Nanocomposite of nickel oxide nanoparticles and polyethylene oxide as printable hole transport layer for organic solar cells † SustainableEnergy&Fuels PAPER

We focused on the low-temperature fabrication ( (cid:1) 100 (cid:3) C) of nanoparticle-based NiO x hole transport layers for organic solar cells. We used high molecular weight polyethylene oxide (PEO) as a sacri ﬁ cial additive for the processing of NiO x nanoparticle (NP) dispersions which is removed from the ﬁ lm by oxygen plasma prior to the active layer deposition. Through comprehensive characterization by optical, electrical and scanning probe techniques we show that an optimal NiO x : PEO ratio allows to disperse the NPs, hindering their aggregation and maximizing device performance. Finally, we demonstrate the practical relevance of our approach by inkjet printing the NiO x :PEO blends to produce hole transport layers (HTLs) for P3HT:PCBM bulk heterojunction solar cells with comparable e ﬃ ciencies to devices produced with high-temperature precursor-based approaches. The utilization of P3HT:PCBM as active layer material is meant to o ﬀ er an optimum standard for the introduction of the presented HTL processing approach and its applicability to a wider range of hybrid systems and multilayer optoelectronic devices.


Introduction
In the eld of printable organic and hybrid electronics, increasing attention has been recently dedicated to novel interlayers that enable the optimal functioning of the electronic device. [1][2][3][4][5][6][7] These interlayers provide an improved energetic alignment of the device electrodes to the semiconducting active layers, reducing contact resistance effects and providing a more efficient charge injection/extraction. [8][9][10][11][12][13] In recent years, high work-function solution-processed transition metal oxides (TMOs) such as non-stoichiometric nickel oxide (NiO x ), molybdenum oxide (MoO 3 ), vanadium oxide (V 2 O 5 ), and tungsten oxide (WO 3 ) have become promising candidates for hole injection/extraction interlayers for photovoltaic and organic light emitting devices. 14 Particularly for photovoltaics, these TMOs have successfully been used to replace the commonly used PEDOT:PSS buffer layer due to their better chemical stability, lower acidity and hygroscopicity, as well as their high optical transparency and facile synthesis. 4,6,15,16 NiO x is a p-type semiconductor and exhibits low electron affinity (1.8-2.1 eV), high work function (5.0-5.6 eV) and wide optical band gap (>3.0 eV). In contrast to MoO 3 , WO 3 , and V 2 O 5 , the valence band of NiO x is well aligned for hole transport with the highest occupied molecular orbital (HOMO) levels of many typical p-type conjugated polymers. 17 Recently, NiO x layers, obtained through a sol-gel process from nickel organic salts precursors, have been successfully utilized in organic (OPV) and perovskite photovoltaic devices, showing improved device performance and stability. [18][19][20][21][22][23][24] However, the used sol-gel process required a high temperature treatment (250-400 C) to induce the decomposition of the precursor into a crystalline NiO x nanoparticle (NP) layer followed by a plasma treatment which further oxidizes the nanoparticles. The latter treatment enhances the non-stoichiometry of mixed nickel oxidation states, and, in-turn, the layer's conductivity. 25,26 Such sol-gel systems can be printed using inkjet, leading to injection layers which display performances in various organic and dyesensitized solar cells that are comparable to spin-coated systems. [27][28][29] In spite of these encouraging results, the high processing temperatures that are required to produce wellperforming layers are not compatible with exible plastic substrates and would be a drawback for the fabrication of costefficient devices by high-throughput printing or coating technologies. 30,31 In this work, we demonstrate a low temperature process ($100 C) for NiO x layers and their integration in organic solar cells by inkjet printing, via use of NiO x nanoparticles blended with high molecular weight polyethylene oxide (PEO) as a sacricial processing additive that do not require any high temperature treatment.

