2021 The Authors. Nano Select published by Wiley-VCH GmbH include introducing additional layers to improve the charge transfer towards the respective electrode. [2] The additional layer could be another charge transport layer, or a surface modification layer to adjust the interfacial energy level alignment. [7,8] Self-assembled monolayers (SAMs) have been widely used as interface modifiers in various electronic devices. [9][10][11][12][13] Specifically, in PSCs, SAMs have also been used, though more often in the n-i-p device configuration, and considerable improvements in the performance were reported. [14][15][16][17] The SAM formation process typically involves immersion of the substrate in a solution of modifying molecules. [18][19][20] In the n-i-p configuration, the SAM is formed on the surface of electrontransporting layer, such as TiO 2 , to realize its effect on the device performance. [21][22][23] There are reports that the SAM modification influences the morphology of the perovskite layer. For instance, Zuo et al. reported a pin-hole-free perovskite film on a C3-SAM (3-aminopropanoic acid)-modified ZnO substrate when compared with an unmodified ZnO underlayer. [24] There are also reports where the SAM modification did not have any effect on the morphology of the perovskite film. [15] Hence, depending on the functionality of the SAM used, the effect of SAM modification can be varied, influencing the morphology/crystallinity or the energy level alignment, or both.
In the cells with p-i-n configuration, the commonly used hole-transporting layer (HTL) poly(3, 4ethylenedioxythiophene)polystyrene sulfonate (PEDOT: PSS) is acidic in nature and relatively unstable. The PEDOT:PSS is suggested to degrade the device performance. [25][26][27] Metal oxide layers as charge transport layers have shown promising improvement over their organic counterparts and specifically, in case of p-i-n configuration, nickel oxide-based HTL have been shown to have advantages in terms of stability and device performance. [28][29][30][31][32][33] These metal oxide layers can be readily modified through functional monolayers. [34,35] Among the possible modifiers, phosphonic acids are known to anchor sturdily on metal oxides as compared to benzoic acid, silanes etc. [36][37][38] Wang et al. modified the NiO x layer with a series of benzoic acid SAMs and observed an improved device performance in 4-bromobenzoic acid-based device. [35] However, they reported a large change in perovskite film morphology/crystallinity as the impact of the SAM modification. It is difficult to delineate whether the improved device performance is a result of better band alignment or changed perovskite morphology/crystallinity, or even different thickness as a result of varied surface wettability after the SAM modification. As the device power conversion efficiency is a composite result of open circuit voltage (V oc ), short circuit current (J sc ) and fill factor (FF), which can all affected by the morphology/crystallinity of perovskite and its contact with other layers and in turn be modulated by SAM modification, minimizing the number of variables involved may lead to better correlation. [39,40] Here in this report, we modified the nickel oxide surface with a series of para-substituted phenylphosphonic acid SAMs as the HTL in the fabrication of PSC to investigate their effect on the device performance. The different substituents (electron-donating CH 3 O-, H-and electronwithdrawing CN-groups) impart different molecular dipoles to the molecules so that when anchored to the NiO x surface, the work function (WF) of the NiO x HTL can be tuned with respect to the valence level of the perovskite layer. The substituents are chosen such that the hydrophilicity of the modified surfaces is not much different, so that their effect on other variables such as perovskite film thickness and crystallinity upon deposition can be minimized. Their effects on the device performance were evaluated and correlation with respect to the band alignment of NiO x or charge extraction was studied.

Materials
All chemicals used in the study are analytical grade.

NiO x preparation
The NiO x was deposited by a slightly modified procedure from that reported previously. [32,41] Thus, the precursor solution for nickel oxide was prepared by mixing 1 M equivalent of ethanolamine and nickel acetate tetrahydrate in ethanol and stirred for 24 hours at 50 • C, after which a homogeneous dark-green solution was formed. The solution was filtered using 0.45 µm PVDF filter. ITO substrates (1.5 cm × 1.5 cm) were cleaned step-by-step with detergent solution, DI H 2 O, ethanol, acetone in sequence, followed by UV-ozone treatment for 25 minutes. The NiO x precursor solution was spin-coated on ITO substrates at 3000 rpm for 30 seconds. The substrates were annealed at 250 • C, for 1 hour and then cooled to room temperature before further process.

