Controlled Crystallinity of TiO 2 Layers Grown by Atmospheric Pressure Spatial Atomic Layer Deposition and their Impact on Perovskite Solar Cell Efficiency

Atmospheric Pressure Spatial Atomic Layer Deposition (AP-SALD) is an upcoming deposition technique suitable for a variety of materials and combines the bene ﬁ ts of a regular atomic layer deposition with a signi ﬁ cantly increased deposition rate at ambient conditions. In this work, amorphous and anatase TiO 2 layers are fabricated by AP-SALD via systematic variation of process conditions such as temperature, reactant (H 2 O and O 3 ), and posttreatment. The formed layers are characterized for their structural and optoelectronic properties and utilized as a hole-blocking layer in hybrid perovskite solar cells. It is found that TiO 2 layers fabricated at elevated deposition temperatures possess strong anatase character but expose an unfavorable interface to the perovskite layer, promoting recombination and lowering the shunt resistance and open circuit voltage of the solar cells. Therefore, the interface is essential for e ﬃ cient charge extraction, which can be signi ﬁ cantly improved by controlling the process parameters.

Thereby, ALD is of particular interest, as it allows the reproducible fabrication of pin-hole free layers of high quality on flat films and conformal coatings on nanostructures. Additionally, the resulting layer properties, like thickness and crystallinity, are precisely adjustable by the processing conditions such as temperature, pressure, purge, and exposure time as well as by the choice of precursor and reactant. Typically, TiCl 4 , titanium tetraisopropoxide (TTIP), and tetrakis dimethylamino titanium are used as precursors and H 2 O, H 2 O 2 , and O 3 as reactant. Spatial atomic layer deposition at atmospheric pressure (AP-SALD) is the next evolutionary step of ALD and extends the precise thickness control of ALD by a fast vacuum free and upscalable deposition in ambient environment [27,28].
Here, we demonstrate the controlled fabrication of amorphous and anatase TiO 2 by AP-SALD and relate the influence of processing parameters to structural, optical, and electronic properties and their implication on the performance of perovskite solar cells.

TiO 2 Layer
Preparation. TiO 2 layers were fabricated on a homemade AP-SALD system with one precursor and one reactant line, and three curtain lines, each with an outlet area of 3.5 mm 2 . TTIP (>97%, Sigma Aldrich) was used as metal source precursor and H 2 O or O 3 as reactant. In order to avoid condensation, the titanium precursor, transport tubes, and deposition head were preheated to 85°C. Bubble flow rate for N 2 was set to 150 Nml/min for TTIP and 100 Nml/ min for H 2 O. Flow rate of the N 2 transport gas stream was set to 300 Nml/min in both cases, whereas the flow rate of the N 2 curtain line was set to 600 Nml/min. O 3 flow rate, with oxygen as transport gas, was manually adjusted in line with the H 2 O flow rate. Substrates were held on a heated stage by vacuum and passed the deposition head twice for one cycle. The distance between deposition head and substrate was adjusted to about 250 μm. For layer fabrication with an incorporated temperature gradient, the first layer was deposited at the initial temperature which then was smoothly increased/decreased until the final temperature was reached.
Layers for Raman spectroscopy and profilometer measurements were fabricated on microscope glass slides. Layers for atomic force microscopy (AFM), scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy (EDX), spectroscopic ellipsometry (SE), X-ray and ultraviolet photoelectron spectroscopy (XPS and UPS), and Kelvin probe (KP) were fabricated on polished silicon (111, n-type, Active Business Company GmbH) wafers, and for ultraviolet-    3 International Journal of Photoenergy Layers for XPS and UPS have been used as fabricated without cleaning or sputtering to exclude change in atomic composition and bindings.
As reference, a~70 nm thick TiO 2 layer was deposited by spin-casting a commercial sol-gel paste (Ti-Nanoxide BL/ SC, Solaronix) and sintered at 450°C for 30 minutes.

