Introduction

The perovskite can be represented by the formula ABX3, where A is a monovalent cation of Cs (cesium), MA (methylammonium), FA (formamidinium), B is most often cation Pb (Lead), rarely Sn (Tin), and X is a halide anion (Br, Cl, I) [1]. Perovskites owe their attractiveness in photovoltaics to its extraordinary properties such as long lifetime of carriers, high optical absorption, ease of band gap control, tolerance to defects in small concentrations and low-cost production [2]. Perovskites are divided into two groups, a much more popular organic–inorganic (also called hybrid) and inorganic, which differ in the A-site cation. The A-cation in organic–inorganic group is most often FA, MA or a mixture of them. Their advantage over solely Cs-based inorganic perovskite is larger ionic radius that facilitate obtaining the desired cubic structure at room temperature according to Goldschmid’s tolerance factor. Having the larger A-site cation allows to grow easily the perovskites with larger iodine halide instead of bromine. In consequence, the iodine results in perovskite materials with narrower energy gaps, closer to the optimal energy gap according to Shockley–Queisser limit [3], enable to achieve higher efficiencies than inorganic cells. On the other hand, inorganic perovskites based on Cs and Br display the wider band gaps and suffer from lower efficiency limits. Nevertheless, their big advantage over the hybrid perovskites is greater stability under long exposure conditions, resistance to high temperatures [4], resistance to degradation in air [5], phase stability [6] as well as resistance to electron beam [1]. The inorganic perovskites have some potential new application and are worthy of interest. A numerous studies were conducted on them due to their greater stability compared to organic perovskites, which gives a better chance of obtaining durable cells suitable for commercialization. ETL (electron transport layer), a crucial component in the perovskite cells, is responsible for transporting the electrons generated in the absorber layer to the electrode while blocking holes [7, 8]. In cells with the n-i-p structure, ETMs (electron transport materials) must avoid optical losses, hence material with a wide energy gap, i.e. greater than 3 eV, should be selected. In the case of ETL, energy band alignment is another important feature. A conduction band edge of ETL should be located below the conduction band edge of the perovskite, which will enable the transport of electrons from perovskite to the electrode (Fig. 1a). At the same time, the energy difference between them should be as small as possible to avoid significant voltage losses. On the other hand, the ETL valence band edge should be located much lower than the perovskite valence band edge in order to prevent holes from flowing in the same direction as electrons to avoid their recombination. Well-matched interfaces will allow the collection of photogenerated charges from the absorber layer with low voltage losses [9]. In order to enable fast and effective transport of electrons to electrode, the material should be characterized by high electron mobility and low recombination. Therefore, the ETL should be as thin as possible while ensuring good coverage of the TCO (transparent conductive oxide) electrode and effective hole blocking; such films can be deposited conformally and non-conformally depending on kind of deposition method. In case of non-conformal deposition, the optimal thickness will also depend on the used TCO electrode (mainly roughness).

Figure 1
figure 1

a Band arrangement of the cell, b cell manufacturing process diagram.