Experimental section
The NiO x dispersion (Avantama, Switzerland) was sonicated for 5 minutes before use and ltered with a 0.2 mm PTFE lter. PEO (M w ¼ 1 000 000 g mol À1 , Sigma Aldrich) was dissolved in ethanol at a concentration of 1 mg ml À1 and kept at 50 C in ambient atmosphere for at least 1 h before use. The PEO solution was then added to the NiO x dispersion to obtain blends NiO x :PEO with the content of NiO x ranging from 10 to 99.9% wt. These solutions were spin-coated in ambient condition with u ¼ 2000 rpm, a ¼ 1000 rpm s À1 and t ¼ 40 s just aer mixing. The lms were then annealed on a hot plate at 100 C for 10 min and treated with oxygen plasma for 1 min. For inkjet-printed NiO x or NiO x :PEO layers, the NiO x :PEO compositions were kept the same as in the spin-coating process, though diethylene glycol was added to the ink formulation in a ratio of ethanol : diethylene glycol 1 : 3 vol to adjust the viscosity. The NiO x or NiO x :PEO thin lms were printed using a DiMatix DMP 2831 inkjet printer with a 10 pL cartridge using a custom designed waveform. The jetting voltage was xed at 21 V. The maximum jetting frequency was set to 3 kHz, the print head temperature was set to 40 C, and the substrate was kept at ambient temperature. The printing of a uniform NiO x or NiO x : PEO lm was achieved with an optimized drop spacing of 15 mm. The inkjet-printed layers were dried in a vacuum chamber at 2 Â 10 À3 mbar and subsequently treated as the spin-coated counterparts. For the active layer deposition, samples were moved into a nitrogen-lled glove-box.

Solar cells fabrication and characterization
Glass/ITO substrates were rst cleaned in detergent, water, acetone and 2-propanol under sonication for 15 min, respectively, and treated by oxygen plasma for 5 min. P3HT (M w ¼ 72 800 g mol À1 , RR ¼ 98%) and PC 60 BM (99%, Solenne), were separately dissolved in 1,2-dichlorobenzene at a concentration of 20 mg ml À1 and kept at 50 C in nitrogen atmosphere for at least 18 h. Mixed solutions were prepared at least 2 h prior to fabrication. The P3HT:PCBM ratio was kept at 1 : 0.9. The spincoat parameters of the P3HT:PCBM active layer were u ¼ 800 rpm, a ¼ 500 rpm s À1 and t ¼ 120 s and delivered layer thicknesses of z90 nm. Subsequently, the samples were annealed in the glove-box at 150 C for 10 min. For the cathode contact, 10 nm of Ca followed by 100 nm of Ag were evaporated in a vacuum system with a base pressure of 1 Â 10 À7 mbar. For the J-V curves, a sun simulator (with spectral distribution AM 1.5G) was used as a light source. The intensity of the lamp was calibrated with a silicon reference cell (RERA Systems RR 106 O) to be 1000 W m À2 . The current was measured with a source meter unit (Keithley 2636A). Series and Shunt resistances were calculated from linear tting of the current measured under illumination at open circuit and 0 V voltage bias, respectively.

Thin-lm characterization
Thin lm thicknesses were characterized using a spectroscopic ellipsometer (SENtech Senpro) at a wavelength from 350 to 1050 nm and angles from 50 to 70 . The NiO x , NiO x :PEO and PEO complex refractive index, N ¼ n + ik was obtained using a Cauchy parameterized model with the help of the Sentech soware. UV-vis spectroscopy was measured by illumination with an AvaLight-DH-S-BAL light source and detection with an AvaSpec ULS3648 spectrometer. The transmittance spectra were measured in the range between 300 to 1000 nm on quartz and ITO substrates. Experimental details of the transient absorption measurements and the following analysis are collected in the ESI. † The Kelvin probe characterization was performed using a KP020 single point system by KP Technology. The system was equipped with a gold tip with a diameter of 2 mm, and its energetic resolution is z20 meV. The free surface energies (SFE) were determined via contact angle measurements with a KRÜSS DSA 100 drop shape analyzing system. Deionized water and diiodomethane droplets with a nominal volume of 0.75 mL were placed on the substrate. The contours of the drops and their contact angles, as well as the SFE calculations, were extracted with the help of the KRÜSS soware. The photoelectron spectroscopy characterization was performed using a PHI VersaProbe II scanning XPS microprobe. The spectrometer is equipped with a monochromatized Al-Ka X-ray source, an Omicron HIS 13-helium discharge lamp, and a concentric hemispherical analyzer. Detail spectra of the core level lines were recorded with a pass energy of 11.75 eV, for the secondary electron edges 0.58 eV were chosen. The spectra