22% Efficiency Inverted Perovskite Photovoltaic Cell Using Cation‐Doped Brookite TiO₂ Top Buffer

Abstract

Simultaneously achieving high efficiency and high durability in perovskite solar cells is a critical step toward the commercialization of this technology. Inverted perovskite photovoltaic (IP‐PV) cells incorporating robust and low levelized‐cost‐of‐energy (LCOE) buffer layers are supposed to be a promising solution to this target. However, insufficient inventory of materials for back‐electrode buffers substantially limits the development of IP‐PV. Herein, a composite consisting of 1D cation‐doped TiO₂ brookite nanorod (NR) embedded by 0D fullerene is investigated as a top modification buffer for IP‐PV. The cathode buffer is constructed by introducing fullerene to fill the interstitial space of the TiO₂ NR matrix. Meanwhile, cations of transition metal Co or Fe are doped into the TiO₂ NR to further tune the electronic property. Such a top buffer exhibits multifold advantages, including improved film uniformity, enhanced electron extraction and transfer ability, better energy level matching with perovskite, and stronger moisture resistance. Correspondingly, the resultant IP‐PV displays an efficiency exceeding 22% with a 22‐fold prolonged working lifetime. The strategy not only provides an essential addition to the material inventory for top electron buffers by introducing the 0D:1D composite concept, but also opens a new avenue to optimize perovskite PVs with desirable properties.

A composite consisting of 1D cation‐doped TiO₂ brookite nanorod embedded by 0D fullerene is investigated as a top modification buffer for inverted perovskite photovoltaic (IP‐PV) cells. The resultant IP‐PV displays an efficiency exceeding 22% with a favorable stability. This work opens up more opportunities in expanding the material inventory for charge transfer layer in perovskite solar cells development and application.

Organic–inorganic halide perovskites have captured growing attention in various energy‐harvesting sectors by showing their primitive success in achieving highly efficient photovoltaic cells.^{[ 1 ]} Extensive researches in the past decade contribute to a surge of power conversion efficiency (PCE) from 4% to 25.2% (recorded in the year of 2020) at the laboratory‐perovskite‐cell.^{[ 2 ]} Although such an evolution represents a big accomplishment, the research has predominantly focused on achieving a record‐high PCE, with stability and durability often being insufficiently emphasized. Competition with incumbent photovoltaic (PV) technologies require perovskite cells to be highly stable and durable to accommodate with the industry standard of 20 year survival in outdoor conditions. Solicitation of year‐scale maintenance of a PCE floor from an initial efficiency over 18% has been released from U.S. Department of Energy (DOE).^{[ 3 ]} Advanced perovskite PV technology with both desirable PCE and durability is thus considered as a paramount step bridging the ongoing research to future commercialization.

The perovskite solar cells are thin film PV device built by layer‐by‐layer processing with incorporation of multiple functional charge extraction/transport/buffer layers and the perovskite light‐harvesting layer.^{[ 4} , ⁵ , ⁶ , ^{7 ]} Conventionally, “n‐i‐p” structured perovskite PV device (with n‐type buffer layer being coated at the front transparent electrode and p‐type being the back contact buffer) has been widely studied, but the commonly used compact TiO₂ front and 2,2′,7,7′‐Tetrakis[N,N‐di(4‐methoxyphenyl)amino]‐9,9′‐spirobifluorene (Spiro‐OMeTAD) back buffer layers have high‐temperature sintering concern^{[ 8 ]} and high‐cost issue,^{[ 9 ]} respectively. These would virtually rise the levelized cost of energy (LCOE) for perovskite PV products.^{[ 10 ]} Moreover, the ionically doped Spiro‐OMeTAD^{[ 11 ]} usually exhibit strong hygroscopic behavior that will accelerate the moisture‐assisted degradation of the cell.

Alternatively, inverted “p‐i‐n” structured devices (with p‐type being the front buffer and n‐type being the back contact buffer) with low‐temperature processed inorganic buffer layers have been developed aiming at a lower LCOE and higher durability. In specific, inorganic p‐type semiconductors such as NiO_x,^{[ 12 ]} CuSCN,^{[ 13 ]} Cu₂O,^{[ 14 ]} CuO,^{[ 15 ]} PbS,^{[ 16 ]} and graphene oxide (GO),^{[ 17} , ^{18 ]} etc., have been introduced as front anode buffers in inverted perovskite cells. However, the back‐contact buffers lack efficient material candidates. So far, most widely used back contact electron extraction/transport materials are the fullerene‐based materials.^{[ 16 ]} While the dilemma between efficient out‐of‐plane electron transit and the in‐plane full film‐coverage represents a major challenge in those fullerene‐based top buffers (or back contact buffer).^{[ 16 ]} Replacement by inorganic metal oxide (IMO) or the IMO:fullerene composite would be a promising solution, as they could provide suitable material configuration for better‐matched energy level with different photoactive layers,^{[ 19 ]} more compact buffering to inhibit corrosion of the metal electrodes caused by the ionic diffusion (e.g., MA⁺, I⁻, etc.)^{[ 20 ]} from perovskite, and optimized electrical contact to maximize the electron extraction and transportation.^{[ 8 ]} Recently, Choy and co‐workers have developed a thick TiO₂ backbone film as the top cathode buffer which suppressed the monomolecular Shockley−Read−Hall recombination loss and ion diffusion induced degradation.^{[ 21 ]} And various 0D n‐type IMOs (ZnO, SnO₂, and Zn₂SnO₄)^{[ 22} , ²³ , ²⁴ , ^{25 ]} have also been used as top buffering in inverted perovskite cells. Although higher durability has been observed in perovskite PV cell using these IMOs top buffering, the PCE remains compromised (≤20%^{[ 8} , ^{21 ]}) and inferior to their “n‐i‐p” structured counterparts (21–22%^{[ 26} , ^{27 ]}).