Full size image

Titanium dioxide (TiO2) is a n-type semiconductor with an electron affinity around 4 eV, band gap 3.2 eV and effective density of states in conductive band 1021 cm−3 [10]. Titanium dioxide layers are the most popular electron transport layer (ETL) in perovskite solar cells [11, 12], especially in inorganic devices due to its stability and high optical transparency [13]. There are few methods we can use to produce c-TiO2 (compact, planar layer of TiO2 responsible of selective transport) like atomic layer depositions [14], spray pyrolysis [15], spin coating [9], dip coating [16], magnetron sputtering [17], etc. The major way to produce thin TiO2 layer are solvent methods especially spin coating because of it fast production time, low cost and easy application in the R&D stage. The main challenge associated with solvent-based methods, such as spin coating, is the non-conformal deposition on substrates; for instance, when depositing titanium oxide onto a rough surface like FTO (fluorine-doped tin oxide), the TiO2 layer tends to preferentially accumulate within troughs rather than evenly coating the peaks of the surface [9]. In case of perovskite solar cells, in addition to the planar structure that includes only a c-TiO2 layer, a more common approach involves utilizing a bilayer mesoporous structure composed of both c-TiO2 and m-TiO2 (mesoporous TiO2). Mesoporous layer is a scaffold for the growth of perovskite and often it covers some imperfections of the planar layer like uneven coverage or small holes. In planar structures, the morphology and quality of the obtained ELT layers is much more critical, despite that, planar structures are being intensively developed because of lower costs and simpler production. Eliminating the mesoporous layer eliminates an additional high-temperature process, making the technology more commercially friendly. Therefore c-TiO2 has to be even more well-optimized and refined in case of its process parameters. Poespawati and team deposited a TiO2 layer with different spin speeds, in a composition with a MAPbCl2I hybrid perovskite; they achieved a cell with 7.2% efficiency at 3000 rpm [18], Xigyue Liu and team used a water-based chemical bath to create the TiO2 layer and produced cells with CsPbBr3 with an efficiency of 7.59% [19]. Thanks to the use of the ALD technology, which allows to obtain conformal layers, Lu and his team achieved an efficiency of 13.6% with a TiO2 layer thickness of 10 nm in a cell with an organic–inorganic perovskite MAPbI3−xClx, cells made with a thicker TiO2 layer obtained from spin coating with the same perovskite achieved 8.61% efficiency [20]. Slawek et al. investigated optimal thickness of spin-coated bis(acetylacetonate) diisopropoxide precursor for planar MAPbI3 cells proving its strong quasi-parabolic dependency on efficiency [9]. Zhao and team compared the layers deposited with the spin coating from the precursor titanium isopropoxide with the layer deposited via magnetron sputtering. The cells made with CsPbBr3 as an absorber achieved 4.81% and 5.48% efficiency, respectively, at a high open-circuit voltage of 1.385 V and 1.439 V, respectively [17]. Various precursors are used to obtain TiO2 layers; however, the titanium(IV) bis(acetylacetonate) diisopropoxide [9], titanium n-butoxide [21], and titanium isopropoxide [22] can be mentioned as the most frequently reported in the literature. The parameters of cells made on TiO2 layers depend mainly on their thickness, but also on the type of perovskite and titanium dioxide precursor. In work of Qin and team, three different precursors were tested: tetrabutyl titanate in ethanol, titanium diisopropoxide bis (acetylacetonate) in 1-butanol and titanium isopropoxide in isopropanol. Planar layers obtained by spin coating from previous precursors were coated by mesoporous TiO2 layer. Cells with organic–inorganic perovskite as absorber achieve high performance of 17.88% for variant with tetrabutyl titanate as compact layer, which has high conductivity, good optical properties and small charge recombination [23]. In the above works, the authors did not considered carefully the types of FTO substrates used, and the impact of their roughness on the optimal thickness of ETL layers and the further construction of subsequent layers. In addition to roughness, commercial FTO substrates differ in the glass used, conductivity and thickness of the FTO layers themselves, all these factors have a significant impact on the operation of the photovoltaic cells made with their use. Most of the works on the use of TiO2 as a layer blocking electrons applied by the sol–gel method used in a system with mesoporous layers, whose application is cumbersome process and requires high-temperature, however, helps in the construction of subsequent layers by hiding some imperfections of the previous layers and anchoring perovskite. Recent studies on titanium dioxide for perovskite solar cells includes application of different forms mesoporous building blocks like nanorods, nanofiber in contrast to basic nanoparticles [24]. Another strategy is doping [25] or interfacial engineering [26, 27].

For CsPbBr3 inorganic perovskites vast majority of scientific reports concerned the mesoporous structure [28,29,30]. Here, we investigated a planar structure and draw attention on some meaningful details. Much of the research in this field is based on the popular titanium isopropoxide precursor, while the titanium ethoxide precursor is much less commonly used, despite its similar properties and applications. In the following work, two precursors titanium(IV) ethoxide and frequently used titanium(IV) bis(acetylacetonate) diisopropoxide were tested in a wide range of thicknesses. The influence of the precursor on the morphology and electrical parameters of the cells was presented, and simulations were performed to explain the issue of the optical fit.

Experimental section

Materials

Materials used are as follows: Titanium(IV) bis(acetylacetonate) diisopropoxide 75%, Sigma-Aldrich, butanol, Chempure, ethanol p.a., Chempure, titanium(IV) ethoxide 99%, Merck, HCl 35–38%, Chempure, lead (II) bromide 98%, TCI, DMF p.a., Chempure, cesium bromide 99%, TCI, methanol p.a., Chempure, spiro-OMETAD 99%, Sigma-Aldrich, 4-tert-butylpyridine (4TBP) 98%, Sigma-Aldrich, bis(trifluoromethane)sulfonimide lithium salt (LiTFSI) 99%, Sigma-Aldrich, isopropanol p.a., Stanlab, pure gold, FTO Greatcell solar materials TEC 8, chlorobenzene p.a., Chempure, polished silicon wafers, Hellmanex® III HellmaAnalytics. Merc., and acetonitrile p.a., Chempure.