and secondary electron edges are referenced in binding energy with respect to the Fermi edge and the core level lines of in situ cleaned Ag, metal foil. The transmission electron microscopy (TEM) samples were prepared by drop-casting the diluted (1 : 2 in ethanol) solution of NiO x nanoparticles or the NiO x :PEO blends onto a copper grid, and imaged using a FEI Tecnai T20 system with a LaB 6 electron source, operated at a working voltage of 200 kV. High-resolution top view SEM images were obtained using the Zeiss Ultra Plus high-resolution scanning electron microscope (HRSEM), equipped with a Schottky eld emission electron source. The images were acquired using both secondary electrons and backscattered electrons detectors, at an accelerating voltage of 2 kV and at a working distance of $3.5 mm. Samples were prepared by spin-coating on silicon wafer substrates. The EDS mapping were performed at 4 kV accelerating voltage and at 6.5 mm working distance. The EDS measurements were conducted with Ka and La signals of carbon and nickel, respectively. The samples were fabricated with a thickness of 400 nm for instrumental accuracy. Due to the excessive thickness, oxygen plasma treatment was not performed on these samples.  Table S1 and Fig. S1 of the ESI †). The measured thicknesses are in agreement with the nominal thicknesses ($10 nm) reported in the literature for NiO x HTLs used in OPV devices. 22,32 All HTLs were annealed at 100 C and oxygen plasma treated prior to the deposition of the active layer. This is a necessary step for the formation of the more conductive species NiOOH by further oxidation during the plasma treatment and consequently for the optimized performance of the devices (see Fig. S2 †). 19,26,33 In the same gure we plotted for comparison a device on bare plasma-cleaned ITO (i.e. no HTL), showing a clear improvement in device performance with NiO x as HTL, and a device with PEDOT:PSS, with a PCE of 3.00% that conrms the quality of the P3HT:PCBM active layer. The utilization of P3HT:PCBM as active layer material, although not exhibiting a high performance compared to state-of-the-art organic and perovskites based PVs, is meant to offer an optimum standard for the introduction of the presented HTL processing approach.

Results and discussion
From Fig. 1b and c we can deduce that devices with either neat NiO x or PEO HTLs show lower performances than any of the devices with HTLs composed of NiO x :PEO blends. While the short-circuit current (J sc ) is not affected by the HTL, the power conversion efficiency (PCE), ll factor (FF), and open-circuit voltage (V oc ) show a bell-shaped dependency with PEO content, with the best overall performance at a NiO x : PEO ratio of 40 : 60% wt. Through the oxygen plasma treatment, a work function of z5.0 eV is obtained independent of the PEO content with exception of the neat PEO on ITO (3.8 eV) (see Fig. S3 †). This signicantly lower WF in the neat PEO HTL causes an energy level mismatch that leads to a loss in V oc (0.3 V) with respect to that of the best working blend. Devices with a neat NiO x NP HTL are characterized by a distinct S-shape of the JV curve, with substantial losses in FF in comparison to devices with the optimized blend HTL. These losses are commonly attributed to an imbalance in the charge carrier extraction, 34,35 suggesting an excessive interface resistance induced by a poor HTL morphology. For the NiO x :PEO blends the presence of PEO seems to be necessary for better photovoltaic performance.
To assess the reason for the bell-shaped efficiency dependency with HTL composition, we extracted the shunt (R sh ) and series resistance (R s ) of all devices, Fig. 1d. We identify two contradicting dependence relationships between blend composition and the device performance which show the best compromise at a NiO x : PEO ratio of 40% wt. In the case of NiO x content >40% wt, the device performance is limited by an increasing R s . R s is caused by injection barriers in the device, so we can infer that denser NiO x NP layers are not favorable for efficient hole extraction from the solar cell. In the case of HTL blends with NiO x content <40% wt, R s values are relatively little affected by composition, however, R sh strongly depends on composition with a maximum at 40% wt. We conclude that a further decrease in NP concentration < 40% wt does not allow for the formation of a dense nanoparticle layer, thus providing energetically inferior paths for the photocurrent which leads to a decrease of the R sh and consequently a reduction in the output of the solar cell.
