1 Introduction

Planar metal halide perovskite solar cells (PSCs) have been pushing the record-breaking power conversion efficiencies (PCEs) over 26%, as well as durable stability and compatibility with large-scale manufacture [1,2,3]. The reasons behind such significant achievements are associated with the strategies of integrating low-bandgap and less-trap-state formamidine-based perovskite with widely regarded as efficient charge transport layers [4]. In particular, regular planar PSCs indispensably necessitate high-quality and compatible electron transport layer (ETL) owing to their high light transmittance, suitable energy level, and low-temperature processability, which are crucial to not only the growth of the top perovskite grain but also extraction and collection of the photogenerated electrons to the electrode [5]. While noting that widely adopted metal oxide ETLs are inevitably endowed with inherently inferior electron mobility (for example, TiO2: usually at the level of 10–5 ~ 10–4 cm2 V−1 s−1) due to low-temperature process that could arouse charge carrier accumulation and recombination loss at buried interface, resulting in less-than-ideal efficiency and unrobust environmental stability [6]. Several state-of-the-art engineering strategies have been developed to address the imperfections of such ETLs, as exemplified by employing nitrogen-doped TiO2 ETLs with reduced the electrical series resistance, as well as high-lattice-matching SrSnO3 as the ETLs enabling ordered beginning of the growth of perovskite to synchronously rule the buried defects and carrier dynamics in PSCs [7, 8].

Creation of p–n junction either in the active layer or at upper/buried interface has been exploited as an effective strategy to tune the charge carrier transport in PSCs [9, 10]. For example, embedding 2D graphdiyne or 0D fluorinated-gold-cluster at grain boundaries within the perovskite films enables the construction of p-n heterojunctions, which provides an extra channel to favor the exciton separation and charge transport [11, 12]. It is reported that the TiO2 with Ti vacancies shows inherent p-type conductivity with high charge mobility, demonstrating a nearly sevenfold increase over the normal n-type TiO2 [13]. Motivated by embedding p–n junction in perovskite to alleviate the carrier loss, it would thus be highly promising to construct the p–n homojunction in the ETLs by introducing Ti-defected TiO2 in n-type TiO2 for further pronounced electron conducting capability and highly efficient and stable TiO2-based planar PSCs. The challenge however remains on not only the technical embedding of such p-type TiO2 in n-type TiO2 ETLs, but also the understanding of such homojunction influencing the carriers transport in the ETL.

Present work demonstrates an effective strategy of constructing Ti0.936O2@TiO2-based p–n homojunction to improve electron mobility and photovoltaic performance of planar PSCs through embedding laser-derived p-type Ti0.936O2 in TiO2 ETL. Such embedding of Ti-defected TiO2 could modulate the crystallization kinetics of the TiO2 matrix by restraining the rutile phase that is detrimental to light stability of PSCs, contributing to the formation of high-quality TiO2 ETLs. The formed p-n homojunction enables also not only elimination of the interfacial lattice distortion between Ti0.936O2 and TiO2, but also more efficient transport of charge carriers at both surfaces and boundaries of TiO2 ETLs through localized build-in electric fields, thus reducing the recombination loss (Scheme 1). More importantly, such a novel Ti0.936O2@TiO2 composite ETL has also exerted a significant influence on the construction of less-trap-states and larger-grain perovskite films. Benefiting from these merits, we obtain highly efficient formamidinium lead iodide (FAPbI3) PSCs with PCE up to 25.50%, which ranks as far as we know among the top in records of TiO2-based planar PSCs. Owing to the synchronous regulation of the Ti0.936O2 in photocatalytic activity of TiO2 ETLs and film quality of perovskite layers, we have also obtained highly stable FAPbI3 PSC that maintains over 95% of their initial efficiency at maximum power point under continuous illumination for 500 h, as well as mixed-cation PSCs with pronounced environmental stability over 3000 h under RH of 40%. We believe this study provides an efficient alternative of improving the carrier conducting capability for charge transport layers and their optoelectronic devices, from the viewpoint of p–n homojunction engineering.