Preparation of TiO2 layers and solar cell fabrication

The manufacturing procedure is shown in Fig. 1b. FTO glass substrates (TEC 8) were washed sequentially in Hellmanex solution, isopropanol and demineralized water for 15 min and dried. The washed substrates were treated with O2 plasma for 15 min. Based on preliminary adjustments, solutions with HCl stabilization were prepared for each tested variant with the following concentrations: titanium(IV) bis(acetylacetonate) diisopropoxide (Tiacac) 0.30, 0.41, 0.48, 0.55, and 0.66 M/dm3 and titanium(IV) ethoxide (Tieth) 0.38, 0.50, 0.59, 0.67, and 0.79 M/dm3. After filtering the solutions with 0.22 µm PTFE filters, layers were deposited on previously prepared FTO substrate (2 × 2 cm) and additional polished Si wafers using spin coating 3000 rpm 15 s. The substrates were then preheated at 200 °C for 10 min followed by calcined at 500 °C for 30 min in tube furnace. The perovskite layer was applied by sequential spin coating. After O2 plasma treatment for 15 min, 1 M r–r PbBr2 in DMF was spin-coated on substrates 2000 rpm 30 s and annealed at 80 °C for 30 min, followed by spin coating 2000 rpm 30 s of 0.07 M solution CsBr in methanol and annealed at 250 °C for 5 min, and the step was repeated 10 times to obtain a CsPbBr3. The obtained perovskite layer was then washed with isopropanol and heated at 250 °C for 5 min. A layer of spiro-OMETAD was deposited with solution of spiro-MeOTAD (0.2944 g of spiro-MeOTAD, 4 ml of chlorobenzene, 118 µl of 4 TBP, and 66 µl of LiTFSI solution (520 mg, 1 ml of acetonitrile)) in the glow box in a nitrogen atmosphere at 2000 rpm 30 s. As the last step, a 80-nm-thick gold electrode was deposited by thermal evaporation. The device area is 0.25 cm2.

Characterization

The nominal thickness and refractive index of the obtained TiO2 layers were measured on a Sentech SE800 PV spectroscopic ellipsometer at the angle of 70° in range of 280–980 nm. As an optical model, air/roughness/TiO2/SiO2/Si stackers were used for adjustments. To describe the optical properties of the TiO2 layer, the Tauc–Lorentz model with four oscillators was used. Haze coefficient were determined from transmittance measurements, obtained in range of 300–800 nm, using a UV–Vis–NIR PerkinElmer Lambda 950S spectrophotometer. The morphology and roughness of the layers were determined using Innova multimode atomic force microscopy and scanning electron microscopy tabletop TM3030 (15 kV), Hitachi High-Tech. Moreover, the microstructure characterization of perovskite solar cell was done by a TECNAI G2 SuperTWIN FEG (200 kV) transmission electron microscope (TEM), whereas thin foil for TEM analysis was prepared using focused ion beam (FIB) technique. Photo-emission Tech AAA class solar simulator under standard test conditions (AM 1.5 G, 100 mW cm−2, 25 °C) was used to determine the cell operating parameters (device area 0.25 cm2). Electrochemical impedance (EIS) measurements were performed in an LASC probe station using a potentiostat (Bio-Logic SP-150e) equipped with a frequency response analyser. We measured the EIS in the frequency range from 1 MHz to 100 mHz under dark conditions by applying small-perturbation signal of 30 mV with a DC bias of 1 Volt. X-ray diffraction (XRD) patterns were measured by applying Bragg–Brentano (Ɵ–2Ɵ) geometry in Bruker D8 Discover diffractometer with CoKα radiation. The Ansys Lumerical FDTD software were used for optical simulations. As a simulated model, glass/FTO/SnO2/SiO2/TiO2/air stackers were used, simulated data were collected from 4 µm2 simulated surface.