Perovskite Solar Cell Preparation.
The perovskite solar cells (MAPbI, CH 3 NH 3 PbI 3 ) were fabricated via a two-step deposition method. All chemicals were used as purchased without further purification. ITO coated glass substrates were ultrasonically cleaned for 5 min in detergent, acetone, and isopropanol consecutively, followed by N 2 drying. Subsequently, a TiO 2 blocking layer was deposited on the substrates either by AP-SALD or via the sol-gel process as reference. In case of modification by [6,6]-phenyl-C61butyric acid (PCBA, Solenne BV), the substrates were emerged into a PCBA bath (0.2 mM in chloroform) for 1 hour. They were then transferred to a glove-box where moisture and oxygen levels were kept below 5 ppm. PbI 2 solution (1 M in anhydrous dimethylformamide) and substrates were preheated and maintained at 60°C before spin-coating. Hot PbI 2 solution was spin-cast on the substrates at 5000 rpm for 15 s, which were then quickly transferred onto a hotplate at 70°C for drying. Subsequently, substrates were dropcoated with 6.54 w/v % methylammonium iodide (Solaronix) in anhydrous isopropanol for 20 s, spun at 4000 rpm for 20 s, and annealed at 100°C for 10 min. Spiro-OMeTAD (80 mg/ml, Merck) was spin-cast at 4000 rpm for 120 s, and WO 3 (3 nm)/Ag (100 nm) were finally deposited on top by thermal evaporation.

Results
To aim for high quality TiO 2 thin films and their desired application in solar cells, we comprehensively investigated the role of varying deposition parameters on layer formation, crystallinity, and interfacial effects. Therefore, we systematically varied the deposition temperature and the reactant (H 2 O and O 3 ) during the AP-SALD process and analyzed the resulting structural and optoelectronic properties, as well as basic solar cell characteristics.

Structural Analysis.
The layers for structural analysis were fabricated at sample temperatures between 100 and 400°C with a fixed number of cycles underneath the deposition head. The resulting thickness was determined by profilometer measurements and spectroscopic ellipsometry, yielding a growth per cycle (GPC) rate of about 0.7-1 Å/cycle for H 2 O and slightly lower for O 3 as reactant (see Table 1 and Figure S1 in SI).
XPS measurements of thin (75 cycles~6 nm) and thick (1000 cycles~80 nm) layers confirm that samples at all temperatures consist of TiO 2 with comparable carbon contamination of the surface determined by the C 1s peak at 285 eV. The major titanium peak Ti 2p 3/2 is detected at a binding energy of 458.8 eV and can be assigned to an oxidation state of Ti 4+ , whereas the corresponding O 1s signal is in accordance shifted to 530.3 eV (see Figures 1 and 2 as well as Figure S2 to S4 in the SI) [29][30][31]. Deconvolution of the O 1s peak into an Ti 4+ (530.4 eV) and an Ti 3+ (532 eV) contribution, the Ti 3+ content is notably decreasing with increasing temperature, indicating a reduction of defects for both reactants [32,33].
SEM images of investigated layers for sample temperatures between 100 and 350°C are shown in Figure 3 (more details in Figure S5 to S16 in the SI). For H 2 O as reactant, deposited layers start to crystallize in the anatase phase at sample temperatures of 200°C. Thereby, an amorphous layer is formed prior to nucleation of small anatase crystallites, which is also evident from the transmission electron microscopy and the selected area electron diffraction of the corresponding regions shown in Figure 4. As temperature increases above 250°C, the crystallization process is strongly enhanced and crystallites with preferential growth orientation are formed. First, the crystal growth is leading to flat terrace-like grains with sizes of 200 to 300 nm. For further increased temperatures, the crystal shape is changing to a plate-like structure with dimensions of about 100 nm. A similar crystallization process was described for TiCl 4 as Ti-precursor in regular ALD systems [23,34,35] and recently was also  With O 3 as reactant, the crystallization process starts at higher temperatures and leads to significantly larger grain sizes of more than 1 μm. As in case of H 2 O, the grain-size decreases with increasing temperatures, but the terrace-like facets remain more discernable. The structural change combined with the temperature-dependent variation in growth rate suggest a transition from a space-limited (due to a small amount of nucleation sites) to a general crystal growth with increasing temperature, superimposed by Ostwald ripening at intermediate and decomposition of the precursor at elevated temperatures [34,36,37]. Furthermore, for both reac-tants this transition is associated with a change in surface roughness, which is increasing from about 1 to 10 nm root mean square (RMS). This is evident from AFM (see Figure S17 in the SI) measurements and consistent with fit parameters extracted from the utilized SE model (compare Table 1). The different results between H 2 O and O 3 in terms of morphology and growth rate are attributed to a different chemisorption of the incoming TTIP precursor and the respective surface reaction mechanism, as well as the resulting surface area available for the process [38][39][40][41].
For both reactants, the change from amorphous to the anatase structure is verified by Raman spectra, as seen in Figure 5.    [42,43]. The significant increase in Raman intensity with higher temperature is attributed to an increasing crystallinity as indicated from SEM images in Figure 3. This is also consistent with EDX measurements showing a raise in the atomic percentage of titanium and oxygen relative to the silicon substrate for increasing temperatures (see Figure S18 and Table S1 in the SI) due to an increase in layer density. Thereby, the atomic percentage of Ti is increasing linearly from 4.5 to 10% for H 2 O starting at deposition temperatures of 100°C, whereas for O 3 the values only start to rise above 200°C. This goes along with a decrease of the O : Ti ratio from about 3.5 to 2.8, assuming the contribution by a native SiO 2 layer on the Si substrate to be insignificant.
We note that intensities in Raman and EDX are sensitive to variations in layer density and thickness. However, the relatively small variation in layer thickness does not explain the significant increase of the Raman intensity towards higher process temperatures. In addition, XPS, which is not thickness-dependent, also shows increased signal intensities while the C 1s peak is not systematically affected (see Figure 2). Hence, the change of atomic percentage is attributed to an increased packing density due to enhanced crystallinity. Table 1 is summarizing relevant properties extracted from structural and optoelectronic measurement techniques.     International Journal of Photoenergy From UPS and KP measurements, the work function (WF) and valence band maximum (VBM) were determined at energies of about −4.6 eV and −7.5 eV, respectively. In combination with the optical bandgap determined from UV-Vis, the conduction band minimum of fabricated TiO 2 layers is located at an energy of about −3.9 to −4.0 eV and therefore the Fermi level is steadily 0.7 eV below the conduction band. The corresponding measurements are presented in Table 1 and Figure S24 in the SI.