Here in this scenario, we report the utilization of 1D cation‐doped brookite TiO₂ nanorod (NR) with embeddings of 0D fullerene as the top cathode buffering for inverted perovskite solar cell. Such a composite design was inspired by the effectiveness of 1D TiO₂ NR embedded by 0D C₆₀ on enhancement in photo‐electrochemical response in photocatalytic field.^{[ 28 ]} In solar cells, the embedded 0D fullerene (C₆₀) could fill the interstitial space of the NR network, making it a more compact capping layer to protect the perovskite as well as enhance the electrical contact with the top electrode. In addition, we use the cation‐dopant of Co and Fe in TiO₂ NR to further modify the electronic property such as energy level and conductivity, which helps improving the charge extraction and transportation at the cathode side. Overall, we employed this C₆₀:M‐TiO₂ (M = Fe or Co) composite as the top buffering layer and demonstrated the highly efficient inverted perovskite solar cell, which exhibits a high PCE over 22%. The device also exhibits 22‐fold prolonged lifetime under continuous radiation of Xenon arc lamp (AM 1.5, including UV irradiance) and a survival PCE of 21.5% after 2‐month storage in ambient atmosphere with Realtime relative humidity (RH) ranging from 20% to 87%, indicating a greatly enhanced device durability. Our attempt in TiO₂ based 0D:1D composite opens a new dimension in simultaneously achieving highly‐efficient and ‐durable inverted perovskite PV cell, which would further inspire a broader material inventory for nanomaterials in application of electrode buffer in PV cells.

Figure  1A shows the inverted “p‐i‐n” structured solar cell devices, with a 70 nm thick layer consisting of C₆₀:M‐TiO₂ composite being the top electron transfer layer (ETL). 1D TiO₂ NRs are well‐known nanomaterials^{[ 29 ]} and have been confirmed to be superior to other dimensional nanostructures such as 0D nanoparticles, as they can offer a direct oriented path for electron so that the charge movement could have less transversal recombination loss.^{[ 30} , ^{31 ]} In addition, to maximize the charge transport and electrical band structural benefits, we further dope the 1D TiO₂ NRs with foreign elements, including cations of Fe and Co, which can tune lattice oxygen vacancy and evoke changes in electronic structure, surface potential, and other opto‐physical properties.^{[ 32 ]} Our prior study has revealed that single transition metal (M) dopants in a broad range could be homogeneously doped into monodisperse single‐crystalline brookite‐phase TiO₂ NRs (M‐TiO₂), which has led significant change in optoelectronic properties and made a difference in photocatalytic application.^{[ 32 ]} These homogeneously Fe‐ and Co‐ doped TiO₂ NRs are schematized in Figure 1B‐i). The single transition metal (M) dopants does not change the 1D nanostructure or the overall TiO₂ crystal phase. We use the transmission electron microscopy (TEM) to visualize the nanoscopic geometric feature. As shown in Figure 1B‐ii to iv, there is no physical dimensional change upon doping. Both Fe‐TiO₂ and Co‐TiO₂ maintain their NR feature inherited from the pristine TiO₂. Figure 1C shows the X‐ray diffraction (XRD) spectra of different NRs. In accordance, all the NRs share the X‐ray scattering feature of brookite‐phase TiO₂ (denoted with the lattice plane in Figure 1C). There is no significant change in lattice parameter which agrees well with the fact of low‐doping concentration and thereby the vast majority of the crystal maintains its TiO₂ brookite‐phase. To confirm this low‐doping concentration, we employed the energy dispersive X‐ray spectroscopy (EDS) in conjunction with scanning electron microscopy (SEM) to quantify the element composition. Figure 1D compares the EDS result of Fe‐TiO₂ and Co‐TiO₂ nanorods where characteristic X‐ray emitted from electron bombardment with Fe and Co are detected, respectively. The quantification of chemical composition is also shown in the tables inserted in Figure 1D. 7.8 and 6.7 mol% of Co and Fe, respectively, have been verified to be incorporated within the M‐TiO₂ NR. Thus, single transition metal doping of Co and Fe has been successfully introduced into TiO₂ NRs.

Figure 1.

A) Cross‐sectional SEM image of FTO/NiO_x/Perovskite/C₆₀:M‐TiO₂ NR composite (M = Fe or Co), showing a homogeneous layer of C₆₀:M‐TiO₂ RD as the top buffering layer for solar cell. B) Nanoscopic feature of M‐TiO₂ NR. i) Schematic showing the pristine and single transition metal doped M‐TiO₂. Corresponding TEM of ii) TiO₂, iii) Fe‐TiO₂, and iv) Co‐TiO₂. C) XRD spectra of different M‐TiO₂ NR, lattice planes of brookite‐phase are assigned. D) EDS results of Fe‐TiO₂ and Co‐TiO₂, collected from Figure S1 in the Supporting Information.

Next, we use these M‐TiO₂ NRs as matrix for fullerene imbedding and further apply the composite as cathode buffering in inverted perovskite solar cell. The motivation of utilization of fullerene as filling in the NR matrix is to ameliorate the morphological and micro‐electrical property, as the film only consisting of NRs contains a large ratio of incontinuity and interspatial gap, which will cause severe charge scattering or trapping loss and significantly limit the charge transport.^{[ 33 ]} In this study, both fullerene and M‐TiO₂ NRs are dissipated in nonpolar solvent such as chlorobenzene, which could be further spin‐casted on top of perovskite to form a homogeneous buffer layer of C₆₀:M‐TiO₂ composite (detailed in the Experimental Section) without degrading the bottom perovskite layer. We then compare the microscopic feature of different ETL buffer layers. Figure  2 displays the SEM images of different ETL buffer layers spin‐coated on top of perovskite layers. It could be seen that homogeneous ETL buffer layers with thickness of ≈70 nm have been realized using either pristine C₆₀ or the C₆₀:M‐TiO₂ composite materials. We found full and homogeneous coverage by different ETL buffer layers on top of (FAPbI₃)_0.95(MAPbBr₃)_0.05 perovskite. Interestingly, it is found that the M‐TiO₂ NR tends to have a short‐distance parallelly stacking pattern with amorphous fullerene embedded between gaps (schematized in Figure S2, Supporting Information). Such spontaneously formed parallelly stacking pattern of NR is similar to those of folded chains’ lamellae packing in polymer semi‐crystals,^{[ 34 ]} which might be due to the rearrangement of geometry in accordance to minimize the overall free energy as such a parallel packing tends to more efficiently reduce the surface energy.^{[ 35 ]} Overall, the C₆₀:M‐TiO₂ composite materials exhibit superior surface morphology and great film uniformity on top of perovskite.