Results and discussion

Precursor solutions

The HCl addition is required to prevent direct hydrolysis during mixing, as well as to avoid the ageing process during storage and improve the repeatability [31]. It was found that 0.4 ml HCl per 10 ml of ethanol is sufficient for titanium ethoxide in ethanol. In case of Tiacac, the addition of 0.2 ml of HCl per 10 ml of butanol was enough to stop changes in solution. In order to determine the nominal thickness of the obtained layers, TiO2 solutions were deposited on polished silicon substrates under the same conditions as on FTO glasses. The thickness of the obtained layers and the complex refractive index were measured using spectral ellipsometry techniques. The layers obtained from the Tiacac precursor have a lower refractive index than Tieth, 2.22 and 2.28, respectively, making it better suited optically to the other layers, which will be discussed later in this article. For both precursors, the refractive index is practically constant regardless of thickness (Table 1).

Table 1 Thickness and refractive index in TiO2 layers versus precursor concentration. Refractive index given at λ = 632.8 nm
Full size table

It was found that relation of final layer thickness with respect to precursor concentration was different for two precursors as shown by slope of linear fit in Fig. 2. A factor that may cause this discrepancy due to the applied spin coating method may be the viscosity of the solutions. Kinematic viscosity tests with Ubbelohde’s viscometer revealed that Tiacac solutions exhibited higher viscosity. Furthermore, the viscosity of the tested solution primarily depends on the viscosity of the solvent (supplementary information Table S1). Higher viscosity was confirmed for Tiacac precursor; however, this stood in is opposition to the obtained results. For this reason, an additional experiment was conducted in which solvents in solutions were swapped to obtain samples from both precursors. As a result of these studies, it was found that the layers obtained from Tieth solutions in butanol, at the same concentration (0.406 M/dm3) as in the case of ethanol were thinner: from ethanol solvent a layer with a thickness was 48.5 nm, while for butanol, only 37.1 nm was achieved. In the case of Tiacac, the layers obtained from solutions in ethanol were thicker than those from solutions in butanol, for a concentration of 0.505 M/dm3 in ethanol a layer of 75.6 nm was achieved, and for butanol 53.4 nm. It can be seen the dominant influence of type of solvent despite the confusing viscosity results.

Figure 2
figure 2

Correlation between thickness and solution concentration.

Full size image

In order to explain this phenomenon, a hypothesis was proposed regarding the influence of solvent evaporation rate during centrifugation and the dynamic concentration of the precursor. Ethanol has a boiling point of 78.4 °C, while butanol has a boiling point of 117.7 °C. Ethanol-based solutions allow thicker layers to be achieved because ethanol evaporates faster than butanol, i.e. already during centrifugation, drastically increasing viscosity and blocking further thinning of the solution layer.

Morphology of samples

The specificity of applying layers using the spin coating method is the uneven distribution of the layer thickness; during centrifugation, on rough substrates (as in the case of TEC 8 substrates), precursor more likely remains in the valleys than on the tops, thus low level of conformality is usually obtained. Thickness tests using an ellipsometer are carried out on polished silicon wafers, which gives us the result of only nominal thickness for comparative purposes. The TEM cross section showed nanocrystalline microstructure of TiO2 layer deposited on the rough FTO substrate (Fig. 3.). Observations are allowed on the measurement of the thickness of the TiO2 layer, and the image concerns a sample for the precursor Tiacac with a nominal thickness of 62.8 nm. In this example at the top of the FTO roughness, it is about 35 nm thick, and in the valley between the vertices, it is twice as thick about 70 nm. Significant differences between point thickness measurements (35 nm) are characteristic of the deposition method and the used substrate.

Figure 3
figure 3

Cross section of cell with 62.75 nm TiO2 from Tiacac precursor.

Full size image

In another step, the concentrations were adjusted in order to investigate the coverage of FTO roughness by TiO2 ETL of similar thicknesses. To assess the quality of FTO coverage and surface roughness, AFM images were acquired. Three-dimensional representations of surface topography can be found in supplementary materials Fig. S1. The Rq was a key parameter of choice for sample comparison (Fig. 4).

Figure 4
figure 4

Roughness of obtained TiO2 layers depending on the thickness.