Scheme 1
scheme 1

Schematic illustration of the effects of embedding coherent p–n homojunctions on carrier dynamics at buried interface compared with those of conventional works [7, 8, 16]

2 Experimental Section

2.1 Materials

Unless stated otherwise, all materials were purchased from Sigma-Aldrich without further purification. Fluorine-doped tin oxide (FTO) coated glass substrates (around 1.5 cm × 1.5 cm) with partial etching were purchased from OPV•Tech. Spiro-OMeTAD (2,2′,7,7′-tetrakis[N,N-di(4-methoxyphenyl)amino]-9,9′-spirobifuorene, ≥ 99.8% purity), 4-tert-butylpyridine (tBP, ≥ 99.9% purity), and lithium-bis (trifluoromethanesulfonyl) imide (Li-TFSI, ≥ 99.9% purity) were purchased from Xi’an Polymer Light Technology Corp.

2.2 Device Fabrication

2.2.1 Preparation of Ligand-Free Ti 0.936 O 2 Nanocrystals in Desired Solvents

The initial anatase Ti0.936O2 powder was synthesized using solvothermal treatment of tetrabutyl titanate in an ethanol-glycerol mixture and then thermal calcination according to precious studies [13]. A certain amount of Ti0.936O2 powder (2 mg) was then transferred into a vessel with 10 mL deionized water (0.2 mg mL−1), subsequently subjected to a non-focusing nanosecond pulsed laser irradiation (Quantel, repetition rate: 10 Hz, pulse width: 8 ns) with a wavelength of 355 nm, along with continuous ultrasonic treatment to obtain the colloid concentration of 0.2 mg mL−1. Subsequently, the colloids with 0.2 mg mL−1 concentration were diluted to 0.1 and 0.05 mg mL−1, respectively, which further suffer from laser irradiation to keep the colloids homogeneous. The tailored laser fluence ranges from 0 to 300 mJ pulse−1 cm−2 and the concentration of raw Ti0.936O2 powder ranges from 0.05 to 0.2 mg mL−1 at fixed irradiation time of 3 min and cooling temperature of − 20 °C.

2.2.2 Preparation of Compact ETLs

The etched FTO substrates were washed successively in detergent, deionized water, acetone and ethanol under continuous sonication, then dried with N2 flow employing a compressed nitrogen gun, and then treated under oxygen plasma for 10 min to remove organic residues on FTO substrates.

For the preparation of target ETL: an approximately 50 nm thick TiO2 compact layer deposited on clean FTO substrate is prepared adopting in situ chemical bath co-deposition by adding different volume of Ti0.936O2 colloid solutions (3%, 6%, and 9% volume ratio to TiCl4 precursor abbreviated as 3%-Target TiO2, 6%-Target TiO2, 9%-Target TiO2) to TiCl4 aqueous solution (2.25: 100 volume ratio of TiCl4: H2O), along with FTO substrates suffering from thermal treatment at 70 °C for 1 h, and then anneal at 150 °C for 1 h.

For the preparation of pristine ETL: identical TiCl4 aqueous solution without adding Ti0.936O2 colloid solutions, along with FTO substrates are subjected to similar thermal processing mentioned above.

2.2.3 Fabrication of Perovskite Solar Cells

CsFAMA type perovskite: CsI (0.0625 M), FAI (1.0125 M), PbI2 (1.075 M), MABr (0.175 M) and PbBr2 (0.175 M) is dissolved in a mixture of DMF: DMSO (4:1 v/v) with a successive stir at 55 °C for 2 h to prepare precursor solution at a concentration of 1.25 M. The spin-coating process was performed in a nitrogen glove box. The as-prepared precursor (30 μL) was dropped onto the TiO2/FTO substrate followed by a consecutive two-step spin-coating process at 2000 and 4000 rpm for 10 and 30 s, respectively. During the second spin-coating step, 200 μL anhydrous chlorobenzene (CB) was immediately poured on the substrate 10 s prior to the end of the program. Subsequently, the intermediate phase film is heated on a hotplate at 100 °C for 1 h.

FAPbI3 type perovskite: FAI (1.80 M), PbI2 (1.80 M), FAHCOO (0.12 M), and MACl (0.52 M) are dissolved in a mixture of DMF: DMSO (volume ratio = 4:1) at 55 °C for 2 h to prepare the perovskite precursor (1.80 M). The as-prepared precursor (30 μL) was dropped onto the TiO2/FTO substrate followed by a one-step spin-coating at 6500 rpm for 60 s. During spin coating, 200 μL chlorobenzene (CB) is immediately poured on the substrate 20 s prior to the end of the program. Subsequently, the intermediate phase film is heated on a hotplate at 150 °C for 10 min.

