Nickel oxide interlayer ﬁ lms from nickel formate – ethylenediamine precursor: in ﬂ uence of annealing on thin ﬁ lm properties and photovoltaic device performance † , 10949 – 10958 |

An organometallic ink based on the nickel formate – ethylenediamine (Ni(O 2 CH) 2 (en) 2 ) complex forms high performance NiO x thin ﬁ lm hole transport layers (HTL) in organic photovoltaic (OPV) devices. Improved understanding of these HTLs functionality can be gained from temperature-dependent decomposition/ oxidation chemistries during ﬁ lm formation and corresponding chemical structure-function relationships for energetics, charge selectivity, and transport in photovoltaic platforms. Investigations of as-cast ﬁ lms annealed in air (at 150 (cid:2) C – 350 (cid:2) C), with and without subsequent O 2 -plasma treatment, were performed using thermogravimetric analysis, Fourier transform infrared spectroscopy, ultraviolet and X-ray photoelectron spectroscopy, and spectroscopic ellipsometry to elucidate the decomposition and oxidation of the complex to NiO x . Regardless of the anneal temperature, after exposure to O 2 -plasma, these HTLs exhibit work functions greater than the ionization potential of a prototype donor polymer poly( N -9 0 -heptadecanyl-2,7-carbazole- alt -5,5-(4 0 ,7 0 -di-2-thienyl-2 0 ,1 0 ,3 0 -benzothiadiazole) (PCDTBT), thereby meeting a primary requirement of energy level alignment. Thus, bulk-heterojunction (BHJ), OPV solar cells made on this series of NiO x HTLs all exhibit similar open circuit voltages ( V oc ). In contrast, the short circuit currents increase signi ﬁ cantly from 1.7 to 11.2 mA cm (cid:3) 2 upon increasing the anneal temperature from 150 (cid:2) C to 250 (cid:2) C. Concomitantly, increased conductivity and electrical homogeneity of NiO x thin ﬁ lms are observed at the nanoscale using conductive tip-AFM. Similar V oc observed for all the O 2 -plasma treated NiO x interlayers and variations to nanoscale conductivity suggest that the HTLs all form charge selective contacts and that their carrier extraction e ﬃ ciency is determined by the amount of precursor conversion to NiO x . The separation of these two properties: selectivity and conductivity, sheds further light on charge selective interlayer functionality.


Introduction
High efficiency bulk heterojunction (BHJ) organic photovoltaic (OPV) devices oen require contacts modied with hole or electron charge-transport interlayers in order to increase the charge carrier collection efficiency above that of the unmodied transparent conducting oxide or metal contact. 1,2 The efficiency of charge collection interlayers relies upon their thin lm conductivity, 3 work function (F w ), 4-7 alignments of interlayer valence and conduction band edges with the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies across the heterojunction, [8][9][10][11][12] as well as the degree of heterogeneity for the contact surface. 13 The combination of these mechanisms provide for selective charge collection in competition with bimolecular and surface recombination under low internal electric elds (i.e. near open-circuit conditions). [14][15][16] In particular, the importance of selectivity and contact-extraction efficiency becomes increasingly important in solution-processed photovoltaic platforms as free carrier mobilities of photo-active layers increase. Metal oxides formed with Mo, V, W, Ni or NiCo 6 have been shown to exhibit favorable attributes as hole transport layers (HTLs). Photovoltaic applications in organic-, [17][18][19][20][21][22] colloidal quantum structure- 23,24 and methyl ammonium lead halide-based platforms [25][26][27] utilize HTLs with high transmission at operational wavelengths and work function (F w ) values equal to or in excess of the donor IP value, although presumably due to different mechanisms for nand p-type oxides. The grand challenge in contact design remains to effectively create solution-processed deposition methods for high performing thin lm devices while still maintaining material and interfacial functionalities outlined above.