Optical and Electronic
Meanwhile, the optical bandgap and the extinction coefficient are consistently shifting to lower energies with increasing deposition temperatures as expected for an increasing crystallinity [8], and the change in crystallinity results in a significant increase in refractive index from about 2.07 to 2.54 and 2.57 at 530 nm for H 2 O and O 3 , respectively. This has to be considered for light propagation and standing wave formation within solar cell devices, which can modulate the resulting short circuit current density (J SC ) [44,45].
The influence of varying TiO 2 properties on solar cell characteristics was exemplarily inspected on perovskitebased solar cells (MAPbI 3 , 2-step process) utilizing AP-SALD layers with the most promising conditions or variations. Fabricated devices have been compared to a commercial solgel reference without further optimization (see Figure 6), yielding flat, at about 70 nm thick compact layers with a surface roughness and morphology comparable to amorphous AP-SALD layers fabricated at temperatures of 200°C and below. More detailed device characteristics of fabricated solar cells are shown in Figures S25 to S30 in the SI. Compared to the reference sample prepared with commercial TiO 2 paste, the series resistance is higher and shunt resistance is lower in AP-SALD-based solar cells, which is directly reflected in a decreased fill factor. The slightly higher series resistance is attributed to the complete absence of unintentional doping and a more stoichiometric TiO 2 crystal formation with less oxygen vacancies during deposition. These vacancies are known to be responsible for the intrinsic conductivity of TiO 2 [46]. The decreased shunt resistance, on the other hand, is attributed to the TiO 2 interface of as-processed layers to the perovskite. The absence of dopants and impurities within the bulk and at the surface of the fabricated layers presumably leads to an unfavorable energy landscape for electron extraction, which was already observed for TiO 2 and other metal oxides [47][48][49]. Thus, a surface passivation by PCBA strongly enhanced open circuit voltage (V OC ) and fill factor of fabricated devices.
However, this impact of the interface is strongly interfering with a direct association of layer properties to device characteristics. Thus, in spite of significant changes in layer properties like layer thickness, morphology, crystallinity, and stoichiometry, clear systematic dependencies of device characteristics were largely superimposed and are mainly within our given uncertainty of device reproducibility.
Common strategies for interface passivation are the application of very thin organic [50][51][52] or inorganic [53,54] interface layers which are improving charge extraction and avoiding recombination. Also, ultrathin ALD layers of amorphous TiO 2 have been shown to act as suitable passivation layer [55].
Integrating this approach into the AP-SALD deposition by a continuous gradual change of temperature allows for an uninterrupted fabrication process of the TiO 2 , reducing complexity and avoiding a possible contamination of the surface between fabrication steps. Therefore, we investigated the influence of temperature gradients during deposition and compared them to layers deposited at fixed temperatures. For layers fabricated with a temperature gradient from 300 to 100°C we find a significant improvement in all device characteristics (see Figure 7) and attribute this to the formation of a core-shell structure with an amorphous layer covering a crystalline anatase film. This combines the good series resistance of the crystalline phase with an improved interface avoiding recombination [56].