Full size image

To improve the statistical importance, data from over 40 samples or spots were collected and included in further calculations. However despite of that, a large scattering of roughness among the tested samples still persists. For both precursors, a decreasing trend of roughness with increasing layer thickness is observed. This is attributed to the non-conformal growth of the layer, which leads to the filling of valleys and flattening of surface peaks, resulting in surface planarization. Slight differences in the samples from different precursors were observed. Namely, for the same thickness higher roughness was noticed of Tiacac, especially for thinner layers. Another attempt to determine the differences in the deposition on a rough substrate between layers from different precursors involved analysing the geometry of selected roughness peaks. The parameter of interest was the opening angle between the slopes of the roughness peak. If the layer grows conformally, meaning that the thickness increment is uniform across the entire surface, the opening angle of the peak will have lower values compared to non-conformal growth (where the layer covers thicker in lower-lying areas). The more conformal the deposition of the layers, the greater the thickness of the layer on the peaks for the same effective thickness. Indeed, the thickness of the electron transport layer (ETL) at the tips or peaks in planar structures (without a mesoporous layer) can be of particular importance. It is hypothesized that charge transport predominantly occurs through these tips or peaks (Fig. 5).

Figure 5
figure 5

Probable paths of electrons and extraction diffusion to ETL.

Full size image

To investigate the opening angle, over 25 cross sections of roughness peaks were analysed for each precursor using AFM imaging. The cross sections were obtained from samples with a small thickness of approximately 35 nm because it is at these thicknesses that the largest differences in Rq between precursors are observed. The average opening angle for TiO2 layers was found to be 127° from the Tiacac precursor and 132° from Tieth, while for FTO the opening angle was 114°. These results indicate a closer to conformal growth of layers from Tiacac, which may result in more even deposition. Methodology and calculations of average opening angles for TiO2 layers can be found in supplementary material Table S2, Figs. S2 and S3. In order to examine the difference in growth between Tiacac and Tieth, SEM images were taken in backscattered electrons based compositional mode and secondary electrons topographic mode in high magnification of 30,000× and 20000× (Fig. S6). There are no visible differences in compositional mode; however, in topographic mode, Tieth samples have smoother surfaces than Tiacac, which again may indicate more conformal grow of the Tiacac. In other studies, X. Li et al. also reported higher roughness and conformality for Tiacac precursor in comparison with the other tetrabutyl titanate [32]. However, considering their technology, it is hard to judge if they compared layers of the same thickness. In case of this article, this requirements was well-secured by initial concentration to thickness adjustments for each precursor. Additional information about the morphology of the samples can be obtained from the analysis of AFM images in phase mode.

Morphologically, the samples appear very similar in this imaging mode. The folds visible on the slopes of the roughness peaks are a result of probe oscillation occurring during rapid changes in the topography. This effect is related to the measurement method rather than the samples themselves (Fig. 6). However, on flat areas where no oscillations occur, “cracks” were observed (indicated by red circles) on the Tieth precursor samples, which were not observed for Tiacac. The occurrence of cracks intensifies with increasing layer thickness, suggesting that their formation may be related to the presence of stress in the TiO2 layer from the Tieth precursor.

Figure 6
figure 6

AFM images in phase mode left Tieth 64.6 nm right Tiacac 62.8 nm.

Full size image

Optics

The discussed TiO2 layers, which are used as layers of selective electron transport in cells with a n–i–p structure, are involved in the transmission of light to the absorber layer, i.e. perovskite; therefore, it is important to minimize optical losses caused by their introduction into the cell structure. Optical losses can occur by absorption of solar radiation in the layer, its reflection caused by the difference in refractive index between adjacent layers or thickness sensitive interference. Optical losses caused by the ETL layer might lower the photocurrent, as less light reaching the perovskite absorber. Using concentration of solutions and two different precursors, we can manipulate the parameters that contribute to optical losses and determine the most optimal configuration.

As assessed a detailed optical properties of the of the TiO2 layer, optical simulations were carried out in the Ansys Lumerical FDTD programme in the range of 300–800 nm on the surface of 16 µm2 in the light direction —> glass > SnO2 > SiO2 > FTO > TiO2 > PVK; due to the specificity of the simulations, the light source was placed in the glass layer, and the reflection at the air–glass interface was taken into account in the correction (Fig. 7).

Figure 7
figure 7

Refractive index diagram of simulated cells with scale bar.