NiO x , is one of few p-type metal oxides that has transversed numerous energy relevant technologies such as catalysis, batteries, fuel cells and photovoltaics. Hence, it is of fundamental interest and several organometallic precursor formulations compatible with solution processing have been identied for thin lm formation. Examples of these are: nickel acetate tetrahydrate complexed with methanolamine (275 C); 28 nickel nitrate hexahydrate with monoethanolamine (500 C) 29 and nickel formate dihydrate with ethylenediamine (250 C). 30 Lowering the processing temperature required to convert these precursors to the oxide allows use of plastic substrates, which in general cannot tolerate prolonged processing above 150 C. 31 There is considerable literature precedent for the decomposition of nickel formate to form Ni and NiO. [32][33][34][35][36][37][38] Diamine complexation with nickel formate lowers the thermal requirement for decomposition and thus enables formation of NiO x at lower temperature. Solutions made with the complexed organometallic precursor in ethylene glycol and water allow fabrication of NiO x thin lms by spin coating the nickel formateethylenediamine-ethylene glycol-water (Ni(O 2 CH) 2 -en-egwater) ink followed by thermal annealing in air. Formation of NiO x by this method is unique as it produces conformal, high performance thin lms with few processing steps.
A detailed understanding of the interconnected decomposition chemistry with the material and interface functionality can drive metal oxide ink development beyond empirical approaches. For example, exposure to reactive oxygen during annealing may further reduce thermal post-treatments. Zhai et al. demonstrated this relaxation in processing conditions for the acetate precursor, below 150 C. 39 As a direct result of the lm growth and processing, NiO x interlayers strongly affect the OPV device performance. 28,30,40 Aer annealing in air and treating with an O 2 -plasma, NiO x outperforms a benchmark HTL of poly(ethylene dioxythiophene):poly(styrene sulfonate) (PEDOT:PSS) in prototypical OPV devices using the BHJ poly[N-9 0 -heptadecanyl-2,7-carbazole-alt-5,5-(4 0 ,7 0 -di-2-thienyl-2 0 ,1 0 ,3 0benzothiadiazole] (PCDTBT): [6,6]-phenyl-C71 butyric acid methyl ester (PC 70 BM). 20 When NiO x interlayers are included in OPV devices, the surface chemistry, band edge energies and mid-gap defect states determine the surface electrical properties and charge selectivity towards holes. Detailed spectroscopic analyses of these solution-deposited NiO x thin lms have shown that these are complex NiO x surfaces, with a wide range of possible oxide stoichiometries that inuence their optoelectronic properties, and their interactions with semiconductors such as those found in organic and hybrid photovoltaic platforms. 8,[25][26][27]41 Previous UPS and XPS measurements on these lms correlated surface hydroxyl species and their dipolar character with an increased band gap energy and improved band edge alignment with BHJ lms. 8,41 More specically, the NiO x surface formed from decomposition of these solution precursors is comprised predominantly of a mixture of NiO x , Ni(OH) 2 and NiOOH, as revealed by XPS characterization. 41 The dipolar character of this modied surface leads to a high F w and favorable energetic matching to the highest occupied molecular orbital (HOMO D ) hole-transport energy level of PCDTBT, while the wide band gap, and an apparent lack of mid-gap states, functions to block reverse electron transfer from the lowest unoccupied molecular orbital (LUMO A ) of the fullerene electron acceptor. 8,20,41 Furthermore, as these processing conditions for the NiO x interlayers led to variations in the measured local density of states observed in UPS, this resulted in higher hole selectivity and lower leakage currents in hole only devices. 41 Through improved charge selectivity and limiting carrier injection from the contact, these NiO x interlayers lower leakage current and increase shunt resistance in OPV devices. 14,42 However, systematic investigation of precursor decomposition in relation to device performance has yet to be addressed and hence, is the focus of this paper.