Figure 2.

SEM images of different ETL buffer layers (on top of perovskite) consisting of A) pure C₆₀, composite of B) C₆₀:TiO₂, C) C₆₀:Fe‐TiO₂ and D) C₆₀:Co‐TiO₂, where i) are the cross‐sectional SEM images and ii) are the corresponding top‐view images.

As an ETL buffer layer, efficient electron extraction ability that can facilitate photocarrier flowing from perovskite to ETL and excellent electrical properties to transport those carriers with minimized dissipation loss inside the ETL are required for high‐performance perovskite solar sell. We employ the photoluminescence (PL) study to check the charge extraction ability of different ETL buffer layers. Figure  3A shows steady‐state PL spectra of perovskite itself and perovskite coated with different ETL materials (acting as quencher in the PL study). We use a constant excitation light source with identical wavelength (505 nm) and intensity for all the samples. Without any quencher, perovskite solely exhibits the highest PL intensity with a peak of ≈775 nm, a fingerprint of (FAPbI₃)_0.95(MAPbBr₃)_0.05.^{[ 36 ]} While with the presence of C₆₀ quenching layer, the PL intensity exhibits significant decreased yield which is due to the charge extraction from (FAPbI₃)_0.95(MAPbBr₃)_0.05 to the C₆₀ quenching layer (ETL buffer).^{[ 11 ]} Replacing the C₆₀ with C₆₀:M‐TiO₂ composite materials further decreases the PL, suggesting a more efficient photocarrier quenching process at the (FAPbI₃)_0.95(MAPbBr₃)_0.05/C₆₀:M‐TiO₂ interface. We then cross‐check the results through the time‐resolved PL (TRPL) spectroscopy. Figure 3B compares the TRPL spectra of perovskite only and perovskite coated with different ETL materials. For all the bi‐layer of (FAPbI₃)_0.95(MAPbBr₃)_0.05/C₆₀:M‐TiO₂, there is a single‐exponential decay featuring the fast charge extraction across the interface. This fits well with the ideal bi‐layer model of single‐exponential photocarrier quenching derived from continuum theory,^{[ 37 ]} as some reports also display a multi‐exponential decay containing a secondary slower process (longer photocarrier lifetime) which is corresponding to the photocarrier decay inner the bulk perovskite itself other than the interface^{[ 1} , ^{38 ]} (this should be excluded here). We then calculate the carrier lifetime based on following equation

$$n(t)=N\exp \left(-\frac{t}{\tau }\right)$$ (1)

with n(t) being the PL emission photon numbers collected at time t, N being the initial photon number and τ being the lifetime. Table  1 compares the lifetime of photo‐generated carriers in perovskite after being coated with different ETL buffer layers. Compared to the perovskite itself (τ = 341 ns), perovskite coated by pristine C₆₀ exhibits a reduced lifetime of τ = 138 ns; perovskite coated by C₆₀:TiO₂, C₆₀:Fe‐TiO₂ or C₆₀:Co‐TiO₂ exhibits further reduced lifetime of τ = 68, 54, and 41 ns, respectively. These results indicate a noticeably enhanced photocarrier quenching process in those (FAPbI₃)_0.95(MAPbBr₃)_0.05/C₆₀:M‐TiO₂ samples, where the photocarriers could be sufficiently extracted from (FAPbI₃)_0.95(MAPbBr₃)_0.05 to C₆₀:M‐TiO₂.

Figure 3.

A) Steady‐state PL and B) time‐resolved PL spectra of perovskite only and perovskite coated with different ETL materials. Mobility measurement based on space‐charge‐limited current (SCLC) method: C) corresponding J _SCLC‐V ² plot of electron‐only devices with structure of FTO/Al/ETL/Al. D) The electrical conductivity measured from devices of FTO/ETL/Al at different temperatures.

Table 1.

Photocarrier lifetime in perovskite when coated with different ETL buffer layers and electron mobility of ETL buffer layers measured from SCLC method

ETL buffer	Carrier lifetime in perovskite [ns]	Electron mobility [cm2 V−1 s−1]
C60	138	8 × 10−5
C60:TiO2	68	6 × 10−4
C60:Fe‐TiO2	54	8 × 10−4
C60:Co‐TiO2	41	9 × 10−4