Full size image

The aim of the simulation was to determine the optimal thickness at which the smallest optical losses due to light reflection would occur. The optical models of the TiO2 layers were used on the basis of ellipsometric measurements. Other models were based on literature data and our preliminary simulations. As part of the optical simulations, three variants of layer thicknesses were tested for both precursors, corresponding to the nominal thicknesses: 50 nm, 60 nm, and 70 nm; for each of them, the reflection and transmission at the ETL/perovskite boundary (TP) were recorded. In terms of effective transmission to the perovskite, an upward trend is clearly visible with the thickness of the TiO2 layers. Tiacac is slightly better than Tieth, achieving a higher TP_eff (range of integration 300–540 nm) - higher by 1 percentage point for each simulated thickness (Fig. 8a). In the case of the simulated effective reflection, cells with a layer of Tiacac precursor have a lower effective reflection of about 6.2% compared to 6.7% for Tieth. Surprisingly in simulations, the effective reflection is practically independent of the thickness of the simulated layer, and there were only slight differences of 0.1%. To confirm the results of optical simulations, measurements of transmission, scattered transmission and reflection were made, from which effective reflection and haze coefficient were then calculated. The UV–Vis measurement results confirm the simulation findings, indicating that the effective reflectance is not dependent on the thickness of the layers (Fig. 8b). Additionally, the Tiacac precursor’s layers exhibit lower reflectance compared to Tieth, with respective values of 6.90% and 7.15%. The results of effective transmittance do not show evident results from type of precursor nor thickness (Fig. 8c). The results of the simulations are consistent with the UV–Vis measurements and both methods indicate better optical properties, namely lower reflection with the layers deposited from the Tiacac precursor.

Figure 8
figure 8

a Dependence of simulated effective reflectance and transmission to perovskite on thickness of TiO2 layers in range of 300–540 nm. b Dependence of measured effective reflectance on thickness of TiO2 layers in range of 300–540 nm, c Transmission of obtained layers (including perovskite) in range of 300–540 nm.

Full size image

For FTO/TiO2 samples, determination of the haze coefficient, which is a measure of sample dispersion, gives additional information about the morphology of the sample, in particular about its roughness.

The haze coefficient (in range of 400–800 nm) of the TiO2 layers decreases with increasing thickness for both precursors, which can be related to the discussed roughness (Fig. 9). For thin layers of approximately 38.1 nm, the haze value is 55.73% for Tiacac, compared to 30.72% for Tieth at a thickness of 35.1 nm. However, for thicker layers of approximately 90 nm, the extrapolated straight line should indicate the similar reduction in roughness for both precursors. The decrease in the haze coefficient of samples with increased thickness coincides with the Rq values obtained from AFM. This correlation suggests that fogginess of the sample is determined by its surface roughness.

Figure 9
figure 9

Haze of obtained TiO2 layers depending on the thickness.

Full size image

Current–voltage characteristics

Both morphology and optical studies indicate the superiority of the Tiacac precursor over the Tieth precursor. Solar cells with layers made from Tiacac precursor should present a denser layer of perovskite without pinholes due to its better morphology. Additionally, based on optical results, a higher current density could be expected. To finally settle the issue of thickness and precursor, solar cells based on CsPbBr3 were made. The obtained cells were subjected to a test of current–voltage characteristics I–V. At least three cells were fabricated for each tested variation, with each cell consisting of four measurement points; this resulted in a total of 12 data points per variation (Table 2, Fig. 10). All tested samples with layers form Tiacac precursor have higher open-circuit voltage (Voc) and then sample (with the same thickness) from Tieth; for both precursors, Voc decreases with the thickness of the layer, and for the Tiacac precursor, the decline is lower. The highest Voc voltage was obtained by the Tiacac sample of 38.1 nm and was 1.337 V, and for the second precursor, the highest Voc value is 1.331 V for the similar thickness of 35.1 nm. Cells from Tiacac have generally better fill factor (FF). The dependence of FF on thickness is described by an asymmetric parabola. The highest fill factors for Tieth and Tiacac are 68% and 72%, respectively, for thicknesses of 53.64 nm and 62.75 nm.

Table 2 Averaged results of measurements of characteristic I–V
Full size table
Figure 10
figure 10

Averaged results of measurements of characteristic I–V.