Here, we study the effects of varying the annealing temperature between 150 C and 350 C for thin lms spin coated from the Ni(O 2 CH) 2 -en-eg-water formulation. The effects of incomplete precursor decomposition are important to understanding their inuence on the interlayer optoelectronic properties and the ability to collect photocurrent in OPV devices. We observe changes to both the chemical and electronic properties of the resulting NiO x thin lms that correlate with large changes in short-circuit photocurrent (J sc ) and little to no changes in opencircuit photovoltage (V oc ) in PCDTBT:PC 70 BM OPV devices. Decomposition/oxidation reactions for the lms were investigated by thermal gravimetric analysis (TGA), differential scanning calorimetry (DSC), Fourier transform infrared absorption spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS). By increasing the anneal temperature for thin lms spincoated from the Ni(O 2 CH) 2 -en-eg-water ink, from 150 C to above 250 C, and subjecting the lms to an O 2 -plasma, amorphous thin lms are formed with: (i) increased conductivity as measured by conductive AFM; (ii) increased surface oxygen content (O/Ni ratio revealed by XPS); (iii) increase of the NiO x band gap; (iv) high performance in OPV devices, as revealed by analysis of their series resistance and J sc . V oc is shown to be affected primarily by the surface oxidation chemistry of NiOOH even if the precursor decomposition/oxidation is incomplete, while losses observed in J sc depend primarily upon the nanoscale conductivity threshold reached upon decomposition of the Ni-formate-diamene complex. These studies decouple the underlying oxide formation from the surface effects by O 2plasma treating for photovoltaic device applications.

Ink preparation
Preparation of the Ni(O 2 CH) 2 -en-eg precursor formulation for NiO x lms has been reported earlier. 30 To summarize, nickel formate (1 g) was combined with ethylene glycol (10 ml) followed by ethylenediamine (0.87 ml). The mixture was heated and shaken multiple times, and then ltered at near ambient temperature through a 0.45 mm pore lter. The ink was a deep purple color, consistent with the violet color reported for Ni(O 2 CH) 2 (en) 2 . 43 For spin coating, the ink was diluted 1 : 1 by volume with water (nanopure).

Thin lms & devices
Patterned ITO substrates were rst solvent cleaned in acetone and isopropyl alcohol followed by an O 2 -plasma treatment. NiO x lms were deposited by spin coating the Ni(O 2 CH) 2 -en-eg-water ink at 4000 rpm onto the ITO substrates and immediately annealing the lms in air at 150-400 C for one hour. Aer annealing, all NiO x layers were exposed to O 2 -plasma treatment for 2 minutes at 155 W and 800 mTorr. The 1 : 4 ratio PCDTBT:PC 70 BM solution was prepared in 1,2-dichlorobenzene under an inert atmosphere at a total concentration of 35 mg ml À1 . The solution was stirred at 90 C for 8 hours before cooling to 60 C followed by immediate use, which is a variation on a previously reported procedure. 20 Spin coated active layers were deposited on top of the NiO x HTL lms at a spin rate of 2000 rpm for 120 seconds. The coated substrates were annealed at 70 C on a hot plate for one hour. Top electrodes composed of Ca/Al (20 nm/100 nm) were thermally evaporated using an Angstrom Engineering thermal evaporator with a base pressure below 1 Â 10 À7 Torr to produce 0.11 cm 2 devices. Films of NiO x were prepared on freshly O 2 -plasma cleaned Au substrates for AFM and C-AFM studies.

AFM & C-AFM
Scanning probe measurements employed an Asylum Research MFP-3D Atomic Force Microscope in conductive mode (c-AFM) using a Pt/Cr coated conductive tip (ElectricMulti75-G by Budget Sensors Inc.) with a radius less than 25 nm. Both height topography and c-AFM were obtained simultaneously. To obtain a good c-AFM signal, NiO x lms were deposited on top of Aucoated glass substrates rather than ITO, since ITO has nonuniform conductive regions 44 and thermally-induced electrical degradation at small length scales. 13 The Au substrate used for calibration had a highly uniform c-AFM prole at very small sample-to-tip bias (VST). All c-AFM measurements on NiO x lms were at ambient conditions with identical scan parameters such as scan speed and drive amplitude, using a VST of 40 mV.

Photoelectron spectroscopy
Experiments were performed on a Kratos Axis Ultra X-ray photoelectron spectrometer equipped with a monochromatic Al K-alpha X-ray source (hv 1486.6 eV) and a He UV source (hv 21.22 eV). Linear calibration of the binding energy scale for the detector was performed following the procedure outlined by M. P. Seah. 45 A bias of À10.00 V was applied to the sample during UPS experiments to spectrally separate the lowest kinetic energy electrons and secondary electrons from the local environment. An Ar sputter-etched, atomically-clean gold sample was measured before characterization of the NiO x samples to establish the Fermi edge of the spectrometer.