As an efficient ETL buffer layer, another important role is to transfer the electron to the cathode with minimal loss, which requires the buffer material to have highly efficient electron transfer capability. We study the electrical transfer properties of these C₆₀:M‐TiO₂ ETL buffer materials. Figure 3C displays the J _SCLC‐V ² plot obtained from an electron‐only device of fluorine‐doped tin oxide (FTO)/aluminum (Al)/ETL/Al, where the J _SCLC is the dark current density of the electron‐only diode and V is the applied voltage. The electron mobility (μ) of ETL could be calculated from the equation of (using the Mott–Gurney square law):^{[ 39 ]}

$$J_{\mathrm{SCLC}}=\frac{9}{8}{\epsilon }_{\mathrm{r}}{\epsilon }_0\mu \frac{V^2}{L^3}$$ (2)

with ε _r being the dielectric constant, ε ₀ being the permittivity of free space, L being the thickness of ETL, respectively. The results are shown in Table 1. The pristine C₆₀ ETL displays an electron mobility of 8 × 10⁻⁵ cm² V⁻¹ s⁻¹, which is comparable to prior reported values.^{[ 40} , ^{41 ]} Meanwhile we found the C₆₀:M‐TiO₂ composite ETL exhibits one magnitude higher electron mobility (e.g., C₆₀:Fe‐TiO₂ displays a μ of 8 × 10⁻⁴ cm² V⁻¹ s⁻¹), as TiO₂ has been reported to have a higher mobility than fullerene.^{[ 42 ]} In addition, by single transition metal doping, we found the C₆₀:Fe‐TiO₂ and C₆₀:Co‐TiO₂ display slightly higher mobility than the C₆₀:TiO₂. Our prior studies have revealed that the homogeneous foreign transition metal doping in TiO₂ could help to tune the electron band structure which has been confirmed to affect the photoelectronic behavior in photocatalytic application.^{[ 32 ]} Physically, the transition metal doping in the lattice will adjust the electronic band structure in the momentum space, while the band curvature (particularly the extreme of the band) is closely related to the effective mass of charge carrier which is inversely proportional to the carrier mobility.^{[ 43 ]} Hence the slight change in mobility of C₆₀:M‐TiO₂ composite can be ascribed to the single transition metal doping effect. To further distinguish the electrical property between pristine C₆₀ ETL and C₆₀:M‐TiO₂ ETL, we measure the electrical conductivity of different ETLs at various temperatures. As shown in Figure 3D, C₆₀:M‐TiO₂ ETL exhibits slightly higher electrical conductivity (e.g., 0.66 mS m⁻¹) than pristine C₆₀ ETL (0.56 mS m⁻¹) at room temperature (RT). As temperature increases, the C₆₀:M‐TiO₂ ETLs exhibit an increasing in conductivity (e.g., from 0.66 mS m⁻¹ at 25 °C to 2.25 mS m⁻¹ at 100 °C for C₆₀:Co‐TiO₂), suggesting a thermally activated conductive behavior and a higher carrier density at elevated temperatures. In contrast, electrical conductivity of the pristine C₆₀ ETL slightly decreases along with the temperature's increase (shown in Figure S3 (Supporting Information), where conductivity displays a drop from 0.56 mS m⁻¹ at 25 °C to 0.52 mS m⁻¹ at 100 °C), presumably due to the activation of defects in C₆₀ film at higher temperatures.^{[ 44 ]}

Above we have shown the C₆₀:M‐TiO₂ ETL buffer layers displaying superior electron extraction and transfer properties in comparison with pristine C₆₀ ETL. Additionally, for efficient perovskite solar cells, the energy level alignment between perovskite photoactive layer and the ETL is a crucial factor determining the overall device performance. We employ the Kelvin probe force microscopy (KPFM) to microscopically investigate the surface of different ETL buffer layers. Figure  4A compares the topographical images of different ETLs and their CPD (corresponding contact potential difference) images. CPD reflects the potential difference between work function of tip (Φ _tip) with respect to the surface potential of sample, which could be used to calculate the quasi‐Fermi‐level of electron (QFLE) of the sample by knowing the tip's work function.^{[ 45 ]} QFLE defines the population of electrons separately in the conduction band and valance band, when their population are displaced from equilibrium due to heat, light, bias, etc., which could more precisely reflect the electronic nature of a material in real case.^{[ 43 ]} Figure 4B shows the distribution plot of CPDs obtained from different samples. We use Gaussian distribution to fit the histogram of CPDs, as detailed in Note S1 (Supporting Information). It can be seen that there is an obvious difference in CPD distribution between pristine C₆₀ ETL and the C₆₀:M‐TiO₂ composite ETLs. As listed in Table S1 (Supporting Information), pristine C₆₀ ETL exhibits a mean value and variance of CPD of 0.54 and 0.157 respectively. While the C₆₀:M‐TiO₂ ETL exhibits higher CPD but narrower variance. For example, C₆₀:Co‐TiO₂ displays a mean value and variance of 0.98 and 0.061 respectively. The narrower CPD in those C₆₀:M‐TiO₂ ETLs suggest a higher microscopic uniformity in terms of the electronic property. We further quantify these CPDs in terms of value of QFLE in Figure 4C (detailed in Note S2 (Supporting Information) for the conversion of CPD into QFLE), along with the root‐mean‐square (RMS) roughness (R _q) obtained from the corresponding topographical images. The R _q of all the ETLs display a similar value locked in a narrow range of 3.2–4.6 nm. For example, the pristine C₆₀ ETL displays a R _q of 4.6 nm and C₆₀:Fe‐TiO₂ ETL displays a R _q of 4.2 nm. Taking consideration of the thickness of these ETLs (≈70 nm in Figure 2), the small R _q of <5 nm for all the C₆₀:M‐TiO₂ composite ETLs suggests a high film homogeneity which is expected to have a good electrical contact with top metal electrode. On the other hand, we found the pristine C₆₀ ETL exhibits a typical QFLE of −4.53 eV (could be considered as the Realtime conduction band minimal). In contrast, C₆₀:M‐TiO₂ composite ETLs exhibit higher QFLE. Specifically, C₆₀:TiO₂, C₆₀:Fe‐TiO₂, and C₆₀:Co‐TiO₂ display the QFLE of −3.83, −4.04, and −4.09 eV, respectively.

Figure 4.

A) KPFM measurement on different ETL buffer layers. The top row shows the AFM topographies and the bottom the corresponding KPFM images. B) CPD distribution of different ETLs. C) Comparison of R _q and quasi‐Fermi‐level of electron (QFLE) of different ETLs.