SAM preparation
The NiO x -ITO substrates were immersed in 1 mM ethanol solution of different phosphonic acid molecules at 50 • C for 60 minutes. The substrates were then dried with N 2 gun before further heated at 150 • C for 3 hours to ensure covalent bond formation. Then the substrates were let to cool down naturally till room temperature, thoroughly rinsed with ethanol to remove any physisorbed multilayers and dried with N 2 gun.

Device fabrication
The device fabrication was carried out in a N 2 -filled glove box. Equimolar amounts of PbI 2 (461 mg), MAI (159 mg), and DMSO (72 µL) were dissolved in 630 µL of DMF and stirred for 8 hours at 70 • C. The clear golden-yellow solution was filtered using 0.22 µm PVDF filter and spincoated on the pristine or SAM-modified NiO x substrate at 4000 rpm for 25 seconds. Two hundred microliters of CB was dripped at 8-11 seconds on top of the substrate after the onset of perovskite precursor spinning. The substrate was then annealed at 70 • C for 1 minute and at 100 • C for next 5 minutes. A homogenous perovskite film has a shiny black texture. The next step was to spin-coat the electron transport layer. A 2% PCBM solution in CB (stirred at 70 • C for 8 hours) was spin-coated at 3000 rpm for 30 seconds and the substrate was subjected to annealing at 100 • C for 10 minutes. A 100 nm Ag film was deposited thermally through a patterned mask, leaving an active area of 10 mm 2 .

Characterization
Contact angle was measured by a Rame-Hart goniometer using sessile drop method. Atomic force microscopy (AFM) images were recorded using a multimode atomic force microscope (Digital Instruments, Nanoscope III) with tapping mode. X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) measurements were carried out using a PHI 5000 Versa Probe apparatus equipped with an Al Kα X-ray source (1486.6 eV). UPS was measured to acquire the work function with UV light source having He l emission at 21.2 eV, ca. 50W, with a take-off angle of 90 • . Work function measurement was additionally done using Riken photoelectron spectrometer AC2. The cross-sectional and surface morphologies of films were recorded using Ultra plus (a member of Carl Zeiss) Field-Emission Scanning Electron Microscope (FESEM) at an accelerating operating voltage of 10 keV. The X-ray diffraction (XRD) was recorded using a Bruker D8 Advance with LYNXEYE detector and Cu absorber. The PSCs were measured under 1 sun irradiation using a Newport 91192A source meter with an AM 1.5G light spectrum at an intensity of 100 mW cm -2 . A mono

RESULTS AND DISCUSSION
The SAM of phosphonic acid on NiO x was prepared by dipping the NiO x -ITO substrates in a solution of phosphonic acid molecules in ethanol for 1 hour. The phosphonic acid head group is known to bind to the oxide surface through the hydroxyl groups and the three acid molecules are expected to bind to the NiO x surface in a similar binding mode, with different tail groups ensuring different dipole moments (μ). [37] The reference molecule phenylphosphonic acid (PPA) has a relatively small μ as compared to other two molecules. The 4-methoxyphenylphosphonic acid (MPPA) molecule, which carries an electron-donating group, has a positive dipole pointing away from the surface. The 4-cyanophenylphosphonic acid (CNPPA) having an electron-withdrawing group has the dipole pointing toward the surface. The μ values for the three structures are tabulated in Table 1. [42,43] As a qualitative characterization of SAM formation on NiO x , CA measurement and roughness measurement using AFM were carried out on the pristine and the modified NiO x surfaces. As shown in Figure 1 and Table 1, the hydrophilicity of the three modified surfaces were rather similar, but less so than the pristine NiO x surface. Also, the surface roughness (R rms ) did not significantly change, as shown in Figure 1 and Table 1. Nevertheless, for all the SAM-modified NiO x substrate, the surface became smoother than pristine NiO x . These may or may not influence the morphology of the perovskite film to be coated on the surface, as will be discussed later. The UV-vis absorption spectra were also measured for the monolayermodified NiO x films (shown in Figure S1). The SAM has little contribution to the UV absorption of NiO x .