In addition, for FAPbI3 type perovskite, 3 mg ml−1 phenethylammonium iodide (PEAI) was deposited on the perovskite/TiO2/FTO substrate via a one-step spin-coating process at 4000 rpm for 30 s, followed by deposition of Spiro-OMeTAD.

Spiro-OMeTAD solution was prepared by dissolving 72.3 mg Spiro-OMeTAD, 29 μL 4-tert-butylpyridine (tBP) and 18 μL lithium-bis (trifluoromethanesulfonyl) imide (Li-TFSI, a stock solution of 520 mg mL−1 in acetonitrile) into 1 mL chlorobenzene. And then 30 μL solution was spin-coated on the perovskite/TiO2/FTO substrate at 6000 rpm for 30 s. Finally, the Au electrode (80 nm) was deposited on the top of devices by thermal evaporation using a shadow mask. Each electrode of devices exhibits the area of 0.05 cm2, for which the effective area would be corrected by optical microscopy.

2.3 Characterization

The scanning electron microscopy (SEM) images were obtained using a field emission SEM (FEI Nova). Atomic force microscope (AFM) was carried out using a Bruker Dimension Icon. High-resolution transmission electron microscopy (HRTEM) was conducted employing an FEI Tecnai F30 transmission electron microscope at 300 kV, equipped with an Oxford Instruments EDS detector and a high angle annular dark field (HAADF) STEM detector. The Raman spectra were recorded by a Raman microscope at an excitation laser wavelength of 532 nm (Renishaw). The X-ray diffraction (XRD) patterns were recorded on a X’pert PRO (PANalytical) adopting a Cu Kα (λ = 0.15406 nm) as the X-ray source. The absorption was characterized by the ultraviolet–visible (UV–vis) spectrophotometer (Perkin-Elmer Lambda 35 UV–vis-NIR). The steady-state photoluminescence (PL) and time-resolved photoluminescence (TRPL) spectra were recorded by a pulse laser excitation source at the wavelength of 470 nm (Horiba FluorologFL-3). The electrical impedance spectroscopy (EIS) was characterized applying a bias of 0.8 V in the dark in a frequency range from 1 MHz to 0.1 Hz (CHI660E). For Mott-Schottky analysis, capacitance–voltage measurements were performed at a frequency of 1 kHz (CHI660E). X-ray photoelectron spectroscopy (XPS) measurements were conducted on an Axis Supra (Kratos). Ultraviolet photoelectron spectroscopy (UPS) was characterized by a VG Scienta R4000 analyzer and the HeI (21.22 eV) emission line employed for excitation at a bias of − 5 V. The contact angles measurements were conducted by a data physics OCA-20 contact-angle system at ambient air. Temperature dependent admittance spectroscopy (TAS) was performed on a precision impedance analyzer at various temperatures (T = 210–320 K) in the dark. A Keithley 2400 source meter was used to record the JV curves and maximum power point tracking under simulated AM 1.5G illumination (100 mW cm−2) produced by a xenon-lamp-based solar simulator (Oriel 67005, 150 W Solar Simulator), which was calibrated with a monocrystalline silicon reference cell (Hamamatsu S1133). The devices were measured both in reverse scan (+ 1.2–− 0.1 V) and forward scan (− 0.1– + 1.2 V) with a scanning rate of 0.2 V s−1. The EQE was conducted by employing a Enlitech EQE measurement system (QE-R3011). A Keithley 2400 source was used to measure the dark IV characterization of the electron-only devices for calculating the defect density using SCLC model.

2.4 Statistical Analysis

All quantitative values are shown as means ± standard deviation. All quantitative experiments were carried out using at least three replicates for each group. The statistical analysis was conducted by the t test, and a p value of less than 0.05 was considered as statistical significance. The error bars correspond to the standard deviation of data points from individual samples.