TGA & DSC
The ink (nickel formate-ethylenediamine-ethylene glycolwater) was placed in a Pt pan at 120 C to evaporate the bulk of the water and ethylene glycol solvent with minimal disruption of the nickel formate-ethylenediamine complex. This procedure was repeated twice to obtain an initial mass of 13.87 mg of the NiO x lm precursor (less ethylene glycol). The pan temperature was increased at 10 C min À1 under dry synthetic air (20% O 2 , 80% N 2 ) in a TA Instruments SDT Q600 operated in TGA/ DSC mode.

FTIR
Transmission spectra were measured using a Thermo-Nicolet 6700 FTIR. A liquid N 2 -cooled mercury-cadmium-telluride (MCT) detector and a KBr beamsplitter were used. Scans were collected to provide data with a resolution of 2 cm À1 . For each measurement 100 scans were averaged for both the sample and the background. Absorbance spectra were calculated from the sample and background using the Beer-Lambert equation.

Ellipsometry
Thin lm optical properties and thicknesses (9.5 AE 0.5 nm) were characterized using an M-2000 spectroscopic ellipsometer (J.A. Woolam Co Inc.) at wavelengths of 250-1000 nm and angle of 65-75 degrees. Spectroscopic ellipsometry data were processed with the aid of WVASE soware. The NiO x complex refractive index constants (N ¼ n À ik) were obtained using a Lorentz parameterized model, which is consistent with the Kramers-Kronig relations. Thicknesses were veried using a Dektak 8 stylus prolometer.

Decomposition and oxidation of Ni(O 2 CH) 2 (en) 2
Decomposition and oxidation processes for the metal organic precursor depends on the nickel complex formed in solution and spin cast into lms. We follow the conversion to NiO x with TGA and DSC shown in Fig. 1. Previous literature isothermal studies have reported the complete decomposition/oxidation of Ni(O 2 CH) 2 $2H 2 O (at 240 C-280 C in air) 38 and Ni(O 2 CH) 2 (at 215-250 C in oxygen and 242-262 C in air). 35,36 For example, Ni(O 2 CH) 2 $2H 2 O was converted to NiO in less than an hour at 240 C-280 C in air. 38 This is consistent with the expectation that the lms annealed in this study at 250 C or 300 C in air should be comprised of NiO x , which is also supported by the decomposition/oxidation temperature for the Ni(O 2 CH) 2 (en) 2 complex observed by TGA/DSC in Fig. 1. Narain reported the synthesis of Ni(O 2 CH) 2 (en) 2 , which should be robust at 120 C without any loss of the ethylenediamine ligand. 43 Therefore, the DSC/TGA measurements (Fig. 1) were performed in a Pt pan on an ink sample aer two rounds of evaporation of the bulk of the ethylene glycol and water solvents at 120 C, leaving a sufficient mass of primarily the Ni(O 2 CH) 2 (en) 2 complex. On heating in synthetic air (80% N 2 , 20% O 2 ), a small endothermic peak is evident near 125 C, with a corresponding mass loss of ca. 5%, which is interpreted as a loss of residual solvent (water, ethylene glycol, excess ethylenediamine). Between 180-240 C, an additional 77% of the initial mass (81% of mass at 150 C) is lost in an endothermic process, which is comparable to the expected 72% mass change for conversion of Ni(O 2 CH) 2 (en) 2 to NiO. Mass loss is due to a combination of evaporation of residual solvent (likely ethylene glycol, boiling point 197 C), loss of the ethylenediamine ligands, which have been shown to leave stepwise, [46][47][48] 49 and Ni(en) 3 SO 4 (ref. 50) reported temperatures for the nal stage decomposition/oxidation to NiO of 300 C, 325 C, 410 C, and 466 C respectively. On further heating, there is some additional mass lost (ca. 2% of the initial mass) until 375 C, when an exothermic transition is seen and correlates with a gradual mass increase. Hence, the Ni(O 2 CH) 2 (en) 2 complex is to date, preferred for lower temperature formation of solution-deposited NiO x but photovoltaic applications require subsequent surface treatments to increase the work-function. 7,30

Chemical composition characterization
Surface elemental composition determined from XPS measurements conrmed predominant NiO x composition as well as the presence of C and N due to the O 2 -plasma treatment and ambient exposure prior to measurement (see Table 1). O 2 -plasma is known to remove adventitious organic compounds. However, the N content in the lms likely originates from the ethylenediamine, which decreases as the anneal temperature increases from 150 C to 250 C. Yet, aer correcting for the C 1s binding energy (BE) the BE values for N 1s peak centroids are located at 406.6 eV (see S1 †) and are too high to be explained by the presence of unreacted amine groups. A more likely assignment consistent with the O 2plasma treatment is one or more forms of near surface N-O species such as -NO 3 , which can be identied using vibrational spectroscopy.