The transient absorption (TA) spectra, aer excitation with 500 nm, of P3HT:PCBM lms deposited on the optimal 40% wt NiO x :PEO HTL are shown in Fig. 2, while the ones deposited on different NiO x : PEO ratios are depicted in Fig. S4a-c. † At early time delays (0.2 ps), the TA spectra are dominated by the ground state bleaching (GSB) of P3HT below 640 nm, and the stimulated emission (SE) of P3HT excitons centred at 700 nm. At longer time delays (aer about 10-20 ps), the SE is replaced by a at positive band due to charge absorption and the GSB has a signicantly reduced amplitude. In agreement with literature, 36-38 these spectral changes are due to P3HT exciton dissociation by charge transfer to PCBM. The unusual decrease of the GSB during charge transfer has previously been attributed to a different degree of delocalization of the P3HT excitons and charges. 36 We have quantied the time constant for the charge transfer by globally analysing the dynamics at all probe wavelengths with an exponential function including an offset. The corresponding amplitude spectra are shown in Fig. 2b and S4d-f, † with the fast component corresponding to the charge transfer and the offset corresponding to long-lived charges. Charge transfer is fastest (2.4 ps) for the 40% wt NiO x :PEO HTL and slower (around 6-7 ps) for the other NiO x : PEO ratios, as  illustrated in the inset of Fig. 2b. This points to a slightly modied morphology of the P3HT:PCBM blend with the different HTLs, leading to different time scales for exciton diffusion through P3HT domains before they reach a PCBM interface for dissociation. Although the TA results point to the smallest P3HT domains for the optimal 40% wt NiO x :PEO HTL, the similar absorption spectra of all samples deposited on NiO x :PEO indicate that the aggregation/crystallinity of P3HT remains comparable (Fig. S4g †). Moreover, the charge transfer in all samples is much faster than the exciton lifetime in neat P3HT lms (about 400 ps), 39,40 so that no signicant difference in exciton recombination loss is expected, in agreement with their similar short-circuit current. There is also no sign of charge recombination within the 1 ns experimental time window in any of the samples, showing the absence of fast geminate charge recombination (which would occur within a few nanoseconds). 37 We conclude that small morphological changes in P3HT:PCBM are induced by the different HTLs, but that they do not lead to signicant differences in the charge generation efficiency that could explain the important changes in device efficiency. X-ray Photoelectron Spectroscopy (XPS) measurements were performed to investigate the surface chemical composition of the NiO x layers before and aer the plasma treatment, as it may strongly inuence the interfacial energy alignment with the BHJ material, thereby the nal photovoltaic performance. Fig. 3 presents the O 1s, C 1s and Ni 2p XPS spectra for NiO x :PEO HTL with various compositions. Before and aer plasma treatment, the spectral distribution of the O 1s and the Ni 2p peaks ascribed to NiO x are not heavily affected, as in previously reported NiO x HTL lms prepared by sol-gel methods. 19,26 This suggests that the ratio between different oxidation states in the NiO x NPs remains unaltered aer plasma treatment. Interestingly, the Ni 2p peak is observed to increase in intensity for the blends while remaining constant for the neat NiO x layer. Simultaneously, the C 1s and the O 1s (534 eV) component of PEO are substantially reduced aer plasma treatment. This can be explained by the PEO being etched awayat least partlyduring the plasma treatment i.e. decapping the surface of the NiO x NP component. This is corroborated by the improvement in the device performances observed aer plasma treatment as shown in Fig. S2. † The XPS characterization and the effect of the plasma treatment on device performances suggest that the main role of the PEO in the blends is to improve the arrangement of the NP in the lm, acting as a sacricial processing additive. Similar device performance improvements have been  reported for OPVs with ZnO(NP):poly-vinylpyrrolidone interlayers through surface etching by UV-ozone treatment. 41 The NiO x :PEO HTLs were further investigated by transmission electron microscopy (TEM) and scanning electron microscopy (SEM) with further energy dispersive X-ray spectroscopy (EDS) analysis to clarify the role of PEO in the lm formation. Fig. 4a-h shows TEM images of layers of NiO x :PEO with 100%, 70%, 40%, and 30% wt NiO x , where the NPs appear as darker objects dispersed in a bright PEO matrix. The images were acquired on layers thinner than the ones used for devices to avoid imaging disturbance. Aer plasma treatment, the suspended part of the lm was mostly lost due to the etching of PEO and the lack of structural support from the substrate (Fig. S5 †). The average interplanar spacing for a NiO x NP, presented in the inset in Fig. 4b, corresponds to the h111i planes of NiO structure (d hkl ¼ 0.24 nm), conrming the crystalline nature of the NiO x NPs. [42][43][44] The neat NiO x layers ( Fig. 4a and b) are dense with several focal planes, indicating the existence of multiple layers of NPs. Blending with PEO causes a gradual dilution and a decrease of the agglomerate sizes of the NPs from an average of $55 nm to $30 nm as observed in Fig. 4c-h, although some large agglomerates are still present in the blend with 30% wt NiO x .