3 Results and Discussion

3.1 Laser Embedding of p-Type Ti0.936O2 Nanocrystals in TiO2 ETLs

Laser manufacture of size-tailored Ti0.936O2 nanocrystals by irradiation of their raw sub-micrometer counterpart in liquid and subsequent embedding in the TiO2 matrix through the chemical bath deposition method are shown in Figs. S1 and S2. Subsequent to the optimization of laser fluence and concentration of laser process (Fig. S3), the transparent Ti0.936O2 colloid solution with clear Tyndall scattering is obtained with well-dispersed nanocrystals with an average diameter of 3.5 nm (Fig. 1a). The crystal structure of the as-prepared Ti0.936O2 nanocrystals, which was determined by HRTEM and corresponding Fast Fourier transform (FFT), represents the lattice spacing of 0.24 nm that corresponds to the typical plane (004) of Ti0.936O2 (Fig. 1b), further confirmed by their identical Raman spectroscopy (Fig. S4). These results indicate that Ti0.936O2 nanocrystals well inherit the properties of their bulk counterpart. In addition, the elements mapping extracted from TEM-energy-dispersive spectroscopy (TEM-EDS) suggests homogeneous distribution of all elements throughout the entire Ti0.936O2 nanoparticles without any segregation (Fig. S5). Furthermore, the XPS demonstrate stable surface composition and chemical state of the Ti0.936O2 during laser irradiation (Fig. S6) [13]. This is consistent with the recent work that reflects unchanged Ti-vacancy of laser-processed Ti0.936O2 characterized by electron paramagnetic resonance (EPR) spectroscopy [14].

Fig. 1
figure 1

a TEM image of Ti0.936O2 nanocrystals (inset: nanocrystals size distribution diagram and Mie-scattering image of colloids). b HRTEM and corresponding FFT of Ti0.936O2 nanocrystals. c SEM images (inset: AFM images) of pristine TiO2 and 6%-target TiO2 films. d XRD patterns of different ETLs. e Dark IV measurement of the electron-only devices (inset) displaying VTFL kink point related to the trap density. f Electron mobilities of different ETLs using the SCLC model, the inset shows the device structure of ITO/Al/ETLs/Al. g Mott-Schottky plots of Ti0.936O2@TiO2 matrix. h UPS results of Ti0.936O2 nanocrystals (inset: band gap and energy level of Ti0.936O2 nanocrystals). i Different TiO2 films with UPS Fermi edge (left) and the cut-off energy (right). Scale bar: a 50 nm; b 1 nm; c 1 μm (inset: 1 μm)

Subsequently, the deposition of Ti0.936O2@TiO2 ETL was conducted adopting a facile one-step chemical bath co-deposition, where laser-generated sub-5 nm Ti0.936O2 nanocrystals could be in situ embedded in the TiO2 matrix (Fig. S7). In brief, different contents (3%, 6%, and 9% volume ratio to TiCl4 precursor denoted as 3%-target TiO2, 6%-target TiO2, 9%-target TiO2) of Ti0.936O2 colloids with concentration of 0.1 mg mL−1 were incorporated into TiCl4 solution to fabricate the TiO2 composite ETLs. Scanning electron microscopy (SEM) and atomic force microscopy (AFM) were performed to evaluate the surface morphologies of corresponding TiO2 films. The results reveal that the 6%-target TiO2 film exhibits smoother and flatter surface with reduced roughness from 33.8 to 21.5 nm in comparison with the pristine film, as shown in Figs. 1c and S8. In order to further explore the influence of embedding Ti0.936O2 on crystallization kinetics of TiO2 ETLs, the settled TiO2 powders after chemical bath co-deposition were collected for systematic XRD analysis. As shown in Fig. 1d, the initial TiO2 ETL consists of polymorphic phases of rutile and anatase, while the rutile phase in the ETLs is significantly reduced after the embedding of Ti0.936O2. It should be noted that the fast nucleation rate is more inclined to form the anatase phase at the initial stage of TiCl4 hydrolysis, whereas the slow nucleation rate is conducive to the directional arrangement of aggregates, resulting in the generation of a more stable rutile phase [15]. It could thus be deduced that the addition of Ti0.936O2 favors the rapid nucleation, which enables not only the formation of smaller TiO2 grains for their compact deposition, but the inhibition of rutile phase and anatase–rutile hetero-phase junction for reducing the unwanted photocatalytic degradation of PSCs under continuous light soaking [16, 17].