Exposure of the NiO x thin lms to O 2 -plasma predominantly affects the exposed surface creating similar structures for all the lms while providing less effect on the subsurface material. We analyzed FTIR spectra taken for NiO x thin lms to understand the O 2 -plasma effects for the whole system. FTIR spectra are shown in Fig. 2 for lms spin-coated from the Ni(O 2 CH) 2 -en-eg-water ink, comparing the as-deposited lm ('no anneal') to lms annealed for one hour in air at 150 C, 200 C, 250 C or 300 C. Band assignments for chemical constituents of the lm precursor are listed in Table 2. The major band assignments reported in the literature for the fundamental vibrations of the formate group in nickel formate (dihydrate) are the n 1 n(CH) mode at ca. 2900 cm À1 , the intense n 4 n as (COO) mode at ca. 1570 cm À1 , and asymmetric deformation (n 5 d(C-H)) and symmetric stretch (n 2 n s (COO)) modes between 1400-1350 cm À1 . 34,51-53 For liquid ethylene glycol, the major band assignments reported in the literature are the strong n(OH) stretching mode at 3400-3150  cm À1 , the strong asymmetric (n as (CH)) and symmetric (n s (CH)) stretch modes at 2935 cm À1 and 2875 cm À1 respectively, the strong d(CH 2 ) mode at ca. 1450 cm À1 , and the very strong n(CO) and n(CC) modes at 1100-1050 cm À1 . 54 (NH)) and symmetric (n s (NH)) modes at 3350-3150 cm À1 , the strong d(NH 2 ) mode at 1613 cm À1 , the u(NH 2 ) mode at 1318 cm À1 , and the strong n(CN) mode at 1025 cm À1 . These ethylenediamine ligand band assignments for Ni complexes are consistent with values reported for other transition metal organometallic complexes [56][57][58] and also liquid ethylenediamine. 59 The bands at ca. 1020-1030 cm À1 (n(CN) and n(CO)) and ca. 3200-3350 cm À1 (n(OH)) indicate ethylene glycol and/or ethylenediamine, 54,55,57,59,60 and are discernible only in the no-anneal lm and the lm annealed at 150 C as shown in Fig. 2a. The FTIR spectra indicate that ethylene glycol and ethylenediamine are virtually eliminated by a one hour anneal in air at 200 C. Bands at 1627 cm À1 (d(NH 2 )), and 1338 cm À1 (n 5 d(C-H)) and (n 2 n s (COO)) and 1587 cm À1 (n 4 n as (COO)), also seen in Fig. 2a, indicate ethylenediamine and formate respectively. 34,[51][52][53]57,59,60 The formate bands are present in the FTIR spectra for the as spun lm and the lms annealed at 150 C or 200 C without an O 2 -plasma treatment. A one hour anneal in air at 250 C and 300 C eliminates the formate from the lms resulting in a near featureless spectra consistent with NiO x except for broad bands at ca. 3570 cm À1 that are surface hydroxyls. [61][62][63] This result is consistent with the TGA/DSC data described above.
