Chemical vapour deposition (CVD) of nickel oxide using the novel nickel dialkylaminoalkoxide precursor [Ni(dmamp′)2] (dmamp′ = 2-dimethylamino-2-methyl-1-propanolate)

Nickel oxide (NiO) has good optical transparency and wide band-gap, and due to the particular alignment of valence and conduction band energies with typical current collector materials has been used in solar cells as an efficient hole transport-electron blocking layer, where it is most commonly deposited via sol–gel or directly deposited as nanoparticles. An attractive alternative approach is via vapour deposition. This paper describes the chemical vapour deposition of p-type nickel oxide (NiO) thin films using the new nickel CVD precursor [Ni(dmamp′)2], which unlike previous examples in literature is synthesised using the readily commercially available dialkylaminoalkoxide ligand dmamp′ (2-dimethylamino-2-methyl-1-propanolate). The use of vapour deposited NiO as a blocking layer in a solar-cell device is presented, including benchmarking of performance and potential routes to improving performance to viable levels.


Introduction
Among p-type semiconductors, nickel oxide (NiO) has good optical transparency and wide band-gap, which result in low visible light absorption losses, and due to the particular alignment of valence and conduction band energies with typical current collector materials it has been widely used in solar cells as an efficient hole transport-electron blocking layer. 1 NiO in solar cells is most commonly deposited via sol-gel and annealed 2 or directly deposited as nanoparticles without annealing, 3 although an attractive alternative approach is via vapour deposition.

Precursor synthesis
The novel precursor [Ni(dmamp 0 ) 2 ] was synthesised following adaptations from two different literature reports in relatively high yields ($86%); 4,13 recrystallisation in THF produced orange crystals of the orthorhombic space group with square planar geometry at the nickel centre, as conrmed by single crystal Xray crystallography. The molecular structure of [Ni(dmamp 0 ) 2 ] (Fig. 1) shows a nickel centre bound to 2 oxygen atoms and 2 nitrogen atoms. The O(1)-Ni(1)-O(1) 1 and N(1)-Ni(1)-N(1) 1 bond angles were both 180 which conrms the square planar geometry of the nickel centre. Both of the Ni-O and Ni-N bond lengths were 1.84 A and 1.95 A, respectively which are comparable with data reported in the literature. 14 4 Similarly, the C-N and C-C bond lengths within the dmamp 0 ligands are also comparable to those outlined in these papers. The structure was consistent with the 1 H and 13 C NMR spectra.
The [Ni(dmamp 0 ) 2 ] synthesised in this work is different to similar precursors reported in the literature 13 because the two methyl groups on the dmamp 0 ligand backbone are positioned on a different carbon atom (Fig. 2).
Thermal analysis of the novel precursor [Ni(dmamp 0 ) 2 ] suggests a melting point of 180 C (differential scanning calorimetry (DSC) curve, blue line) although thermogravimetric analysis (TGA) indicates mass loss is observed from signicantly below that temperature. Subsequently, decomposition occurs in the range of $200-250 C, with an overall mass loss of $75% (Fig. S1 †). The data correlate well with that reported in the literature for similar dmamp-type structures. 15 2 ]-type structures, it is believed that the novel precursor has a higher thermal stability as its end decomposition temperature is greater (240 C compared to 174 C in literature), 15 which suggested it may have preferable properties for ALD deposition. Extensive attempts at producing NiO using a true ALD process, either using precursors previously described in literature for ALD or the novel [Ni(dmamp 0 ) 2 ] precursor synthesised here, were however uniformly unsuccessful and therefore we carried out CVD in order to target thin lm growth of NiO.
TGA/DSC showed that precursor vaporisation occurred from as little as 80 C, and initial experiments indicated at this temperature the vapour pressure was sufficient for vapour transport of the precursor into the reactor without causing precursor decomposition. 1 H NMR analysis of the precursor aer being heated at 80 C for long periods of time (several weeks) showed no signs of decomposition. In order to determine the optimum growth temperature for the NiO lms, CVD experiments were initially performed for 24 hours at growth temperatures of 250-400 C (the long growth times were due to the relatively low vapour pressure of the precursor at 80 C). For lms deposited at 250-350 C the refractive index values for the deposited lms were in line with those reported in the literature  for NiO, which range from 2.3 to 2.9 depending on the deposition temperature, 17 and increased with lm thickness (see ESI †). However for lms deposited at 400 C the value of n decreased. Similar behaviour has been reported by Lu et al., where a decrease of 0.2 in the refractive index of NiO lms was observed in the photon energy region from 1.2 to 3.3 eV when samples were annealed at temperatures above 500 C. 17 The AFM images in Fig. 3 show that as the growth temperature increased, the surface roughness of the NiO lms also increases. Some larger particulates were observed on the lm surface as the deposition temperature exceeded 300 C. Fig. 4 shows the X-ray diffraction (XRD) measurements for lms deposited at 300 C for various lengths of time (see ESI † for details). The XRD patterns show that a 6 hours-deposited CVD lm ($5 nm) is weakly diffracting, most likely because the lms were too thin to produce signicant diffraction, which is consistent with the appearance of a broad background peak attributed to breakthrough to the glass substrate around 20 2q. For lms deposited for 18 hours or longer (>30 nm), NiO peaks became visible, with the (200) reection the most intense (NiO PDF reference number 01-089-5881) indicating some degree of preferred orientation. As the lm thickness increased (i.e. longer deposition time) the intensity of the NiO peaks increased.
To determine the elemental composition and electronic state of the elements within the NiO lms, X-ray photoelectron spectroscopy (XPS) was performed. Fig. 5 shows a high resolution surface scan of the Ni 2p 3/2 peak. The principal core peak of Ni 2+ (red peak) was observed at a binding energy of 855.4 AE 0.2 eV, with satellite peaks 6.1 eV (green peak) and 9.0 eV (purple peak) above the principal peak, which can be attributed to multi-electron excitation. 18 The prominent satellite shoulder 1.8 eV above the Ni 2p 3/2 principal peak (blue peak) is unique to NiO. It has the same shape and FWHM (3.2 eV) as the principal peak. A typical survey XPS spectrum, full Ni 2p peak envelope and O1s spectrum are provided in ESI. † The AFM images in Fig. 6 show that as the deposition time (and hence lm thickness) increased, the surface roughness of the NiO lms also increases. Some larger particulates were observed on the lm surface as the lm thickness approached 70 nm. This resulted in a sharp increase in the surface roughness of the lms from 10 nm to 20 nm RMS (root mean square) roughness for lms of 30 nm and 50 nm thickness respectively.