In order to investigate the effects of the Ti0.936O2 nanocrystals on the optical and electronic properties of the TiO2 ETLs, the optical bandgap of Ti0.936O2@TiO2 films were first evaluated by ultraviolet–visible (UV–vis) absorption spectra and corresponding Tauc plots. The results show that the embedding of Ti0.936O2 enables slight increase in the bandgap of the TiO2 ETLs from 3.20 to 3.22 eV, with insignificant change on the optical transmittance, as shown in Figs. S9 and S10. To evaluate the electronic properties, the defect density (Nt) and the electron mobility (µ) of the TiO2 ETLs were successively examined by the space charge-limited current (SCLC) method. The result shows that the embedding of Ti0.936O2 results in the reduction of Nt from initial 6.48 × 1016–1.39 × 1016 cm−3 (Fig. 1e and Table S1), which may be due to the improved crystallization kinetics of TiO2 and the high-quality ETLs. Moreover, the µ is found to be boosted by two orders of magnitude from pristine 8.63 × 10–5–4.67 × 10–3 cm2 V−1 s−1 for the 6%-target TiO2 ETLs (Fig. 1f), which is consistent with the conductivity (σ) result that indicates higher σ for 6%-target TiO2 due to a large slope (Fig. S11). The improved electronic properties are mainly attributed to the construction of Ti0.936O2@TiO2 p–n homojunction, which is evidenced by an inverted “V-shape” with typical p–n junction feature from the Mott-Schottky plot shown in Fig. 1g [18]. To confirm the p-type characteristic of the Ti0.936O2, the ultraviolet photoelectron spectroscopy (UPS, Fig. 1h) was used to check its electronic structure of the Ti0.936O2. Based on the optical bandgap (3.25 eV) (inset in Fig. 1h), the corresponding Fermi energy level, the conduction band energy level and the valence band energy level are calculated to be − 4.49, − 2.29, and − 5.54 eV, respectively, which identifies the p-type semiconductor feature of the Ti0.936O2 nanocrystals. Such p–n construction greatly accelerates the carrier transport at both the surfaces and the boundaries of TiO2 particles to restrain carrier loss owing to the increase of the depletion width [18, 19]. It is also found that the embedded p-n homojunction is helpful to improve electronic structure of TiO2 ETLs with upward-shifted energy level, enabling a better energy level alignment with top perovskite active layer to lower the interfacial electron barrier (Figs. 1i and S12) [9]. Detailly, the UPS characterization of laser-processed Ti0.936O2 nanocrystals strongly confirms their p-type semiconductor characteristic (Fig. 1h), which were embedded into the n-type TiO2 matrix to form a p–n junction by generating a uniform Fermi level (Fig. S12b). It is worth noting that the formed p–n homojunctions between Ti0.936O2 and TiO2 could create numerous localized built-in electric fields with a direction from n-type TiO2 to p-type Ti0.936O2, as shown in Scheme 1 and Fig. S12b, which enables not only the effective promotion of carrier transport at both the surfaces and boundaries of TiO2 matrix due to the expansion of the depletion width [19], but also the oriented transport of photo-generated charge carriers, which favors for the boosted electron mobility (Scheme 1) [10].

3.2 Effect of Ti0.936O2@TiO2 on the Top Perovskite Films

The surface morphologies of the mixed-cation perovskite (Cs0.05(FA0.85MA0.15)0.95PbI2.55Br0.45, CsFAMA) grown on different TiO2 ETLs were intentionally evaluated by the SEM and AFM characterizations. The results show obviously larger-grain size and smoother surface of target perovskite films compared with those of pristine ones (Figs. 2a, b and S13). Such surface morphology can trigger the reduction of Gibbs free energy for heterogeneous nucleation of perovskite precursor due to the sharply decreased contact angle from pristine 13°–4°, for Ti0.936O2@TiO2 ETLs (see insets in Fig. 2a, b), thus contributing to the high-quality and large-grain perovskite films [20]. Cross-sectional SEM and XRD patterns demonstrate the large grains throughout the entire thickness of the perovskite film deposited on Ti0.936O2@TiO2 ETLs (Figs. S14 and S15), owing probably to the enhanced crystallization kinetics during their grain growth [6]. Furthermore, the steady-state and time-resolved PL spectra were employed to investigate the effects of Ti0.936O2 nanocrystals on the carrier dynamics between the perovskite layer and the ETLs. As shown in Fig. 2c, the CsFAMA perovskite films based on the 6%-target TiO2 present a more prominent PL quenching with a twofold decrease of the PL intensity compared with the control films, demonstrating more efficient electron transfer between the perovskite layer and the ETLs. Similarly, time-resolved PL (TRPL) results (Fig. 2d, Table S2) exhibit that the average carrier lifetimes are calculated to be 130.63–37.38 ns for the control and the target, respectively, indicating the significant reduction of the carrier lifetime and thereby the enhanced electron extraction at the buried interface [16].