A comparison of the impact of the O 2 -plasma treatment, typically used for NiO HTLs, is also included in Fig. 2a and b for lms annealed for one hour in air at 150 C, 200 C, 250 C or 300 C. The intensities of all the ethylenediamine, ethylene glycol, and formate bands were lowered signicantly aer treatment with O 2 -plasma, consistent with an O 2 -plasma being highly efficient at removing organic compounds from materials and surfaces. Aer O 2 -plasma treatment, two new bands emerge in the FTIR spectra located at 2190 cm À1 and 2340 cm À1 . The 2190 cm À1 band is present aer O 2 -plasma treatment in the lm annealed at 150 C (i.e., before the ethylenediamine is eliminated), and is very weak in the lm annealed at 200 C. The 2340 cm À1 band is present in the lms annealed at 150 C, 200 C and 250 C aer O 2 -plasma treatment, although the intensity decreases signicantly with increasing temperature consistent with greater conversion decomposition/oxidation of the precursor to NiO x . Given the oxidizing environment in the O 2plasma and the presence of C and N in the partially decomposed/oxidized lms annealed at 150 C and 200 C, the 2190 cm À1 band is tentatively assigned to the n(C-N) modes for oxygen-bonded cyanate (OCN) groups or nitrogen-bonded isocyanate (NCO) groups to Ni 2+ : rather than the stretch modes of C]N in a carbon nitride lm. 64 For example, n(CN) modes have been reported at ca. 2200 cm À1 for the nickel isocyanate complex [Et 4 N] 2 [Ni(NCO) 4 ], 65 CNO À intercalated in a-Ni(OH) 2 (ref. 66) and a theoretical study of the adsorption of cyanate and isocyanate on a Ni(100) surface; 67 and at 2262 cm À1 and 2200 cm À1 for Ni(NCO) 2 $H 2 O, 68 In contrast, a theoretical study of the adsorption of cyanide on a Ni(100) surface reported the n(CN) mode at only ca. 2000 cm À1 , 69 and experimentally the n(CN) mode for Ni(CN) 2 $2H 2 O was reported at 2172 cm À1 . 70 The 2340 cm À1 band is tentatively assigned to the n 3 n as (CO 2 ) mode of CO 2 trapped in the lms annealed at 150 C-200 C aer O 2 -plasma treatment of the partially decomposed/oxidized Ni(O 2 CH) 2 (en) 2 complex. Similar IR bands have been reported for free CO 2 trapped during the thermal decomposition in air of hexahydrated nickel iron citrate to form ultrane NiFe 2 O 4 particles (2320 cm À1 ), 71 propanol/TaCl 5 gel to form Ta 2 O 5 thin lms (at 2345 cm À1 and 2333 cm À1 ), 72 and zinc acetate dihydrate/sodium hydrogen carbonate mixtures in argon to form ZnO nanoparticles (at ca. 2340 cm À1 ). 73 Both of these modes appear to be eliminated aer O 2 -plasma treatment for the lm annealed at 300 C since decomposition/oxidation of the precursor to NiO x is complete. However, as described above a small percentage of N is still observed in the XPS spectra with high BE values for  [74][75][76] This is also the region where medium strength formate and ethylenediamine bands are anticipated. Conrmation of nitrate cannot be provided by the FTIR spectra aer 150 C anneal and O 2 -plasma treatment. However, the peak position of the ethylenediamine u(NH 2 ) mode shis from 1338 cm À1 to 1358 cm À1 for the as-deposited, and 150 C anneal plus O 2 -plasma respectively, and may indicate possible spectral contribution from -NO 3 ions (see S2 †). Aer the 200 C anneal and O 2 -plasma treatment, the 1300-1400 cm À1 region is nearly featureless. FTIR analysis suggests the trapping of CO 2 and the formation of N-based anions in the lms annealed at the lower temperatures and aer an O 2 -plasma treatment. The connement of CO 2 in solid-state NiO x lms implies that a dense surface barrier forms during the O 2 -plasma treatment. The cyanate species assigned in the FTIR spectra were not identied in the more surface-sensitive XPS measurements. Moreover, nitrates observed by XPS could not be unambiguously identied with FTIR. These complementary surface and through-lm measurements lead to the tentative hypothesis that the low concentration of nitrates are most likely surface conned.