To further confirm the formation of SAM on NiO x surface, XPS measurement was performed on the modified surface. The peaks associated with Ni2p 3/2 , O1s, P2p and C1s are shown in Figure 2 and tabulated in Table 2. The spectra for Ni2p 3/2 and O1s are deconvoluted and the results are shown in Figure S2. The Ni2p 3/2 region has two peaks, one at 853.9 eV, corresponding to the electronic state of Ni 2+ and another peak at 855.4 eV corresponding to the electronic state of Ni 3+ (deconvoluted in Figure S2   under the peak at 854.3 eV). [44] The Ni2p  These changes indicate local environment alterations near the NiO x surface. [20] The peak at 861 eV indicates satellite peak associated with the main peak, occurring due to a sudden change in the Coulomb potential when the electron passes through the valence band. [45] The lower B.E. peak at ∼529 eV in the O1s ( Figure 2B) spectra for all conditions correspond to the lattice oxygen from the underlying NiO x . [37] In Figure S2, the deconvoluted peaks, around ∼530.2 eV is assigned to Ni 2 O 3 and ∼531.2 eV is assigned to NiOOH. [44] The O1s peak shows shifts in case of modified NiO x as compared to pristine NiO x . With PPA modifier as the reference, there is a relative shift of 0.2 eV towards lower B.E. in the O1s peak for MPPA. This lower B.E. shift suggests a higher electron density which is in accordance with the smaller WF of NiO x after modification by MPPA. For CNPPA modifier, a shift of 0.2 eV towards higher B.E. as compared to PPA, suggests a lower electron density. [20] Figure 2C shows the P2p peak which further affirms the presence of PA monolayer on the NiO x surface. The P2p peak is observed at 132 eV for PPA-modified surface, 132.1 eV for MPPA-modified surface and 132 eV for CNPPAmodified surface. The negligible shift in the peak position indicates similar binding mode for all molecules to the underlying oxide layer. The integration of the peak area, however, indicates the CNPPA-modified and MPPAmodified surface have lower coverage (22% and 9% less, respectively) than the PPA-modified surface. A possible explanation is that the much larger dipole moment for CNPPA and then MPPA may not favor a closely packed monolayer as PPA does, due to the electrostatic repulsion. [46] The shifts in C1s peaks ( Figure 2D) corrobo-rate with the expected chemical shifts due to differences in the modifiers and follows the same trend as that of O1s. [20] The additional C1s peak at 285.4 eV for MPPA-modified surface can be due to the methyl group of the OCH 3 moiety in MPPA.
The structure of the device used in the PSC is shown in Figure 3A. One of the major challenges of enhancing the charge flow in a PSC is to optimize the energy level alignment. In this study, the WF of the hole-transporting nickel oxide layer was modified using SAMs (inset Figure 3A) in order to elucidate its effect on the device performance. To analyze the energy levels after SAM modification, UPS measurements of pristine and modified NiO x were taken. The WF values, deduced from the low kinetic energy cut-off region ( Figure S3A) and the Fermi cut-off region ( Figure S3B) for pristine NiO x , PPA-, MPPA-, and CNPPAmodified NiO x , are 5.27, 5.23, 5.17, and 5.39 eV, respectively. Similar trend in WF was obtained using photoelectron spectroscopy in air (PESA) ( Figure S4 and Table S1). Thus, the PPA SAM slightly decreased the WF, whereas the MPPA SAM further decreased the WF, presumably owing to the stronger electron-donating property of methoxy substituent. In contrast, the CNPPA SAM increased the WF value due to the electron-withdrawing substituent. The highest occupied molecular orbital of the standard MAPbI 3 lies at -5.4 eV. [47] Therefore, the modification of NiO x by CNPPA SAM brings the WF of NiO x closest to the valence band of MAPbI 3 . The energy level alignment for different layers along with different modifiers is shown in Figure 3A.
After the SAM modification, the perovskite layer was prepared by spin-coating the precursor solution onto the pristine and differently modified NiO x surfaces and annealing. The SEM images of perovskite film surface on different substrates (shown in Figure 4) exhibit meager difference in film coverage.