The TEM images show that reducing NiO x content leads to a non-continuity of the HTL (on the nano-scale). At a NiO x NP concentration of 40% wt the coverage is however sufficient to reduce the barrier for charge carrier extraction, effectively screening the effects of the bare areas. Moreover, the nanoscale NP distribution can be directly linked to the resistances in the nal devices presented in Fig. 1d. A thick and dense layer of NP, as in the neat NiO x HTLs, causes an increase in the R s and a decrease in the R sh , while the separation between the nanoparticles that can be notable depending on the PEO content causes a discontinuity in the HTL that is reected in the decreased R sh of the devices.
The surface morphology of the blends on a larger scale was investigated through SEM. Fig. 5 presents the secondary electrons (SE) and back scattered electron (BSE) images of the best working HTL blend (40 : 60% wt NiO x : PEO) pristine and aer plasma treatment. The SE images give information about the surface topography of the lm, while the BSE images give additional information on the chemical composition of the sample surface because the contrast depends on the atomic weight, so the Ni-rich zones are bright, and the C-rich zones are dark.
The images before plasma treatment show bright NiO x particles and large organic dark "stain" regions indicating a phase separation of the blend components when deposited into a thin layer. Aer the oxygen plasma treatment, the dark organic "stains" are no longer observable suggesting removal of the PEO leaving behind a NiO x -rich surface. Fig. S6 † shows the corresponding images for three other blend compositions reecting this behavior as well as the neat NiO x HTL.
Element mapping by EDS on SEM images of pristine HTL are displayed in Fig. 6. It can be observed that the continuity and  the homogeneity of the NiO x distribution are preserved for the blend layers with 70 and 40% wt NiO x , while HTLs with 30% wt NiO x shows a separation of domains comprised of NiO x -rich islands.
Finally, devices with inkjet-printed HTL with the optimal blend composition (NiO x 40% wt) were fabricated, aer considering the surface wettability (see Fig. S7 and Table S2 †) and developing a printable ink formulation. Since the fabrication of layers from ethanol-based solutions is challenging due to the low viscosity and fast evaporation of the solvent (boiling temperature, T b ¼ 78.4 C), 30,31,45 we incorporated the more viscous diethylene glycol (DEG) (T b ¼ 197.3 C) to the asreceived NP suspension.
The post-treatments of these layers included vacuum drying, annealing at 100 C and oxygen plasma treatment. The characteristics of devices fabricated with HTLs from the DEGcontaining solutions are displayed in Fig. 7 and relevant parameters are summarized in Table 1. We fabricated spin-coated and inkjet-printed HTLs from the blend as well as neat PEO and NiO x to determine the inuence of the printing technique on the device performance. In all cases, the inkjet-printed HTLs resulted in the best performing devices with higher V oc and J sc , and consequently PCEs. Furthermore, introducing DEG resulted in an improved device performance compared to devices processed from ethanol-only HTLs shown in Fig. 1. Due to its higher boiling point, DEG slows down the drying process, leading to a different lm formation dynamics. For the inkjet printing process, a longer drying time provides the necessary time for coalescence of the deposited drops. Notably, the results show that the inclusion of PEO is necessary to obtain the well performing of the NiO x HTLs, with an average PCE ¼ 2.37%. This performance is comparable to the ones obtained in literature with NiO x precursors (sol-gel method) that require a thermal treatment >200 C. 29 Our approach, thus, provides an excellent low-temperature alternative that enables inkjet printing of NiO x HTLs for solution-processed photovoltaics application.

Conclusions
In summary, we demonstrated a low-temperature (100 C) processing approach for solution-processed NiO x NP-based hole transport layers and its application for organic photovoltaic devices. PEO was utilized as a processing additive, which improves NiO x lm formation and can subsequently be removed prior to device fabrication. We showed that an optimal ratio of 40 : 60% wt between NiO x NP and PEO permits a good dispersion of the nanoparticles without leading to too dense layers that lead to high resistances necessary for maximizing device efficiency. The NiO x :PEO solutions were processed both by spin coating and formulated for use by inkjet printing exhibiting devices performances comparable to precursorbased sol-gel produced NiO x layers that require posttreatments at high temperature. The utilization of P3HT:PCBM as active layer material, although not exhibiting a high performance compared to state-of-the-art organic and perovskites based PVs, offers an optimum standard for the introduction of the presented HTL processing approach. Therefore, we expect that the low-temperature process reported here by blending NiO x with a sacricial polymer additive can be applicable to a wider range of hybrid systems and interfaces in multilayer optoelectronic devices resulting in device architectures more suitable for the industrial processing of printed devices on plastic substrates.

Conflicts of interest
There are no conicts of interest to declare.