Characterization using spectroscopic ellipsometry
In order to study the effect of annealing on the optical properties we employed angular dependent UV-vis reection using spectroscopic ellipsometry. Fig. 3 displays the square of absorbance versus incident photon energy. Absorbance was calculated using the extinction coefficient from Lorentz oscillator model ts to the ellipsometric data. The onset of absorption was extrapolated aer a linear t to the square of the absorption coefficient between 4.1 and 4.4 eV. 77 For the lowest temperature anneal at 150 C the NiO x absorbance is low and barely resembles the absorption edge of a semiconductor, which is consistent with the minimal decomposition/oxidation of the Ni(O 2 CH) 2 (en) 2 complex at this temperature. Nevertheless, tting the onset region results in an estimate for the optical gap of 3.4 eV. For the lm annealed at 200 C, the absorbance increases as the precursor has partially decomposed/oxidized resulting in a 3.8 eV estimate of the optical band gap. For lms annealed at 250 C and 300 C, because the decomposition of the Ni(O 2 CH) 2 (en) 2 complex to NiO x is well advanced the absorption edge is more clearly dened and results in an optical gap estimate of 3.9 eV, which is in good agreement with the accepted band gap of 4.0 eV for NiO. 78,79

OPV device performance
To investigate the impact of the NiO x lm composition as a function of annealing temperature, on OPV performance we utilized these lms as HTLs, annealed at different temperatures on ITO and integrated them into ITO/NiO x /PCDTBT:PC 70 BM/ Ca/Al OPV devices. Current-voltage (JV) measurements were performed under one-sun illumination. The data from these devices are shown in Fig. 4a with calculated performance metrics in Tables 3 and 4.
Power conversion efficiencies (PCE) are shown normalized in Fig. 4b and increased from 0.5% to 5.7% with increasing anneal temperature between 150 C and 250 C. These PCE values trend directly with short-circuit current density and as a function of thermal annealing temperature. Likewise, the PCE and J sc inversely trend with R s as a function of annealing temperature. For the lowest annealing temperatures (150 C and 200 C) the devices suffer from large resistive losses, poor current extraction and low ll factors. At 250 C and above the R s drops substantially and the device performance improves with gains in J sc and FF. This drop in series resistance within the device is commensurate with the decomposition/oxidation of the NiO x layer.
It is important to note that the open-circuit voltages do not appear to trend with annealing temperature. Ultraviolet photoelectron spectroscopy (UPS) measurements (see S3 †) for these solution-deposited NiO x lms aer annealing between 150 C and 300 C all produce lms with very similar work function values that range from 5.4 to 5.5 eV and IP values of 5.7-5.8 AE 0.1 eV, in agreement with earlier reports. 15 This is consistent with the relatively uniform V oc found across the devices when one considers work-function and the interface electronic structure of the contact determining factors of V oc . For the devices annealed at 250 C and above there is very little statistically signicant difference in the device data. As shown in Table 3, there is a modest increase just above the statistical noise from 250 C to 300 C. It is clear that the compositional changes from annealing the NiO x HTL to 250 C signicantly alter the electronic properties and result in enhanced holecollection from the BHJ.