The grain size distribution and average grain sizes are compared (shown in Figure S5 and Table S2). No significant difference in the average grain sizes (∼173-190 nm) was found. Further, the calculated full-width-at-halfmaximum (FWHM) of the (110) peak from the XRD pattern of perovskite films ( Figure S6 and Table S2) shows a similar trend in that smaller FWHM indicates better  crystallinity and larger grain size and vice versa. The cross-sectional images of perovskite films on pristine and modified NiO x surfaces ( Figure S7) show a thickness of approximately 310 nm in all the cases (Table S2). Hence, it can be inferred that the SAM modification did not alter the perovskite film formation much, possibly due to similar wetting property of the surfaces. From these measurements, it is suggested that in correlating the changes in performance, the factors of perovskite thickness, coverage, or crystallinity may not be that critical. Additionally, the thickness for NiO x and PCBM in all cases is ∼60 and ∼70 nm, respectively. To examine the effect of different SAM-modifications on the device performance, NiO x layer with and without SAM modification were used in fabricating devices with the structure: ITO/NiO x /SAM/CH 3 NH 3 PbI 3 /PCBM/Ag (Figure 3A). SAM modification changes the band edge of NiO x , which not only modulates the built-in voltage, but also other parameters. [48] Figure 5A and Table 3 show the J-V curves and performance characteristics of the champion cell for each condition.
The champion cell based on pristine NiO x has a PCE of 17.02% with an open circuit voltage (V oc ) of 1.02 V, shortcircuit current (J sc ) of 22.98 mA cm -2 and a fill factor (FF) of 72.61%. For the PPA-modified device, the V oc remained nearly the same as the pristine NiO x and is in agreement with the nearly unchanged work function of the nickel oxide after modification by PPA ( Figure 3B). For the MPPA modification, the V oc slightly decreased from 1.02 to 0.99 V, while for the CNPPA modification, the V oc increased from 1.02 to 1.06 V. The CNPPA-modified NiO x has a WF closer to the valence band maximum (VBM) of MAPbI 3 and MPPA-modified NiO x has a WF farther from VBM level of MAPbI 3 . The closer WF of CNPPA-modification provides a better built-in voltage, which is proportional to the difference in WF of the HTL and ETL. [16] The PPAand MPPA-modified devices had decreased J sc . Since the perovskite film has similar morphology and crystallinity, we attributed the decrease in J sc to the dipole direction of these molecule pointing away from the NiO x surface, thus inhibiting the flow of holes towards NiO x . The CNPPAmodified device had increased J sc , presumably due to a dipole directed towards the NiO x and thus facilitated the flow of holes to the NiO x surface. [49] The average fill factor was also improved with CNPPA-modification, indicating better interface and efficient charge transfer. [39,40] Overall, the observed trend in PCE was: MPPA-modified NiO x (12.08%) < PPA-modified NiO x (15.35%) < pristine NiO x (17.02) < CNPPA-modified NiO x (18.45%). As is clear from the analogous thickness of perovskite film on pristine and modified NiO x , the major contribution in the device performance arises from the band level alignment and dipole direction of the molecules. Better band alignment resulted in an increased V oc and the appropriate dipole direction resulted in an improved J sc and thus enhanced device performance. [49] The corresponding external quantum efficiency (EQE) spectra are shown in Figure 5B, along with the integrated J sc (values tabulated in Table S3), depicting a similar trend as that observed in the PCE. The enhanced EQE and J sc in the case of CNPPA-modified NiO x based device indicates improved charge transfer at the interface of NiO x and perovskite layer. [16] To further investigate the photo-carrier dynamics and observed efficiency trend due to modification of NiO x by SAM in PSC, steady state photoluminescence (PL) spectra of perovskite films on pristine and modified NiO x were measured, with the excitation wavelength of 470 nm and the range of detection from 700 to 850 nm. As shown in Figure 5C and Table S3, a clear difference in PL intensity was observed depending on the different underlayers on which the perovskite film was deposited. In case of perovskite layer alone with no NiO x layer underneath, the PL intensity is the highest (shown in Figure S8). With NiO x HTL and SAM-modified NiO x inserted underneath the perovskite layer, various extent of quenching of the PL occurred. A maximum quenching of ∼95% was observed in film incorporating CNPPA-modified NiO x . The quenching is suggested to be due to hole-charge transport from the perovskite