Integration in solar cell device
NiO has been used as a bottom hole transport layer (HTL) in inverted perovskite solar cells (p-i-n), as well as a top HTL in conventional perovskite solar cells (n-i-p). 19 In order to perform functional testing of the CVD prepared thin lms, inverted (p-in) perovskite solar cells were fabricated. For comparison, NiO   control lms were prepared by a typical sol-gel method from nickel acetate, analogous to previous reports. 20 NiO control devices achieved a maximum power conversion efficiency (PCE) of 14.1% (Fig. 7), however the NiO CVD devices had a signicantly lower performance (PCE 3.9%).
While NiO is clearly a versatile metal oxide HTL, its surface roughness is of particular importance for p-i-n perovskite solar cells since this has a large inuence on the growth of the perovskite crystals (absorber layer) and to achieve smooth, uniform and dense perovskite layers, it is important that the underlying NiO lm has low surface roughness (RMS < 5 nm). 21 In order to determine whether the signicant decrease in PV performance could be due to a change in surface roughness from the deposition of NiO on FTO, we performed additional AFM measurements of FTO, CVD-deposited NiO/FTO and spincoated NiO/FTO (ESI †). A CVD-deposited NiO lm on an FTOcoated glass substrate had an RMS roughness of 11.6 AE 0.5 nm, very similar to that of the underlying substrate (11.4 AE 0.2 nm) demonstrating good conformality, however a spincoated NiO lm on an FTO-coated glass substrate had an RMS roughness of 9.7 AE 0.1 nm. While the difference was only a few nanometers, the rougher surface of the CVD-deposited NiO lms is thought likely responsible for the drop in performance. 21 Such limitations in the lms also resulted in a poor ll factor (FF), which can be attributed to increased recombination from an increased series resistance in comparison to the control samples. Further work in optimising smooth deposition are expected to help eliminate the factors limiting performance.

Precursor synthesis
Unless otherwise stated, the majority of manipulations were performed under nitrogen using standard Schlenk techniques and a Unilab MBraun glove box. Oxygen-free nitrogen (99.9% purity) was obtained from BOC and used as supplied. All solvents were puried by standard methods with respect to oxygen and water and then stored over activated 3 A molecular sieves until used. Anhydrous nickel(II) chloride hexahydrate (98+%) was obtained from Acros Organics and used as supplied. Nickel(II) chloride anhydrous (98%) was obtained from VWR and used as supplied. Dmamp 0 (2-dimethylamino-2-methyl-1propanol) was obtained from MP Biomedicals and puried by standard methods with respect to oxygen and water using the freeze-pump-thaw technique and then stored over 3 A molecular sieves until used. Methanol (99.8%), ethanol (95%) and petroleum ether (40-60 C) were purchased from Sigma Aldrich and used as supplied. Tetrahydrofuran (THF) and hexane were obtained from a dry solvent system within the chemistry department at University College London (UCL).