Fig. 2
figure 2

SEM top-view images of perovskite films based on a pristine TiO2 and b 6%-target TiO2 films (inset: contact angles of different ETLs dropped by perovskite precursor). Steady-state c and time-resolved d PL spectra of CsFAMA perovskite films spin-coated on different TiO2 layers. e Arrhenius plots of the characteristic transition frequencies. f Trap state density (NT) of the perovskite photovoltaics measured at 300 K. Dependence of g Jsc and h VOC on the irradiation intensity of the devices based on different TiO2 ETLs. i Nyquist plots of the devices based on different ETLs measured in the dark at a bias of 0.8 V. Scale bar: a 500 nm; b 500 nm

In order to further explore the impact of Ti0.936O2@TiO2 ETLs on modulating defect states in top perovskite films, temperature dependent admittance spectroscopy (TAS) was accordingly employed to quantitatively estimate both the energy level and the distribution of trap states (Note S1, Fig. S16) [21]. As shown in Fig. 2e, the defect activation energies (Ea) of different films are extracted from the Arrhenius plots of the characteristic transition frequencies obtained from the corresponding capacitance-frequency curves under different temperatures (Fig. S16a, b), and are calculated to be 0.277 and 0.223 eV for the control and target films respectively. Figure 2f exhibits the energy level and the density of trap states of different perovskite films, demonstrating that Ti0.936O2@TiO2 ETLs effectively reduce the energy level of trap states from pristine 0.22–0.18 eV, as well as their density of states from pristine 3.52 × 1016–2.03 × 1016 cm−3. Subsequently, the short-circuit current density (Jsc) and the open-circuit voltage (Voc) at variable light intensities were measured to gain in-depth understanding of the carrier recombination kinetics in perovskite films. As shown in Fig. 2g, the curves of dependence of Jsc on the irradiation intensity represent similar slopes close to 1, revealing negligible bimolecular recombination within all films [22]. Figure 2h depicts Voc versus light intensity in which the fitted slopes significantly decrease from pristine 1.501.05kT/e for target films, indicating the effectively suppressive trap-assisted recombination that facilitates leakage current (Fig. S17) [22], which is also in good agreement with that of the TAS in Fig. 2f. The EIS was used to reflect interfacial charge transfer capability between the perovskite layer and ETLs. As shown in Fig. 2i, the contact resistance (Rco) decreases from pristine 16,544–12,300 Ω and the recombination resistance (Rrec) increases from pristine 1.05 × 105–1.64 × 105 Ω for the 6%-target TiO2 (Table S3). These results indicate that embedding of the Ti0.936O2 in TiO2 matrix results in effectively improved charge transport and suppressed charge recombination at the buried interface [16, 23, 24].