Full size image

Cells with a TiO2 layer from the Tieth precursor achieve higher efficiencies, mainly due to high current density. The highest efficiency was achieved by Tieth 53.6 nm with parameters Jsc 6.820 mA/cm2, Voc 1.307 V, FF 68%, Eff. 6.08% (Fig. 11) I–V characteristics of the obtained cells. It was a photocurrent as the most decisive factor. The highest current density has been recorded for the Tieth sample 53.6 nm and is 6.820 mA/cm2 for the Tiacac precursor is 6.035 mA/cm2 for a thickness of 38.1 nm. The obtained results contradict the earlier findings regarding the morphology and optical properties of the layers, which indicated the superiority of the Tiacac precursor. In case of crystalline structure, the SEM images of perovskite layer (supplementary material Fig. S4a) looks very similarly for both type of samples; thus, the difference in solar cells cannot be explained by this factor. However, the poorer performance of the Tiacac samples could be attributed to issues with the growth of the perovskite layer. The advantage of cells with Tieth precursor layers is a higher Jsc (short-circuit current density), indicating a better extraction capability. Although in Tiacac cells, the generation of current to a tiny extent may be limited due to the presence of a significant number of holes in the perovskite layer (supplementary material Fig. S4b); however, concentration and size of holes in perovskite layer are not sufficient to bring down generated current density to this level as it was confirmed by similar transmittance as shown in Fig. 8c.

Figure 11
figure 11

AFM images in tapping mode left Tieth 64.6 nm right Tiacac 62.8 nm.

Full size image

Next, the electrochemical impedance spectra (EIS) were measured to know about the influence of ETL thickness and different precursor on the charge transport property in perovskite solar cells (PSCs). The influence of ETL modification in CsPbBr3 solar cells can quantitative analysed by a general equivalent circuit model of charges transport or recombination (Fig. 5A) [33, 34]. The series resistance (RS) and contact resistance (RSH) was mainly correlated to the FTO substrate/ETL interface contact and ETL/perovskite interface respectively. The recombination resistance (RREC) represents the charge recombination in PSCs [31]. The Nyquist plots of the devices measured under dark conditions and all the fitted parameters are shown in Fig. S7 and Table S4. The functions of ETL layer in this PSCs can be separated as (i) transport injected electrons to the FTO (electrode), and (ii) help to separate photogenerated holes and electrons from the perovskite layer by electron injection to the TiO2 layer [35,36,37]. The value of RS is increasing with the increase in thickness in case of both type of PSCs due to the increase in electronic path. However, we can see higher RS for Tiacac precursor. This is assigned to higher conformality of TiO2 from Tiacac, namely the layers are thicker over the roughness peaks, where we suppose that the actual electron transport occurs. In this perspective, a good ETL layers need to limits the recombination loss by successful separation of holes and electrons from the perovskite layer than only limits RS. The value of RREC for thinnest layers shows low values because the tin ETL do not provide sufficient hole blocking properties. For thicker ETL layers, RREC reaches a maximum value here for the thickness of ~ 50 to 60 nm for Tieth ETL and Tiacac ETL and then decreases with further increasing in the thickness of ETL (Table S4). Higher thickness of ETL layer decreases the recombination probability by increasing the electronic path between electrode and perovskite [38, 39]. Therefore, optimum balance between recombination and contact resistance at 53.6 nm thickness of Tieth ETL and 62.8 nm thickness of Tiacac ETL layer shows maximum efficiency of the PSC. It should be emphasized that a parabola of RREC as a function ETL thickness is much curved for Tieth and much broader than for Tiacac. However, on the other hand, the parameter which differs the most the two precursors is contact resistance RSH (Table S4). It is generally much lower for Tieth PSCs which indicates more efficient transport of electron from perovskite. Due to the lower value of RSH and lower level of recombination (higher RREC) in both the devices with optimised ETL thickness, the lower roughness of ETL helps to reduce the interfacial recombination and contact resistance between ETL and perovskite. Figure S7 shows the Nyquist plots of devices with 53.6 nm thickness of Tieth ETL and 62.8 nm of Tiacac ETL. Since the HTL layer remains constant for all PSCs, the variation in RSH between the two devices is expected to originate only from ETL/perovskite interface. As seen, the device with titanium precursor exhibits the lowest RCT and highest RREC values, which indicates the formation of better ETL/perovskite interface contact with the perovskite layer, and reduced charge recombination in PSCs with Tieth ETL due to the better morphology and reduced roughness of Tieth ETL as compare to Tiacac ETL layer [40, 41]

In order to further explore the possible origin of difference in contact resistance (RSH), therefore charge transport, the additional AFM measurements were conducted. The AFM images in tapping mode in submicrometric scale show that the crystals formed on the FTO surface are larger in the case of Tieth (Fig. 11). This observation is confirmed by additional XRD measurements by slightly broadened peaks from Tiacac (Fig. S8). As shown also for spatial ALD crystallinity of TiO2 is important aspects regardless of deposition method [42].