Precursor analysis
1 H NMR and 13 C{ 1 H} NMR spectra were recorded using Bruker AMX 300 MHz and Bruker Avance III 600 MHz spectrometers and were referenced to the residual proton and 13 C resonances of the solvent. Microanalytical data and mass spectrometry data were obtained at UCL. Mass spectra were recorded using Thermo MAT900 and Micromass LCT Premier Spectrometers. Thermogravimetric analysis (TGA) was performed using a Netzsch simultaneous TG-DTA/DSC apparatus equipped with Proteus soware at atmospheric pressure, using aluminum pans under a constant ow of helium gas. The heating rate was 10 K min À1 . X-ray crystallography diffraction data were recorded on an Agilent Super Nova Dual Diffractometer with Cu Ka radiation (l ¼ 1.5418 A) at 150 K. Using Olex2, 22

Synthesis [Ni(dmamp 0 ) 2 ]
Sodium hydride (1.745 g, 72.73 mmol) was dissolved in $100 mL THF and stirred in an ice bath for 2 hours which formed a milky grey slurry. Dmamp 0 (2-dimethylamino-2methyl-1-propanol) (10.63 mL, 86.17 mmol) was added and the solution was reuxed overnight resulting in a clear orange/ brown solution (Fig. 8). The solvent was removed under vacuum to isolate the intermediate salt as a pale yellow solid. The sodium salt was added to a slurry of NiCl 2 (5.30 g, 40.89 mmol) in $50 mL THF at 0 C and then warmed to room temperature. The solution was then reuxed for 2 days, forming a dark purple/black solution (Fig. 8) Thin lm synthesis CVD of nickel oxide thin lms were performed using a ow-type, cold-walled reactor described previously. 25 The reactor was programmed using a custom IGI Systems Lab Interface Input control box which automatically controls all temperatures and heating systems, gas ow rates and solenoid valves. CVD experiments were carried out using [Ni(dmamp 0 ) 2 ]. Pureshield argon gas (99.998%) supplied by BOC, was used as the carrier gas for all depositions. Gas ow rates were controlled using Mass Flow Controllers (MFC's) purchased from Brooks Instrument (GF40 model number), with ow rates varying from 20-700 sccm. The reactor running pressures therefore varied in the range of $1.0-7.0 mbar. Quartz glass slides (obtained from Wuxi Crystal and Optical Instrument Company Limited) were used as the substrate materials. Prior to deposition, substrates were cut into $4.0 Â 2.5 cm pieces, cleaned using iso-propanol (Sigma Aldrich, 99.5%) and air dried before loading into the reactor. Whilst the substrate holder was being heated to the required temperature, the reactor was pumped down under vacuum to achieve a base pressure of $4 Â 10 À2 mbar. Gas ows were then turned on, where the running pressure was recorded. [Ni(dmamp 0 ) 2 ] was introduced into the reaction chamber by passing the carrier gas into the bubbler to assist the transportation of vaporised precursor molecules. To prevent the precursor condensing or reacting in the pipework, the bubbler outlet line was held at a temperature higher than the bubbler temperature but lower than the substrate temperature. Films were deposited by continuously dosing the metal precursor into the reactor under a constant ow of inert gas until the desired reaction time/lm thickness had been reached.

Thin lm analysis
Ellipsometry measurements were performed using a Semilab SE-2000 Ellipsometer. A continuous spectrum of light was generated by a broadband 75 W arc lamp including ultraviolet through visible to near infrared (1.25-5 eV). Microspot optics were used conne the beam spot size to 3.5 mm in the minor axis. Data was recorded at angles of incidence of 60, 65, 70 and 75 . Ellipsometric optical models were constructed within the Spectroscopic Ellipsometry Analysis (SEA) soware. The NiO model consisted of two components, a Tauc-Lorentz and a Bruggeman dispersion function. An air-NiO diffusion layer (effective medium approximation) was also incorporated in order to account for surface roughness of the lms. Fittings were performed within the spectral range 250-990 nm and the obtained refractive indices (n) of the deposited lms were recorded for comparison at a wavelength of 632.8 nm (1.96 eV). Atomic Force Microscopy (AFM) measurements were obtained using a Nanosurf Easy Scan Atomic Force Microscope, with a 10 mm head in non-contact tapping mode. Scan areas were 5 Â 5 mm, with measurements recorded at 250 points per line (20 nm lateral resolution) with 1 s per line scan time. Data extracted from each scan included the arithmetic average roughness (R a ) and the root mean square roughness (R q or RMS). X-Ray Diffraction (XRD) measurements were performed using a Bruker-Axs D8 (GaDDS) diffractometer which operates with a Cu X-ray source, monochromated (Ka 1 and Ka 2 ) and a 2D area X-ray detector with a resolution of 0.01 . The diffraction patterns obtained were compared with database standards from the Inorganic Crystal Structure Database (ICSD), Karlsruhe, Germany. The beam spot is approximately 5 mm 2 , which means that several areas of the sample can be analysed separately. For all thin lms analysed an incident angle of 0.5-1 was used, and the diffracted X-rays were detected at angles from 10-66 . X-Ray Photoelectron Spectroscopy (XPS) analysis was performed using a Thermo Scientic K-Alpha X-ray photoelectron spectrometer with monochromated Al K alpha radiation, a dual beam charge compensation system and constant pass energy of 50 eV. Survey scans were collected in the range 0-1200 eV. XPS data was tted using CasaXPS soware. The principal peaks of interest were Ti 2p, Ni 2p, O 1s, Si 2p and C 1s. The escape depth in this system was in the range of 1-10 nm. Depth proling was carried out via argon ion sputtering.