3.3 Photovoltaic Performance and Stability of Planar PSCs Based on Ti0.936O2@TiO2 ETLs

The regular planar PSCs with the configuration of FTO substrate/Ti0.936O2@TiO2/perovskite/Spiro-OMeTAD/Au were fabricated to further evaluate the effect of the embedding of Ti0.936O2 on photovoltaic performance (See schematic illustration of fabrication process in Fig. S18). Figure 3a shows the current density–voltage (JV) curves of different CsFAMA champion devices measured under illumination of 100 mW cm−2 (AM 1.5G) and the corresponding photovoltaic parameters are listed in Table 1. The device based on 6%-target TiO2 ETLs exhibits a PCE of 22.02% with a Voc of 1.194 V, a Jsc of 23.72 mA cm−2, and a fill factor (FF) of 77.75%, higher than those of devices based on pristine TiO2 (19.94%), 3%-target TiO2 (21.42%), and 9%-target TiO2 (20.66%). In addition, negligible hysteresis is shown in target champion devices with a stabilized power output of 21.90% close to maximum efficiency (Fig. 3b), evidenced by reduction of hysteresis factor from pristine 13.1%–1.5% for the 6%-target TiO2-based devices (Table S4). To verify the universality of Ti0.936O2@TiO2 ETLs in enhancing carrier dynamics at buried interface, α-phase formamidinium lead iodide (FAPbI3)-based devices were further constructed to pursue higher photovoltaic performance. It is worth noting that Ti0.936O2@TiO2 ETLs could encouragingly initiate the fabrication of highly crystalline FAPbI3 perovskite with optimized morphology, and enable enhanced carrier dynamics at the buried interface (Figs. S19, S20, S21. S22, S23, Table S5). Owing to these merits, the target FAPbI3 PSC upon a bandgap of 1.53 eV (Fig. S24) achieved the champion PCE up to 25.50% with a Voc of 1.185 V, a Jsc of 25.79 mA cm−2, and a FF of 83.45%, and highly stabilized PCE of 25.42%, far exceeding that of the control (22.81%), as shown in Fig. 3c and Table 1. Figure 3d further exhibits external quantum efficiency (EQE) spectra, in which enhanced spectral response of target devices in entire range is due to the construction of high-quality and large-grain perovskite films deposited on Ti0.936O2@TiO2 ETLs [25], enabling an increment of integrated Jsc from 24.27 to 25.43 mA cm−2, matching well with the JV results. As shown in Fig. 3e, efficiency distribution histogram of 50 individual PSCs indicates the improved reproducibility of the target devices, demonstrated by enhanced average PCEs from 18.28% to 21.22% and from 21.24% to 23.52% that successively corresponds to CsFAMA and FAPbI3 PSCs, along with synchronous improvement of Voc, Jsc and FF (Figs. S25 and S26). It is worth noting that champion efficiency of 25.50% in present work ranks among the top in records of PSCs based on TiO2 ETLs (Fig. 3f, Table S6). Such significant enhancement in efficiency is partially attributed to the optimized interface band-alignment induced by the embedding of Ti0.936O2 (Fig. 3g), thereby favoring the less charge accumulation at the interface between perovskite and ETL and the increment of voltage output, evidenced by the UPS analyses and capacitance–voltage measurements (Figs. S16c and S27) [26,27,28].

Fig. 3
figure 3

a J-V curves of CsFAMA devices with different TiO2 layers (inset: schematic illustration of device structure). b JV plots of CsFAMA champion devices containing the pristine TiO2 and 6%-target TiO2 layers measured both in reverse scan and forward scan, the insets show stabilized power output at MPP tracking. c JV curves of FAPbI3 devices employing different TiO2 layers with stabilized power output at MPP tracking. d EQE spectra of FAPbI3 champion devices upon the pristine TiO2 and 6%-target TiO2 layers, respectively. e PCE distribution of 50 individual CsFAMA and FAPbI3 devices. f Comparison of efficiency of PSCs employing TiO2 as ETLs. g Energy level diagram for each component of devices upon different TiO2 layers. The energy level structures of Spiro-OMeTAD and Au refer to the literature [23, 24]

Table 1 Photovoltaic parameters of the CsFAMA and FAPbI3 type PSCs upon different TiO2 ETLs

Adopting the p–n homojunction embedding strategy has been demonstrated to be greatly effective for the construction of high-performance device, while its influence on the long-term stability of devices would be further investigated. Figure 4a shows the humidity stability of different CsFAMA PSCs without encapsulation stored in ambient air with relative humidity (RH) of 40% in the dark. The result indicates that the target devices exhibit superior humidity stability, maintaining 85% of initial PCE for 3300 h in comparison with that of control devices (approximately 50% for 700 h). The operational stability of target devices also shows great improvement, retaining 93% of initial PCE over 170 h in contrast with that of control devices (approximately 21% for 35 h), which was measured using maximum power point (MPP) tracking under full-sun illumination in ambient air with RH of 55 ± 5%, as well as excellent thermal stability (Figs. 4b and S28). We further check the environmental stability of various FAPbI3-based devices, due to the ease with which FAPbI3 perovskite could arouse its spontaneous phase transition from α- to δ-FAPbI3 under ambient conditions [29]. For the humidity stability of FAPbI3 PSCs (RH of 40%, Fig. 4c), the control devices continuously degrade by more than 60% of their initial PCE for 1000 h, whereas the target devices could retain 73% of their initial value at the same time. For the operational stability of FAPbI3-based devices under MPP tracking at 60 °C under full-sun illumination in inner atmosphere, the enhanced operational stability in target FAPbI3 device is demonstrated by less than 5% degradation of its initial PCE over 500 h, compared with that of control devices (over 50% after 200 h, Fig. 4d). The PL characterization was further carried out to check the stability of perovskite films under continuous UV irradiation. As shown in Fig. 4e, f, there is a red shift of about 3 nm in the PL peak of pristine perovskite film under UV irradiation of 20 h, while target perovskite films on Ti0.936O2@TiO2 ETLs exhibit a negligible shift, demonstrating suppressive decomposition of perovskite film due to significantly reduced anatase–rutile phase junctions that trigger the photocatalytic property of TiO2 [30], which accounts for improvement of light-illumination stability of target device.

