Dry-pressed anodized titania nanotube/CH 3 NH 3 PbI 3 single crystal heterojunctions: the beneficial role of N doping

Highly ordered, anodically grown TiO 2 nanotubes on titanium supports were annealed in ammonia atmosphere in order to incorporate nitrogen doping (˂ 2 at.%) in the titanium oxide lattice. FESEM micrographs revealed nanotubes with an average outer diameter of 101.5 ± 1.5 nm and an average wall thickness of about 13 nm. Anatase crystals were formed inside the tubes after annealing in ammonia atmosphere for 30 min. With further annealing, rutile peaks appeared due to the thermal oxidation of the foil and rise as the duration of heat treatment was increased. The concentration and chemical nature of nitrogen in the nanotube arrays can be correlated to the optical response of dry-pressed heterojunctions of doped TiO 2 /CH 3 NH 3 PbI 3 single crystals. The N-TiO 2 /perovskite heterojunction with the highest amount of interstitial nitrogen exhibited an improved photocurrent, indicating the importance of the semiconductor doping-based heterojunction optimization strategies to deliver competitive levels of halide perovskite-based optoelectronic devices to be envisioned for urban infrastructures.


INTRODUCTION
Great efforts of scientists are focused on enhancing solar energy utilization worldwide.
Among the numerous photoactive materials studied, titanium dioxide (TiO 2 ) has been proven highly promising, since it is chemically stable, cheap, non-toxic and available in abundance. However, in specific applications, such as semiconductor photocatalysis or photodetection, the material faces considerable limitations due to its light absorption edge lying at 387 nm. A significant body of recent research has been focused on extending the effective light absorption range into the visible region.
A key technique for narrowing the band gap of TiO 2 is the incorporation of a dopant atom such as transition metal [1,2] or nonmetal [3][4][5] in the crystal structure. Among various elements, nitrogen seems to be an appropriate choice, due to its atomic size similar to that of oxygen, small ionization energy and as it is defined by metastable center formation and stability [6]. Many reports have been published about the incorporation of nitrogen into anodized TiO 2 nanotubes by the hydrothermal method [7,8], immersing in nitrogen containing solution [9], ion implantation [10,11], anodization in nitrogen containing electrolyte [12] or thermal treatment in N 2 atmosphere [13], in order to improve the photoelectric or photocatalytic efficiency (Table S1).
Furthermore, several research groups synthesized N-TiO 2 by exposing TiO 2 to hot ammonia gas and studied the influence of the annealing temperature (300 to 600 °C) [14] and the combination of air-ammonia atmosphere [15]. The highest photocurrents were observed in the temperature range of 500-600 °C [14] with mixed, anatase and rutile phase TiO 2 having up to 9 at.% nitrogen incorporated into the lattice [16]. 5 Recently, Bjelajac et al. [17] characterized nitrogen-doped TiO 2 nanotubes decorated with CdS quantum dots for the potential use as a photoanode in quantum dots sensitized solar cells. They anodized the titanium film deposited on FTO glass and annealed it at 450 °C, 30 minutes in NH 3 flow. Analyzed XPS spectra proved the incorporation of nitrogen in the lattice forming TiO 2-x N x , with an N:Ti ratio of ~1:100. Consequently, a redshift from 380 nm to 507 nm was observed.
Unlike the above-listed doping techniques, which require the dopant cation or anion to be part of the hosting lattice, the decoration-based sensitization techniques rely on the creation of new interfaces. Typically, sensitization of the TiO 2 surface can be achieved by depositing different species such as Ag, Au [18], C-N=C, C-NH 2 fragments [19], metal complexes [20], organic dyes [21], quantum dots [22] and organometallic salts [23][24][25] or coating nanotubes homogenously with a thin layer of light absorber material [26,27].
Therefore, it shows a great potential to be used in light-emitting diodes [31], lasers [32], solar cells [28], magneto-optical data storage [33], visible light [34,35], X-ray [36] and even high energy particle detection [37]. Most of the aforementioned devices contain polycrystalline thin films that demonstrated limited optical and physical properties due to grain boundaries and higher defect density. Recently, device performance has been improved by using perovskite single crystals. Due to the absence of grain bounders, low 6 trap densities (~10 -9 cm -³) and long diffusion lengths (3 mm) which decreases recombination [38,39].
In the majority of these applications, the light-sensitive heterojunctions are formed by interfacing the hybrid perovskite with planar or mesoporous films of titania. The textural properties of the mesoporous films will depend on the size and shape (spherical, tubular, rod or wire) of its nanoparticle building blocks. The nanotubular morphology may offer several advantages over nanoparticles: a 30-fold greater electron diffusion length, reduced charge recombination and faster electron transport, due to which the electron energy loss is reduced [24,40,41]. The anodization technique is frequently used for the synthesis of highly aligned arrays of TiO 2 nanotubes perpendicular to the substrate [42]. A significant body of research suggest that by varying the synthesis parameters and post-synthesis treatments one can control and predict the morphological characteristics and the doping level of TiO 2 nanotubes produced by anodization.
Therefore, our aim is to investigate the influence of the annealing time in ammonia atmosphere on the concentration and chemical nature of nitrogen atoms incorporated in the nanotube arrays. Additionally, we seek to provide answers to whether or not the optical response of dry-pressed heterojunctions of doped TiO 2 /CH 3 NH 3 PbI 3 single crystals can be correlated to the amount and chemical nature of nitrogen-dopant incorporated in the TiO 2 nanotube arrays.