Device fabrication
Fluorine-doped tin oxide coated glass (FTO) was rst etched using 2 M hydrochloric acid and zinc followed by sequential cleaning in detergent, deionized water, acetone, ethanol, isopropanol, and nally treated with oxygen plasma for 10 minutes. A compact layer of NiO was then deposited onto the FTO following a previous report. 20 Briey, nickel acetate tetrahydrate (Sigma-Aldrich) was dissolved into 2-methoxyethanol (Sigma-Aldrich, anhydrous), and ethanolamine (Sigma-Aldrich) was used as a stabiliser. The molar ratio between nickel acetate tetrahydrate and ethanolamine was kept at 1 : 1, and the resultant concentration was 0.2 M. The solution was stirred overnight and ltered through a 0.2 mm PTFE lter before deposition. In order to obtain a 20 nm NiO lm, the sol-gel was spin coated onto FTO at 4000 rpm and annealed at 250 C for 20 minutes to achieve a transparent dense lm. For the CVD lms, NiO lms of the same thickness (20 nm) were used as a comparison.
The FTO/NiO substrates were then once again treated with oxygen plasma for 10 minutes before being transferred to a nitrogen lled glovebox. The light absorbing layer, methylammonium lead iodide (MAPbI 3 ), was prepared with slight modication to a procedure previously reported by Ahn et al. 26 The CH 3 NH 3 I$PbI 2 $DMSO adduct solution was prepared by mixing 461 mg of PbI 2 with 159 mg of CH 3 NH 3 I with 600 mg of DMF and 78 mg of DMSO. The solution was stirred for 1 hour at 70 C and ltered before use. The ltered solution was spin coated on the previously prepared NiO lms at 4000 rpm for 30 seconds and aer 23 seconds had elapsed, 0.5 mL of diethyl ether was slowly dripped onto the rotating substrate. The transparent CH 3 NH 3 I$PbI 2 $DMSO adduct lm was heated to 65 C for 2 minutes and 100 C for 10 minutes to obtain a dense CH 3 NH 3 PbI 3 lm. For the electron transport layer, a 23 mg mL À1 solution of PCBM (Ossila, 99.5% purity) in chlorobenzene was spin coated on top of the perovskite lm at 1500 rpm for 20 seconds. A thin layer of 0.5 mg mL À1 BCP (Lumtec, Inc) in methanol was then deposited on top of the PCBM by spin coating at 4000 rpm for 20 seconds. Finally, 100 nm of Ag was thermally evaporated as the counter electrode. J-V measurements were performed under one sun (AM 1.5G) illumination using a calibrated solar simulator with a xenon lamp (LOT). The light intensity was calibrated by silicon reference cell certicated by National Renewable Energy Laboratory (NREL).

Conclusions
NiO thin lms have been deposited by CVD of the novel nickel dialkylaminoalkoxide precursor [Ni(dmamp) 2 ] at substrate temperatures in the range of 250-400 C, where lms with the highest conformality and uniformity tended to be those deposited at 300 C.
AFM and XRD analysis suggest that the lm density and crystallinity increased with lm thickness, which was supported by an increase in the refractive indices of the lms deposited at 250-350 C. However at higher growth temperatures (400 C) lms appeared non-uniform and the refractive index decreased, which suggested that the lm crystallinity had decreased due to a possible change in lm composition and less uniformity in thickness.
Together with the XRD patterns and the AFM images, it can be suggested that lms deposited at low growth temperatures/ thicknesses consisted of mixtures of amorphous and NiO phases; where growth occurred via nucleation sites which grow together and coalesce to form a dense lm.
XPS analysis conrmed the presence of Ni 2+ on the lm surfaces with 2p 3/2 and 2p 1/2 peak binding energies consistent with those reported in the literature. A prominent satellite shoulder 1.8 eV above the principal 2p 3/2 peak was observed which is unique to NiO.
In addition to synthesis, we also show that our approach can also be used to integrate NiO in photoelectrodes. While the performance was limited likely due to surface texture (roughness), the un-optimised devices were able to achieve a power conversion efficiency of 3.9%, highlighting their potential application in energy conversion.

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