Also a nanoroughness from these TiO2 nanograins is much more developed for Tieth. The fine-crystalline structure resulting in the higher surface smoothness of the Tiacac precursor layers at submicrometer scale may cause issues with the wettability of the solid-phase perovskite at atomic scale. This could lead to shrinkage, i.e. reduction in its active and effective contact area and increased surface defect concentration, which could lead to problems with the extraction of charges. This finding directly corresponds to contact resistance (RSH) determined from EIS (Table S4). Similar dewetting phenomenon issued to obtain a so-called metal island film for instance of silver hemispherical nanoparticles from several nanometre film [43, 44]. The driving force was a poor wettability of dielectrics by metal film followed by the annealing at the temperature of a few hundred Celsius, which allow to the rearrangement of films into islands. Due to the issues with cesium solubility in CsPbBr3 perovskite, it is introduced multiple times as given in the experimental section and followed by annealing and diffusion. Therefore, this inorganic perovskite is being processed at higher temperature than most hybrid perovskites, which can trigger the proposed phenomenon. The dewetting process escalates with increasing annealing temperature, and higher temperatures lead to the formation of larger holes. The described relationship was observed in another experiment (supplementary material Fig. S5). More frequent occurrence of holes in perovskite layer in devices based on Tiacac could support this hypothesis of issues with wettability of solid state perovskite (supplementary material Fig. S4b). In case of mentioned thin metal film its thickness was a restraining factor, here the perovskite is a thick layer hence only limited effects can be noticed. Even though in might be enough to influence the operation of solar cells by lowering charge extraction efficiency, as given in this example. We propose that nanoroughness higher for Tieth can hinder dewetting process. For one-step deposition of perovskite, substrate wettability can be improved for example by washing with mixture of solvents as shown by Khalil and Ahmad [45]; however, it might be problematic for sequential deposition of CsPbBr3. Still, further research is needed to understand the role of ETL/ perovskite interface properties in all inorganic perovskite solar cells.

Conclusion

In this work, the solutions for the preparation of TiO2 layers based on two precursors titanium(IV) ethoxide and titanium(IV) bis(acetylacetonate) diisopropoxide were optimized. As part of the study, thickness, morphology and optical parameters (i.e. effective reflection, Haze coefficient, refractive index) of obtained layers were determined. The TEM analysis revealed a non-conformal growth of TiO2 layers deposited by spin coating. To examine the differences in the deposition of layers from different precursor solutions, AFM images were taken, which provided data about Rq and the opening slope angle of roughness peaks. Surface roughness decreases with thickness of TiO2 layer which prove non-conformal grow, although for Tiacac precursor, the decline was more linear, which with smaller opening slope angle of roughness peaks (closer to FTO) indicate more conformal deposition than Tieth. This result was confirmed by SEM imaging. Both optical and morphology parameters indicated the superiority of the Tiacac precursor over the Tieth precursor. However, cell-based on Tieth provides better efficiencies mainly because of generated current density. As additional AFM images at the submicron scale revealed greater nanoroughness layers from Tieth precursor, we formulated a hypothesis regarding poor wettability of inorganic perovskite on TiO2 layers from Tiacac caused by high smoothness of surface. Without kinetic barrier in the form of nanosurface roughness, the effective perovskite surface shrinks during perovskite processing and lead to problems with the extraction of electrons. The EIS equivalent circuit parameters corresponded with the materials and cell parameters, especially lower RSH and higher RREC for Tieth. Undoubtedly, the strong point of this article is that it raises the topic of ETL/perovskite interface problems, especially for CsPbBr3. Even though, the possible nanoroughness can mitigate the hindered charge extraction the cons of this work that the Jsc is still far from the theoretical one. Without mesoporous layer, morphology of planar TiO2/perovskite interface is critical and dominates other layer properties, e.g. optical. The further research should focus on additive engineering and surface passivation of ETL layer to reduce the surface energy which could potentially get rid of its weakness and improve cell efficiency. This approach is good candidate for commercial application because of lower production cost and higher cell stability in comparison to mesoporous structure with hybrid perovskite. The best photovoltaic parameters for titanium(IV) ethoxide are characterized for 53.6 nm thickness layer with the parameters: Jsc 6.820 mA/cm2, Voc 1.307 V, FF 68%, Eff. 6.08%. For titanium(IV) bis(acetylacetonate) diisopropoxide, best photovoltaic parameters was characterized for 62.8 nm thickness layer with the parameters: Jsc 5.866 mA/cm2, Voc 1.311 V, FF 72%, Eff. 5.62%.