Synthesis of undoped and doped TiO 2 nanotubes
7 Titanium foils (10 mm × 15 mm × 0.25 mm, 99.7 %, Aldrich) were degreased by sonication in acetone, ethanol and deionized water (DI) for 60 min (20 min in each). This was followed by rinsing with DI water and drying in air. The TiO 2 nanotube arrays were synthesized in an electrolyte consisting of 0.6 mM HF and 0.1 mM CH 3 COOH in DI water according to the previously reported approach [43]. The electrochemical anodization was carried out in a two-electrode cell using a DC power source (PEQLAB EV 231), where a Ti foil was used as an anode and a thin platinum foil as a counter electrode. The anodization was performed at room temperature, for 30 minutes under 15 V, with mild stirring of the electrolyte. After anodization, the samples were cleaned with DI water and then dried in air.

The synthesis of methylammonium lead iodide perovskite
For the preparation of perovskite single crystals, the precipitation method was used starting from a concentrated aqueous solution [44]  days. An optical image of grey, rhombohedral MAPbI 3 single crystals with 3-5 mm silver gray facets is shown in Figure 1b.

Preparation of TiO 2 /CH 3 NH 3 PbI 3 single crystal photodiodes
The heterojunctions were prepared by dry pressing of a bulk, millimeter-sized MAPbI 3 single crystal against the surface of TiO 2 and N-TiO 2 nanotubes. Pushed electrical contacts were applied by using tungsten needles. The tungsten needles were contacted with the top part of the MAPbI 3 single crystal and titanium plate back-electrode. A schematic representation of the interface and measurement setup can be seen on Figure 1c and d, respectively.

Characterization
The surface morphology of the TiO 2 nanotube arrays was studied using a Tescan MIRA3 XMU FESEM. The UV-Vis diffusion reflectance spectra were recorded in the wavelength range of 300-800 nm on the Shimadzu UV-2600 spectrophotometer equipped with an integrated sphere. X-ray diffraction patterns were determined by Rigaku Ultima IV diffractometer, in a 2θ range from 23 to 60° with a step size of 0.05° and time of 3 s.
Grazing incidence angle was set at 2° to enhance the signal from the film. The X-ray photoelectron spectroscopy was carried out on a SPECS customized UHV surface analysis system containing sputter ion gun, PHOIBOS 100 spectrometer for energy analysis, dual anode Al/Ag monochromatic source and electron flood gun. The XPS spectra were taken using a monochromatic Al Kα line (photon energy of 1486.74 eV) operated at 400 W. The presented results were obtained after the sputtering of the surface in order to reduce the contribution of atmospheric pollution. The sample sputter cleaning was performed using a 3 keV Ar + ion beam, the duration was 1 min in all cases in order to reduce the preferential sputtering effects which typically, in the case of TiO x samples, reduce 'x'. All binding energies were referenced to the C 1s peak at 284.8 eV of surface adventitious carbon.