These elevated QFLE in C₆₀:M‐TiO₂ ETL buffer layers are expected to have better energy level alignment with the conduction band of (FAPbI₃)_0.95(MAPbBr₃)_0.05 perovskite. Figure  5A displays the diagram of energy level alignment in the inverted perovskite solar cell. At the cathode side, there is a larger energy level mismatch between the conduction band minimal (CBM) of (FAPbI₃)_0.95(MAPbBr₃)_0.05 and C₆₀ (with a big difference of 530 meV). In contrast, the contact between (FAPbI₃)_0.95(MAPbBr₃)_0.05 and C₆₀:TiO₂, or C₆₀:Fe‐TiO₂ or C₆₀:Co‐TiO₂ ETLs displays smaller energy level mismatch of 170, 40, and 90 meV, respectively. This better‐matched energy level of electron secures a more efficient interfacial transfer, as verified in above PL studies. In a solar cell the photoelectrons originate from the photoexcitation of valance band (VB) electrons to CBM of perovskite layer, and they have the initial density of state (DOS) distribution at the CBM of perovskite. After that, these CBM photoelectrons will need to transfer across multiple layers/interfaces toward the electrode, through which their electrical potential will be lost. Smaller energy level offset between ETL and perovskite will reduce the potential lost during the transfer of those photoelectrons across the interfaces, leaving a higher residual potential at the electrode and thereby a higher V _OC. We then use these different ETL as the top buffer layers in an inverted perovskite solar cell, with a device structure of FTO/NiO_x/(FAPbI₃)_0.95(MAPbBr₃)_0.05/ETLs/BCP/Ag, where FTO is fluorine‐doped tin oxide, BCP is bathocuproine and Ag is silver. Figure 5B shows the current density–voltage (J–V) curve of solar cells using different ETLs. The photovoltaic parameters are listed in Table  2 . The inverted perovskite solar cell using pristine C₆₀ ETL as top buffer exhibit a J _SC (short‐circuit current density) of 23.04 mA cm⁻², V _OC (open‐circuit voltage) of 1.00 V, FF (fill factor) of 0.768, and PCE of 17.69%. In comparison, using C₆₀:TiO₂ ETL, we found a significant improvement in PCE of 20.89%, with simultaneously improved J _SC of 23.85 mA cm⁻², V _OC of 1.08 V, and FF of 0.811. Moreover, using the C₆₀:Fe‐TiO₂ and C₆₀:Co‐TiO₂ ETL, there is a further improvement in device performance. In particular, the device incorporated with C₆₀:Co‐TiO₂ ETL exhibits a maximal PCE of 22.13% with J _SC of 24.38 mA cm⁻², V _OC of 1.10 V, and FF of 0.825. We have shown above that the Co‐TiO₂ ETL has optimized energy level alignment with (FAPbI₃)_0.95(MAPbBr₃)_0.05 perovskite (Figure 5A), higher electron extraction efficiency and superior electrical properties (Figure 3), which jointly result in the highest device efficiency in the corresponding solar cells. In particular, the higher electron mobility in those C₆₀:M‐TiO₂ ETL ensures an efficient charge transfer and thereby less charge carrier recombination losses throughout the cathode side, leading to a higher FF for the solar cell devices. The enlarged V _OC could be ascribed to the minimized energy level mismatch that reduces the potential loss during the charge transfer across the perovskite/ETL interface, as evidenced by the KPFM study.^{[ 11 ]} Such a smaller energy level offset accompanied by the more sufficient interfacial charge extraction at the perovskite/C₆₀:M‐TiO₂ ETL interface secures less potential loss.^{[ 46 ]} To verify the boosted photocurrent by the C₆₀:M‐TiO₂ ETL, we measured the incident photon‐to‐current efficiency (IPCE) spectra of inverted perovskite solar cells using different ETLs (Figure 5C). We also integrate the spectra with regarding to the solar spectra to calculate the J _SC obtained from IPCE (Figure S5, Supporting Information). The results are listed in Table 2. The J _SC obtained from IPCE of device using C₆₀, C₆₀:TiO₂, C₆₀:Fe‐TiO₂, and C₆₀:TiO₂ ETL are 22.85, 23.70, 24.21, and 24.36 mA cm⁻², respectively. These values are in great consistence to those obtained from J–V curves (as can be seen in Table 2) with a negligible mismatch <1%. Such a slight deviation could be related to instrumental error such as different light spectrum. Over the entire response region from 300 to 850 nm, the C₆₀:M‐TiO₂ ETL based devices exhibits overall higher IPCE intensities than the C₆₀ ETL based device. Considering all the devices use the identical perovskite layer and the only difference is the ETLs, the higher IPCE then originates from the more efficient charge extraction/transport in these C₆₀:M‐TiO₂ ETLs based devices. The photocurrent hysteresis is also studied. Figure 5D displays the J–V curves of inverted perovskite solar cell using C₆₀:Co‐TiO₂ ETLs tested with both reverse and forward scan. The device exhibits a PCE of 21.60% under forward scan and 22.15% under reverse scan, exhibiting a relatively small hysteresis with a hysteresis index ($\frac{\mathrm{PC}{\mathrm{E}}_{\mathrm{Reverse}}-\mathrm{PC}{\mathrm{E}}_{\mathrm{Forward}}}{\mathrm{PC}{\mathrm{E}}_{\mathrm{Reverse}}}$) of 0.025 which is at a small level comparing those reported values.^{[ 47 ]} The hysteresis in perovskite solar cell remains to be a complex phenomenon with potential origins including the chemical composition (e.g., stoichiometry, internal interfaces and trap density of perovskite), electrical history (scanning history), atmospheric conditions (e.g., light soaking, temperature, humidity, oxygen, etc.) as well as their interdependent effects. Prior study has concluded major factors dominating the photocurrent hysteresis in perovskite solar cell to be the trap‐involved charge carrier trapping/detrapping and/or the ionic motions across the interfaces and the interfacial issues are the main origins for the hysteresis.^{[ 48} , ^{49 ]} Here the relatively small hysteresis might be due to the more efficient interface between the (FAPbI₃)_0.95(MAPbBr₃)_0.05 perovskite and C₆₀:M‐TiO₂ ETLs where the optimized interface contact and incorporation of the dense ETL which could block the heavy ion (Pb²⁺, I⁻) motion across the interface and relieve the hysteresis in certain degrees.^{[ 50 ]} Overall, a high efficiency over 22% has been successfully achieved in an inverted perovskite solar cell using the C₆₀:Co‐TiO₂ ETL cathode buffer. We also check the reproducibility of our devices. Figure 5E shows the statistics of PV parameters collected from 12 individual devices (Normal distribution fitting and data colony is shown in Figure S6, Supporting Information). The control device using C₆₀ ETL shows an average PCE of 16.6% with a larger error range of 1.2%. While the device using C₆₀:Co‐TiO₂ ETL displays a highest average PCE of 21.4% with a smaller error of 0.8%, indicating a higher reproducibility. The higher consistence in C₆₀:Co‐TiO₂ ETL based device is much clearer in FF, where it has a range of ±1.6% while the C₆₀ ETL device has a range of ±3.0%. The high consistence and reproducibility might be related to the high electronic uniformity of the ETL film (narrower CPD distribution in Figure 4B and Table S1, Supporting Information) and the ultra‐flat feature (in the cross‐sectional SEM in Figure 2). Overall, the inverted perovskite solar cell using the C₆₀:M‐TiO₂ ETL cathode buffer exhibit higher PCEs than the C₆₀ ETL based device (with an average PCE of 16.6%). In addition, with Fe‐ and Co‐dopant, the average PCE could be further improved to 20.8% and 21.4%, respectively, from that of 19.9% of C₆₀:TiO₂ ETL based cell. Meanwhile, the C₆₀:M‐TiO₂ ETL based device also exhibit highest PCE over 22%, indicating the effectiveness of using this 0D:1D composite cathode buffering material.