Conclusion
The analysis of structural and optoelectronical characterization methods shows a significant influence of temperature and the chosen reactant on crystallinity and grain size of deposited AP-SALD layers. Thereby, low deposition temperatures result in smooth amorphous TiO 2 films, while elevated temperatures lead to polycrystalline anatase layers. With increasing deposition temperature, the preferential crystal growth is increasing the resulting surface area and roughness. Simultaneously, the improving crystallinity leads to a higher refractive index and lower optical band gap, though the energy levels of the layer are only slightly altered.
In fabricated perovskite-based solar cells, the impact of an unfavorable interface restricts a clear association of TiO 2 layer to specific changes in device characteristics and highlights once again the necessity of adequate interface engineering. Thus, a promising approach for gradual core-shell structures was demonstrated, leading to an amorphous passivation layer on top of highly crystalline anatase TiO 2 for improved solar cells.

Data Availability
Data of this work are available on reasonable request.

Conflicts of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Supplementary Materials
In the supplementary information we supply details on the applied methods for the TiO 2 preparation, perovskite solar cells preparation, applied measurement techniques, and more data for the structural analysis. Figure S1: layer thickness as a function of (a) deposition cycles and (b) temperature measured by profilometer, and (c) growth per cycle determined by VASE. Layers in (a) and (b) are fabricated with H 2 O as reactant, whereas GPC was determined for H 2 O and O 2 as reactant. Figure S2: XPS overview and detailed scans of TiO 2 with H 2 O as reactant at varying deposition temperatures. Figure S3: XPS overview and detailed scans of TiO 2 with O 3 as reactant at varying deposition temperatures. Figure S4: XPS overview and detailed scans of TiO 2 with H 2 O as reactant comparing a layer with 75 deposition cycles to a layer with 1000 deposition cycles.  Figure   8 International Journal of Photoenergy S11: SEM (InLens detector) of 1000 cycles SALD with O 3 as reactant at 100°C. Figure S12: SEM (InLens detector) of 1000 cycles SALD with O 3 as reactant at 150°C. Figure S13: SEM (InLens detector) of 1000 cycles SALD with O 3 as reactant at 200°C. Figure S14: SEM (InLens detector) of 1000 cycles SALD with O 3 as reactant at 250°C. Figure S15: SEM (InLens detector) of 1000 cycles SALD with O 3 as reactant at 300°C. Figure S16: SEM (InLens detector) of 1000 cycles SALD with O 3 as reactant at 350°C. Figure S17: AFM of 1000 cycles SALD on polished Si at varying deposition temperatures and reactants. Figure S18: atomic percent of Ti, Si, and O of 1000 cycles SALD at varying temperatures measured by EDX. Figure S19: UPS of TiO 2 fabricated with 1000 cycles and H 2 O as reactant at varying temperatures. Figure S20: UPS of TiO 2 fabricated with 1000 cycles and O 3 as reactant at varying temperatures. Figure S21: transmission spectra acquired by UV-Vis spectroscopy for TiO 2 layers fabricated with (a) H 2 O and (b) O 3 as reactant at varying deposition temperatures. Layer thickness and refractive index show a significant impact on standing-wave formation, while crystallinity is causing a shift in energy bandgap to lower energies. We note that measured absolute bandgap values of anatase are slightly higher than commonly reported in literature. Figure S22: dispersion relation of refractive index and extinction coefficient of TiO 2 deposited at varying temperatures and H 2 O as reactant. Figure S23: dispersion relation of refractive index and extinction coefficient of TiO 2 deposited at varying temperatures and O 3 as reactant. Figure S24: work function, valence band maximum, and conduction band minimum as a function of temperature measured by Kelvin probe, UPS, and UV-Vis. Figure S25: thickness dependent solar cell characteristics for SALD layers fabricated at 200°C and partially modified with PCBA. Figure S26: steady state measurement of solar cells presented in Figure S25. Figure S27: thickness and deposition temperature dependent solar cell characteristics for SALD layers modified with PCBA. Figure S28: steady state measurement of solar cells presented in Figure S27. Figure S29: solar cell characteristics for SALD layers fabricated at fixed and gradient temperatures, modified with PCBA. Figure S30: steady state measurement of solar cells presented in Figure S29. Figure S31: SEM (InLens detector) of 1000 cycles SALD with H2O as reactant for a continuous temperature gradient from 300 to 100°C. (Supplementary Materials)