Photodetector device characterization
Two point resistivity measurements (d.c.) were carried out using a National Instruments

RESULTS AND DISCUSSION
Halide perovskite crystallites are relatively soft materials [45,46], and, hence efficiently sinter together exposed to a various range of pressures at room temperature. For instance, Mettan et al. [44] demonstrated that MASnI 3 crystallites compressed into pellets allow reliable transport measurements and achieve high ZT values of 0.13 at room temperature.
Moreover, Pisoni et al. [47] demonstrated that the resistivity of polycrystalline samples obtained by pressing together an assembly of microcrystallites is only a factor of 2 higher than that of a high-quality single crystal. Even devices, such as perovskite-based photodiodes were recently fabricated by small pressure-assisted mechanical contacts, i.e. by dry pressed perovskite single crystals or pressed pellets on CNT electrodes [35,48]. These mechanically assembled photodiodes detected visible light in the μW/cm 2 intensity range and exhibit an ideality factor bellow 10. Additionally, pressed pellets have been recently applied to build high-performance X-ray photodetectors [49]. These pioneering works suggest, that the quality pressure-assisted mechanical contacts is surprisingly good, and therefore these type of, so far rarely studied heterojunctions have been used in this work.
The heterojunctions were created by dry pressing of a 3 mm MAPbI 3 single crystal on top of the undoped and N-doped TiO 2 nanotubes (Figure 1c and d). In order to investigate the effect of N doping on the photodiode performance, the tungsten needles, which served as electrical leads, were pressed on the top surface of the MAPbI 3 single crystal and onto the Ti foil back electrode, respectively. Highly ordered nanotubes arrays perpendicular to the Ti foil were obtained during anodization (Figure 1a). To study the morphology evolution during heat treatment in ammonium atmosphere, micrographs were statistically analyzed ( Figure S1). It is noteworthy that there was no significant difference in the average pore diameter and the average wall thickness between the nitrogen doped and undoped TiO 2 samples, as can be seen from Table S2. On the other hand, slight differences in outer diameter distribution are observed ( Figure S1). A lower Young's modulus [50] of TiO 2 nanotubes compared to TiO 2 films allows the nanotubes to better accommodate the MAPbI 3 single crystal, thus leading to larger contact areas. Influence of the annealing atmosphere on the crystal structure of titania nanotubes was followed by XRD measurements. The sample annealed in air for 30 min shows the presence of anatase peaks at ~ 25°, 48° and 55° with a low-intensity peak at 27° for the rutile phase ( Figure 2). However, when annealing in an  Table S3. A slight increase in anatase crystal size can be observed with longer durations of annealing. XPS studies were performed in order to investigate the chemical state of nitrogen incorporated in the TiO 2 structure. Table 1 summarizes the total amount of elements in the samples after sputter cleaning. Because of standard surface contamination, a decrease in the amount of carbon after sputtering was registered for all samples. From Table 1, the titanium foil and TiO 2 -undoped sample show a nitrogen peak, just like the samples annealed in NH 3 atmosphere, indicating that the titanium foil contained nitrogen in traces before its exposure to the NH 3 atmosphere. In order to reveal the nature of nitrogen in the titanium foil, N 1s peak ( Figure S2) was deconvoluted in three components: 396.8, 398.8 and 400.6 eV. The peak at 396.8 eV indicates the presence of nitrogen inside the Ti foil because a peak around 396 eV is commonly ascribed to the Ti-N bond. The peak at 398.8 eV is usually associated with Ti-O-N or N-H bond [54], while the peak at 400.6 eV indicates NO species, nitrogen bound to various surface oxygen sites [55]. Therefore, the nitrogen is present inside the foil and on the surface of the foil. On the other hand, the TiO 2 -undoped sample showed only two peaks: 399.8 and 401.9 eV, without the peak around 396 eV (Table S4). The peak at 399.7 eV indicates the incorporation of nitrogen in some form into the lattice on interstitial sites [56][57][58][59]. Hence, it can be assumed that the peak at 399.8 eV observed in the undoped TiO 2 was due to nitrogen diffusion from the titanium substrate to the TiO 2 nanotubes during the anodization process. The peak around ~401 eV is assigned to molecularly adsorbed nitrogen on the surface of the samples, and it indicates contamination [60]. Finally, for