Figure 5.

A) Diagram of energy level alignment of inverted perovskite solar cell with architecture of FTO/NiO_x/perovskite/ETL buffer/BCP/Ag. B) J–V curve and C) IPCE spectra of inverted perovskite solar cell using different ETLs. D) J–V curve of inverted perovskite solar cell using C₆₀:Co‐TiO₂ ETLs. E) Statistics of photovoltaic parameters in the inverted perovskite solar cell with different ETLs.

Table 2.

Photovoltaic parameters of inverted perovskite solar cells using different ETLs

ETLs	J SC a)	J SC b)	V OC	FF	PCE
	[mA cm−2]	[mA cm−2]	[V]		[%]
C60	23.04	22.85	1.00	0.768	17.69
C60:TiO2	23.85	23.70	1.08	0.811	20.89
C60:Fe‐TiO2	24.25	24.21	1.10	0.811	21.63
C60:Co‐TiO2	24.38	24.36	1.10	0.825	22.13

^a) J _SC obtained from J–V curve

^b)J_SC obtained from integration of IPCE spectra.

In order to understand the positive role of C₆₀:TiO₂ ETL in terms of charge carrier dynamics in the inverted device, we also carry out the transient photovoltage (TPV) and transient photocurrent (TPC) analysis. Figure S7 (Supporting Information) shows the transient measurement results of a C₆₀:TiO₂ ETL based inverted device with a comparison from a reference cell using C₆₀ ETL. Briefly, we measure the photovoltage decay (TPV) under the open‐circuit condition with a pulse light for photoexcitation. During the pulse period of illumination, the separation of photo‐generated electrons and holes leads to a photovoltage generation, which decays upon light‐off due to the trap‐assisted charge carrier recombination.^{[ 51 ]} The C₆₀:TiO₂ ETL based device exhibits a slower photovoltage decay with an elongated lifetime of 5.8 µs, over twofold larger than that of 2.4 µs from the reference device (Figure S7A, Supporting Information). This observation is consistent to the enlarged V _OC in the C₆₀:TiO₂ ETL based device, both of which can be ascribed to a reduced trap density and improved extraction ability at the perovskite/C₆₀:TiO₂ ETL interface. In parallel, we also record the TPC of the same devices. Under short‐circuit condition, the photo‐generated carriers can be swept out and quickly recombined with injected carriers at corresponding electrodes and the TPC decay illustrates the transit time of photocarriers across the interfaces. The C₆₀:TiO₂ ETL based device exhibits a faster transit with a time of 206 ns, nearly threefold smaller than that of 616 ns from the reference device (Figure S7B, Supporting Information). Such an enhanced transit can be attributed to the enhanced interfacial electron extraction between the perovskite and ETL, as evidenced by the PL results. Overall, both TPV and TPC results are consistent with the conclusions from the TI/TR‐PL measurements, indicating a more efficient interfacial charge transport by using the C₆₀:TiO₂ ETL and thereby leading to reduced recombination loss and higher device performance.