Conductive tip-AFM
To examine the charge transport across the NiO x lms processed at different temperatures and to better understand holecollecting efficiency, samples were prepared (in an identical way to the lms used in devices) on Au as opposed to ITO substrates. Gold was used to eliminate underlying effects of electrical heterogeneity of the ITO, compared to the uniform and stable background provided on the gold surface. The nanoscale topography and electrical transport of these lms were measured with conductive tip-AFM using a Pt/Ir tip held at the same tip bias for all lms. For these experiments the tip forms the top contact on the Au-NiO x -Pt/Ir tip junction. Fig. 5 shows AFM and c-AFM data for ve different anneal temperatures between 150 C and 350 C. The white pixels in the bottom row of Fig. 5 indicate a good electrical transport or an Ohmic junction, while dark pixels indicate poor electrical conductivity or diode like behavior. Since all the NiO x lms have similar work functions as seen in Table 1, we can assume minimal changes in tip-surface injection/extraction barriers. Hence, we conclude that the measured c-AFM data is indicative of changes in the conductivity of the NiO x . Conformal, thin-lm coatings were observed for each annealing condition on the Au substrates. Moreover, XRD analysis of a series of NiO x thin-lms heated from 150 C to 350 C showed no sign of diffraction peaks. The 150 C anneal resulted in an undulating surface topology (1.46 nm RMS) and a highly insulating lm with no current transfer between tip and the Au ground. In contrast, the lm annealed at 200 C has a much atter surface morphology (0.46 nm RMS) and a small but measurable current, which indicates improved charge transfer with <10% area being electrically active. As the annealing temperature is increased from 200 C to 350 C, increases in the NiO x lm roughness and electrical property are observed. Aer the 250 C anneal, the lm roughness is 0.65 nm RMS and conductive regions occupy most of the surface. The 300 C anneal results in a NiO x lm with larger protrusions (1.35 nm RMS) and a sharp increase in the measured current that saturated the 20 nA c-AFM detection limit. The NiO x lms resulting from a 350 C anneal exhibits slightly lower conductivity compared to the 300 C lm, and a further increase in particle/grain size. The improved through-lm conductivity, as assessed by the area of the saturated pixels, correlate well with lowering of the series resistance and improved current collection for the OPV devices. This suggests that high series resistance for the lower anneal temperatures is a result of the poor conductivity of the HTL, which is a direct result of incomplete precursor decomposition/oxidation.
The large volume fraction of incompletely decomposed precursor results in resistive properties at the nanoscale and macroscale. However, separating improvements to carrier concentration and mobility remains elusive. One would expect    28 These studies show increased series resistance as a direct result of incomplete decomposition of the NiO x precursor complex and how these resistive HTL lms lead to lower J sc in devices with concomitant performance loss. However, for interlayers where conversion to NiO x is performed below the decomposition temperature and is incomplete, V oc remains at high values. The surface chemistry and work function of the NiO x interlayer determines the V oc in OPV devices whether or not the organometallic precursor has fully decomposed to form NiO x . In comparison, non-selective selfassembled molecular interlayer contacts provide paths to electron transfer from the BHJ LUMO levels and reduce quasi-E F splitting, which lead to lower V oc . 7 From the results presented here, we conclude that the nanoscale electrical changes as a function of converted precursor observed seem to strongly affect the ability of charge selective NiO x interlayers to extract holes from the adjacent BHJ and transport those to the external circuit. Recently, the surface polarity of NiO x interlayers was investigated and shown to dominate the interface properties when compared to the interlayer surface roughness and crystal structure. 40 Hence, post treatment and formation of a dipolar surface with low defects is related to the increased polar component of the total surface energy. Data presented here shows that surface composition for these lms are similar. However, differences in their nanoscale conductivity do not strongly affect the V oc . Hence, surface recombination velocity is not signicantly enhanced as quasi-E F splitting seems nominally equivalent for these devices at V oc , which is consistent with steady-state and transient photocurrent studies on similar systems. 42 We hypothesize that the majority of J sc loss observed as the processing temperature is lowered below the precursor decomposition threshold proceeds via recombination in the BHJ and is not mediated by NiO x surface states. If this postulate holds, then charge selectivity and efficient carrier transport are functionally separate and proceed by different mechanisms for this particular active layer. Moreover, these properties are also spatially separate as the selectivity is determined by the surface composition and local density of states that provide a low defect interface and low surface recombination while the interlayer subsurface enables charge delocalization and carrier transport to the transparent electrode. Implications for the separation of selectivity and transport mechanisms could result in designs for bilayer selective contacts and indeed examples exist in literature. 80 This can also help to decouple surface and subsurface effects of decomposition temperature, organometallic precursor formulations and subsequent surface modications for efficient interlayer contacts in photovoltaic technologies. However in more demanding photovoltaic systems with higher carrier mobilities and photogenerated charge densities, it may be necessary to increase the NiO x thickness in order to effectively passivate high carrier density electrodes such as TCOs and metals.