nitrogen-doped TiO 2 samples three N 1s peaks, at ~ 396 eV, ~ 399.7 eV and ~ 401.9 eV, were observed. The binding energy of 396 eV can be attributed to anion N 3that substituted some oxygen site [56,60,61], while the 399.7 and 401.9 eV to interstitial nitrogen and surface contamination, respectively. The total amount of nitrogen decreases with longer heat treatment in ammonia atmosphere, as it is shown in Table 1. Although the nitrogen content decreases, it was observed that the amount of substitutional nitrogen increases (Table S4). This can be due to the fact that a greater amount of nitrogen diffuses from the atmosphere into the structure, resulting in a drop of the concentration gradient of nitrogen with prolonged annealing time. Furthermore, the decreased gradient of concentration is 13 accompanied by a decreased diffusion from the substrate and a subsequent decrease of the amount of interstitial nitrogen.
In addition, it should not be neglected, that interstitial nitrogen diffuses to substitutional sites with longer exposure to ammonium atmosphere. Thus, the amount of substitution nitrogen increases, while the amount of interstitial nitrogen decreases. Moreover, the amount of chemisorbed nitrogen decreases with prolonged annealing time. Therefore, the total amount of nitrogen becomes lower (Figure 3b). In order to determine the influence of nitrogen doping on the optical response of TiO 2 , DRS analyses were performed. As evident in Figure 4a, all samples annealed in ammonia atmosphere exhibit a shoulder in the absorption spectra towards higher wavelengths. The same is not present in the pristine sample. The most pronounced increase in absorption is for the TiO 2 -N30, which contains the highest proportion of nitrogen. This is presumably caused by introducing new localized states in the band gap of TiO 2 due to nitrogen doping.
According to Di Valentin et al. [56] substitutional nitrogen form localized state at 0.14 eV above the valence band edge. While interstitial nitrogen form occupied antibonding states below O2p valance band and bonding states at 0.73 eV above the valence band.
Furthermore, I-V measurements were performed exclusively on the TiO 2 nanotube samples before the addition of perovskite single crystals, to assess whether or not nitrogen doping leads to an increased conductivity as compared to the pristine titanium oxide 14 nanotubes. The resistivity of both, the pristine and doped samples were comparable and above GΩ approaching the limitations of our two-point resistivity measurement setup.
Additionally, impedance spectroscopy measurements confirmed, as seen in Figure 4b, that the conductance of the doped and undoped samples was identical. This suggests that the improved detectivity of the fabricated photodetector devices is not due to the conductivity difference of the TiO 2 electrodes. correlation between photocurrent and the amount of interstitial nitrogen. A similar phenomenon was also observed by Cabrera et al. [62]. They synthesized N-TiO 2 -films with different amount of interstitial and substitutional nitrogen for photodegradation of stearic aide. An increase in photoactivity was observed for films that contained the highest amount of N i . Moreover, the film with the lowest level of N i was the same as the undoped reference sample despite possessing the highest level of N s . Furthermore, it has been demonstrated that TiO 2 doped with interstitial nitrogen is 30% more photocatalytic active than TiO 2 doped with substitutional nitrogen [63]. This is due to better stability of photo-generated electron-holes in interstitially doped anatase while the tendency for recombination is greater in substitutionally doped samples.
To assure that the increase in photocurrent of the perovskite/TiO 2 device is due to the increase in nitrogen content rather than an increase of O-vacancies, titanium 2p and oxygen 1s core levels were studied. As shown in Table S5, the position of the Ti 2p 1/2 biding energy at 458.8±0.1 eV for undoped and N-doped clearly indicates that Ti is in the 4 + valence state.
There is no decrease in the Ti 2p binding energy after nitrogen doping which could indicate the formation of the Ti 3+ state (usually in combination with oxygen vacancies). Furthermore, deconvolution of the O 1s peak reveals that all samples have three contributions: 530.1±0.1 eV which proves the O 2state in the crystal lattice; 531.5±0.1 eV attributed to the chemisorbed surface hydroxyl groups; and ~ 532.5 eV which is probably due to weakly adsorbed species [64].