Lastly, we investigate the device stability under multiple conditions. We take the inverted perovskite solar cell using the C₆₀:Co‐TiO₂ ETL cathode buffer as an example, as it displays the highest PCE. Figure  6A shows the static current density and PCE as a function of the time for the C₆₀:Co‐TiO₂ ETLs devices, biased at its maximal power point (MPP) of 0.96 V and one‐sun illumination (AM 1.5) which is obtained from the power density‐bias curve in Figure S8 (Supporting Information). The C₆₀:Co‐TiO₂ ETL based cell exhibits a stabilized J _mp (current density at the maximal power point) of 23.21 mA cm⁻² over a 300 s period. This gives rise to a steady output power of 22.28 mW cm⁻², corresponding to a PCE of 22.28%, which is in great consistence with the value of 22.13% extracted from the J–V curve. In terms of long‐term stability, the perovskite device needs to have certain resistance to moisture, since the core material (halide perovskite) is sensitive to moisture.^{[ 52 ]} In solar cell device, the buffer layer on top of perovskite is even more crucial in protecting the sensitive perovskite layer underneath. While in typical “n‐i‐p” perovskite solar cells, the ionically doped Spiro‐OMeTAD is usually used as the top buffering on perovskite layer, which however exhibits strong hygroscopic feature that would accelerate the degradation of perovskite.^{[ 11} , ^{53 ]} Here we use the C₆₀:Co‐TiO₂ ETL composite that exhibits a great hydrophobicity than those hygroscopic buffers in “n‐i‐p” devices. Figure 6B shows the water contact angle measurement on different ETLs directly coated on top of perovskite. It is clear that the pristine C₆₀ ETL exhibits a small angle of 48°; while the composite buffer layer of C₆₀:TiO₂, C₆₀:Fe‐TiO₂, and C₆₀:Co‐TiO₂ displays increased contact angle of 73°, 84°, and 90°. Hence, utilization of these C₆₀:Co‐TiO₂ composite top buffer is anticipated to have a stronger moisture resistance in the ambient, as it would be more difficult for the moisture infiltrate into the bottom perovskite layer to trigger the degradation reaction. We then measure the working lifetime of inverted solar cells using C₆₀:Co‐TiO₂ ETL and pristine C₆₀ ETL. Both devices are unencapsulated and are continuously working under an Xe lamp (AM 1.5, without any optical filter) in ambient atmosphere. It is found that the C₆₀:Co‐TiO₂ ETL based device exhibit a much slower decay than that using pristine C₆₀ ETL. Briefly, the pristine C₆₀ ETL device show a quick PCE drop from its original number to 0 within 14 h, while the C₆₀:Co‐TiO₂ ETL device remains 50% of its original PCE even after 110 h irradiation. We also use the linear function to fit the curve, where the slope of the fitting line represents the decay rate (ΔPCE h⁻¹) of the device. The C₆₀:Co‐TiO₂ ETL device displays one‐magnitude slower decay rate (0.5% h⁻¹) than the C₆₀ ETL device (5.6% h⁻¹), suggesting a much higher working stability. On the other hand, we also study the shelf‐life of these inverted devices (without encapsulation) using either C₆₀ or C₆₀:Co‐TiO₂ ETL. All the devices are stored in ambient atmosphere with Realtime relative humidity (RH) ranging from 20–87%. The device PCE is tested once every day by the J–V curve measurement. Figure 6D displays the result. As expected, the control device using the pristine C₆₀ ETL display a faster decay. The PCE drops from 17.1% to 14.4% in the first week and then gradually drops to 0 within one month. In contrast, the C₆₀:Co‐TiO₂ ETL device maintains an extremely excellent performance after 70 days storage, with a decent survival PCE of 21.5%. Overall, utilization of C₆₀:Co‐TiO₂ ETL as the top cathode buffer enables highly durable perovskite cells, which makes a breakthrough on the device shelf lifetime extension from a typical daily‐scale to a decent monthly‐scale. We anticipate the post encapsulation in conjunction of using our method would further prolong the device lifetime, making it possible for transition toward industry.

Figure 6.

A) Static current density and PCE measured as a function of the time for the C₆₀:Co‐TiO₂ ETLs devices, biased at its maximal power point (MPP) of 0.96 V. B) Contact angle measurement of water droplet on different ETLs. C) Stability test under continuous one‐sun Xe lamp illumination in ambient air, measured by tracking the current at MPP of an unencapsulated device for 120 h. D) Device shelf life measurement by testing the J–V curve every day for over two months. Realtime relative humidity (RH) is also included for reference.

In summary, we have demonstrated that the utilization of 0D:1D composite consisting of fullerene and M‐TiO₂ nanorods as the top ETL buffering in inverted perovskite solar cell represents as an efficient way to simultaneously improve the device efficiency and lifetime. Excitingly, we found the composite of C₆₀:Co‐TiO₂ ETL exhibits improved film uniformity, enhanced electron extraction and transfer ability and the better energy level matching with perovskite, as revealed by a suite of spectroscopic and optoelectronic characterizations. Moreover, we elevated the PCE baseline of inverted perovskite solar cell from <20% to ≈22% and demonstrated an extended lifetime from daily‐ to monthly scale, making it possible to accelerate the transition of this technology from lab research to the industry field. In a larger scope, the concept of 0D:1D composite buffering layer (consisting of carbon material and well‐tuned transition metal oxide nanostructures) is expected to open up more opportunities in expanding the material inventory for charge transfer layer in perovskite solar cells development and application.

Experimental Section

Material

Lead (II) bromide (PbBr₂, 99.99%), Lead (II) iodide (PbI₂, 99.99%), and Iron (III) chloride anhydrous (FeCl₃, 98%) were purchased from Alfa Aesar. Cobalt (II) acetylacetonate (Co(C₅H₇O₂)₂, 99%), Octadecene (ODE, 90%), and titanium chloride (TiCl₄, 99.9%) were obtained from AcrosOrganics. Oleylamine (OAm, 70%), Oleic acid (OAc, 90%), Methylammonium bromide (MABr), Formamidinium Iodide (FAI), Ni‐(NO₃)₂·6H₂O (98%), C₂H₂O₄·2H₂O (99.5%), NaOH (96%), Dimethylformanmide (DMF, extra dry, 99%), Dimethyl sulfoxide (DMSO, extra dry, 99%), 1, 2‐dichlorobenzene (DCB, extra dry, > 98%), Chlorobenzene (CB, extra dry, 99.8%), Isopropanol (extra dry, 99.8%), and other chemicals were purchased from Sigma‐Aldrich.