Fig. 4
figure 4

a Humidity stability and b operational stability of different CsFAMA-based devices. c Humidity stability and d operational stability of different FAPbI3-based devices. Normalized PL spectra of perovskite films grown on e pristine TiO2 and f 6%-target TiO2 layers under continuous 254 nm UV irradiation (50 mW cm−2) for 20 h. The error bars represent the standard deviation for 20 devices

It is encouragingly found that the Ti0.936O2@TiO2 ETLs play an important role in the construction of highly efficient and stable PSCs through multiple pathways: (i) the formation of the p–n homojunctions between Ti0.936O2 and TiO2 could not only accelerate the electron transport at both the surfaces and the boundaries of TiO2 matrix due to the increase in depletion width [19], but also significantly enhance crystal quality of the ETLs, resulting in enhanced conduction capability and boosted electron mobility by two orders of magnitude for TiO2 ETLs; (ii) such p–n homojunction enables upward-shifted Fermi level of TiO2 that favors better energy level alignment with the perovskite, leading thus to reduction in voltage loss and promotion of electron extraction at the buried interface [31]; (iii) embedding Ti0.936O2 could greatly improve not only crystallization process of TiO2 for the inhibition of rutile phase that results in light-induced instability (Fig. S29), but surface wettability of TiO2 that initiates rapid nucleation of top precursor for the fabrication of highly crystalline and large-grain perovskite films, which effectively prevent moisture from penetrating at grain boundaries to retard degradation of perovskite and enhance humidity stability of PSCs. Owing to these merits, highly efficient FAPbI3 PSCs delivered champion PCE up to 25.50%, which ranks among the top in records of TiO2-based planar PSCs, as well as prominent moisture (RH of 40% for 3300 h) and light (under MPP for 500 h) stability of unencapsulated PSCs were achieved. It could be inferred from the improvement of all photovoltaic parameters that better energy level alignment, accelerated electron transport and reduced contact impedance at the buried interface principally account for the increment of Voc, Jsc, and FF, respectively [11, 23, 24]. In addition, the significantly eliminated hysteresis phenomenon of the target device is attributed to effective suppression of the carrier transport imbalance within the device, due to the enhanced electron mobility of the ETLs by embedding Ti0.936O2 (Figs. 3b and S30, Tables S4 and S7) [32].

4 Conclusions

In summary, present work has demonstrated an efficient strategy of laser embedding of the p–n homojunctions in TiO2 ETLs to address the issues of insufficient carrier transport at the buried interface for highly efficient and stable PSCs. The embedded p–n homojunction between Ti0.936O2 and TiO2 not only greatly assisted synchronous acceleration of carrier transport at both surfaces and boundaries of TiO2, which principally accounts for boosted electron mobility by two orders of magnitude, but modulates interfacial energy level to rapidly extract electrons that results in reduction of voltage deficit. The laser generated p-type Ti0.936O2 also exerts significant influence on the crystallization kinetics of the TiO2 matrix and top precursor, which significantly inhibits the rutile phase that leads to light instability and initiates the fabrication of large-grain perovskite that enhances humidity stability of PSCs, respectively. The capability of such novel Ti0.936O2@TiO2 ETLs is thus demonstrated by accessing efficient and stable planar FAPbI3 PSCs with a champion efficiency of 25.50% and robust light-induced stability over 500 h, as well as mixed-cation PSCs with a champion efficiency of 22.02% and pronounced humidity stability for 3300 h (RH of 40%). This study exploits a novel pathway of developing highly conductive charge transport layers for state-of-the-art planar PSCs.