These optoelectronic results are in good agreement with the XPS and DRS analysis. The TiO 2 -N30 sample has the highest N i content and the largest absorbance red shift, which results in the highest photocurrent. Consequently, with the decrease of interstitial nitrogen content, the optical and electrical characteristics worsen. Furthermore, our results confirm that the amount of interstitial nitrogen is governing the increase of photocurrent. As an example the TiO 2 -N90 sample, although possessing a higher substitutional nitrogen content then its TiO 2 -N30 counterpart, exhibits lower photocurrent.
Lastly, on-off measurements were done under ambient conditions (Figure 5b). As expected, higher values of photocurrent were achieved using doped TiO 2 nanotubes. In agreement with the I-V measurements, the TiO 2 -N30/perovskite device had the best performance, having ~230 % higher photoresponse compared to the undoped reference sample. Typical figures of merit of a photodetector device, the responsivity R and response times, t rise and t fall , were calculated. R is defined as the ratio of the photocurrent and the intensity of light, while the response times are defined as the times elapsed for the change of the photocurrent intensity from 10 % to 90 % of the peak output photocurrent amplitude.
The on-off curves show a fast response regime increasing the current to 75 % of its maximum value. The current then continues to increase, saturating to the final current value. This leads to a rise time of 2.65 s. When the light source is turned off, the decay time is short -0.33 s. The longer rise time is most probably due to ion migration (pooling) effect in the perovskite single crystal [65], as the measurements were done at relatively high bias voltages. The responsivity for each junction is shown in Figure 5c. Responsivity off the TiO 2 -N30 sample is more than double the value for its undoped counterpart. Lastly, the ideality factor of the photodiode was calculated from the I-V measurements and is shown in dependence of the applied voltage in Figure S4. showed that by introducing nitrogen atoms in the structure of TiO 2 nanorods the efficiency of the solar cell increased by 14.7 %, compared to its counterpart using an undoped TiO 2 ETL. In order to explain this, they studied photoluminescence (PL) and time-resolved photoluminescence (TRPL) spectra of undoped TiO 2 /MAPbI 3-x Cl x and N-TiO 2 /MAPbI 3- x Cl x . PL and TRPL spectra showed better electron transfer on the N-TiO 2 /MAPbI 3-x Cl x interface than on TiO 2 /MAPbI 3-x Cl x interface. In addition, lower resistance and larger recombination resistance of perovskite solar cells based on N-TiO 2 ETL were demonstrated by electrochemical impedance spectroscopy (EIS). The reason for that is the decreasing of the trap density, due to the lowering of the oxygen vacancies concentration.
Therefore, we believe that the reasons behind the increase of photocurrent and responsivity of the fabricated photodiodes using N-doped TiO 2 nanotube electrodes are in line with the conclusions of these earlier works. Effective electron transfer and lower resistance of the devices with an increase of interstitial nitrogen doping result in better device performance. Future measurements, such as PL, TRPL and EIS could prove these statements and clarify the electron transfer process between the perovskite single crystal and the N-doped TiO 2 nanotube electrodes in detail. Nitrogen-doped TiO 2 nanotube electrodes can, therefore, be attractive as electrodes for future perovskite-based optoelectronic device fabrication.

CONCLUSIONS
Well-aligned TiO 2 nanotube arrays were prepared by anodization of titanium foils in an acid electrolyte. Nitrogen doping has been introduced in the titanium oxide nanotubes via high-temperature crystallization in pure NH 3 atmosphere. It has been revealed that the duration of the heat treatment has a direct impact on the amount and nature of the incorporated nitrogen dopant. XPS study demonstrated that nitrogen is incorporated into the nanotubes at both the substitutional and the interstitial sites. The ratios of substitutional and interstitial nitrogen vary as a consequence of diffusion from the titanium substrate and from the ammonia atmosphere as well. Photodiodes have been fabricated by pressing the pristine and N-doped anodized titania nanotubes and millimeter-sized lead methylammonium triiodide single crystals. The photodetectors prepared from the nitrogen-doped samples annealed at 450 o C degrees for 30 minutes showed a greatly improved optical response, highlighting the future potential of the semiconductor doping-based (interstitial and substitutional) heterojunction optimization strategies in the development of halide perovskite-based optoelectronic devices.