layer to the NiO x layer so that radiative relaxation of the excitons to the ground state was diminished. On the other hand, PL quenching for MPPA-modified NiO x is the smallest (∼83%), which means there is least charge transport from perovskite layer to NiO x layer. The quenching increased in the order of MPPA-NiO x < NiO x ∼ PPA-NiO x < CNPPA-NiO x , roughly parallel to the trend in J sc . To further elucidate the trend in the PL quenching, we performed time-resolved photoluminescence (TRPL) measurements for the perovskite films on various NiO x substrates. The TRPL measurement was conducted using an exciton wavelength of 470 nm. The time-resolved PL decay plots (dotted plot in Figure 5D) were fitted using biexponential decay fitting (solid line in Figure 5D) and the parameters are tabulated in Table S4. The relatively faster decay components τ 1 is attributed to charge carrier lifetime at the interface of perovskite and the transport layers, whereas the slower decay component, τ 2 , is related to lifetime in the bulk of the material. Average lifetime is calculated using the equation , where is the relative contribution of each lifetime ( = ∕ ∑ 2 =1 ) in the bi-exponential decay model ( ) = + 1. − ∕ 1 + 2. − ∕ 2 used to fit the TRPL spectrum. The observed quenching trend in PL and lifetime trend in TRPL correlate well with the J sc . [15] Amongst all the conditions, CNPPAmodification enables the NiO x -Perovskite interface to have fastest charge extraction as indicated by the shortest τ 1 , τ 2 and τ avg . The shortest lifetime for CNPPA-modification may indicate reduced trap states/better interfacial defect passivation, or superior interface leading to an increased fill factor. This reduction in trap states or interfacial defect passivation may result from binding of the localized negative charge of the cyano group with the positive Pb ions in perovskite, as suggested in previous report. [50] Electrochemical Impedance Spectroscopy (EIS) was also used to evaluate the recombination in the devices. The EIS spectra were recorded in dark at open circuit voltage for each device. The Nyquist plot in Figure S9 shows that the device with CNPPA-modified NiO x exhibited the largest recombination resistance as compared to those devices with pristine or other modified NiO x . This indicates the reduction in recombination and efficient charge extraction and the trend observed was in agreement with our TRPL and PL quenching results.
The device data statistics was estimated (results shown in box plot in Figure S10). The CNPPA-NiO x and pristine NiO x -based devices show smaller distribution over the device performance, indicating better reproducibility. The dark storage stability of the devices was recorded. All the devices were stable and maintained similar performance for more than 50 days ( Figure S11). The light soaking effect was also measured for all device conditions as shown in Figure S12. All devices showed stable PCE for above 1000 seconds. Hence, SAM modification didn't have any adverse impact on the stability for the devices.

CONCLUSION
In conclusion, a systematic study of the effect of dipolebearing SAM modification of the hole-transport layer of NiO x on the performance of inverted PSC was conducted.
Molecules with different tail groups on the same phosphonic acid head group were used to modify the surface of nickel oxide layer. Phenylphosphonic acid used as the reference molecule did not significantly changes the WF of NiO x . Modification by the acid carrying an electrondonating methoxy group resulted in a lower value of work function, owing to the positive dipole directing away from the oxide layer. The opposite effect was observed on NiO x modified with CNPPA, which increased the WF as a result of a negative dipole directing towards the oxide surface. The modification resulted in similar hydrophilicity of the surface and no major difference in film quality and thickness of the perovskite layer deposited on top. This simplifies the variables from various SAM modifications. A correlation between energy level alignment after SAMmodification and the open circuit voltage values can be seen. The best performing device is the CNPPA-modified NiO x -based device as CNPPA aligns the WF of NiO x closest to the valence band of perovskite, thus giving a larger V oc and enhancing the charge transport from perovskite to NiO x . The dipole direction may also contribute to the improved J sc and the FF. More than 21% increase in the PCE was observed over the reference device based on pristine NiO x .

A C K N O W L E D G E M E N T S
The authors thank the financial support of this work from Ministry of Science and Technology, Taiwan, (Grant No. 109-2113-M-001-006).