Synthesis of TiO₂ and M‐TiO₂ Nanorod

Brookite‐phase TiO₂ NRs were synthesized according to the prior reported approach.^{[ 32 ]} Briefly, 10 mL OAm, 10 mL ODE and 0.48 mL OAc were first degassed under vacuum at 90 °C for 1 h to evacuate moisture and oxygen. Then the mixture was cooled down to 60 °C under nitrogen atmosphere with 1.5 mL Ti‐precursor solution containing 1.0 m OAc and 0.2 m TiCl₄ in ODE being injected into the solution. ODE and OAc in the precursor were pre‐dried and stored in glovebox to avoid undesirable hydrolysis in TiCl₄. The reactor was then quickly heated to 290 °C and being hold at 290 °C for 10 min. Then, another 8 mL Ti‐precursor solution was added dropwise into the reactor with a dropping rate of ≈0.3 mL min⁻¹. After that, the system was cooled down to room temperature and the TiO₂ NRs were collected and washed by addition of isopropanol, followed by centrifugation at 8000 rpm for 8 min. The product was further purified twice by the addition of isopropanol and hexane. For brookite‐phase M‐TiO₂ NRs, the synthesis is similar to that for pristine TiO₂ NRs, except that the TiCl₄‐M‐oleate mixed solution was used for the M‐TiO₂ NRs. Taking Fe‐TiO₂ NRs as an example, the Fe‐oleate precursor was first mixed with the Ti‐precursor with a desired Ti/Fe ratio. And 1.5 mL of this mixed solution was injected into the degassed solution of 10 mL ODE, 10 mL OAm, and 0.48 mL OAc at 60 °C, which was quickly heated up to 290 °C. After 10 min heating, another 8 mL TiCl₄‐Fe‐oleate mixed solution was introduced into the system. The as‐prepared Fe‐TiO₂ NRs were collected and purified in the same way to that of pristine TiO₂ NRs. The Co‐TiO₂ NRs were synthesized in the similar method. It should be noted that this method is a general method to prepare new materials of M‐TiO₂ NRs solely or binary doped by other transitional metals.

Preparation of ETL

The ETL cathode top buffers were spin‐casted from solutions containing either fullerene or NRs or mixture of fullerene and NRs. For C₆₀ ETL, a solution of 10 mg mL⁻¹ C₆₀ in mixed solvents of CB and toluene (1:1 v/v) was spin‐casted at a rate of 1000 rpm for 50 s. For C₆₀:TiO₂ (or C₆₀:M‐TiO₂) ETLs, a solution containing 5 mg C₆₀ and 5 mg TiO₂ (or M‐TiO₂) in 1 mL mixed solvents of CB and hexane (1:1 v/v) was spin‐casted at a rate of 1000 rpm for 50 s. All the solutions were prepared in ambient atmosphere by dispersing different materials in the solvents under ultrasonication for 30 min, followed by filtration with a 0.45 µm polyvinylidene fluoride (PVDF) filter. The solution was used right away.

Device Fabrication

The FTO glass was cleaned by ultrasonication in bath of detergent, deionized water, acetone and isopropanol sequentially, followed by drying in an oven at 100 °C overnight. The pre‐cleaned FTO glass substrates were then treated by UV–ozone plasma for 40 min. Afterward, a ≈20 nm thick NiO_x was spin‐casted from NiO_x nanocrystal solution at a spin‐rate of 4000 rpm for 40 s according to prior reports.^{[ 54 ]} Then the (FAPbI₃)_0.95(MAPbBr₃)_0.05 perovskite photoactive layer was spin‐coated by a one‐step method. Briefly, a 1.2 m mixed (FAPbI₃)_0.95(MAPbBr₃)_0.05 perovskite precursor solution with a mixed solvent of DMF:DMSO (8:2 v/v) was spin‐coated at 5000 rpm for 30 s. During the spinning process, 200 mL of CB was dropped onto the spinning surface (in the last 20 s). After that, the samples were then annealed at 70 °C for 15 min followed by annealing at 100 °C for 50 min in the glove box filled with nitrogen. After cooling down to room temperature, different ETLs were spin‐coated on top of perovskite from above solution with a spin‐rate of 1000 rpm for 50 s. Lastly, a 5 nm thick bathocuproine (BCP) and 100 nm thick silver (Ag) film was sequentially deposited on top through a shadow mask in the vacuum of <5 × 10⁻⁶ mbar. The device area was measured to be 0.12 cm².

Material Characterization

Cross‐sectional SEM images of FTO/NiO_x/Perovskite/C₆₀:M‐TiO₂ NR composite and SEM images of different ETL buffer layers on top of perovskite were obtained from LEO 1530 MERLIN (FESEM). The TEM images of TiO₂ and M‐TiO₂ NRs were obtained from FEI Spirit (120 kV). XRD spectra of TiO₂ and M‐TiO₂ NRs were obtained from the Philips Xpert Pro X‐ray diffractometer. Energy dispersive X‐Ray spectra (EDS) of M‐TiO₂ NRs and their corresponding SEM images were obtained from a field emission scanning electron microscopy (FE‐SEM, FEI Sirion‐200 SEM) at an acceleration voltage of 5 kV. PL spectra of perovskite only and perovskite coated with different ETL buffer layers were obtained from a FLS1000 Photoluminescence Spectrometer, Edinburgh Instruments. KPFM images of different ETL buffer layers were obtained from Bruker Innova AFM platform.

Device Characterization

The solar cell device performance was characterized by measuring the J–V characteristics under one‐sun illumination (AM 1.5), provided by a Xenon solar simulator (Newport) in a glove‐box. The J–V scan rate is programed to be 100 mV s⁻¹. The solar spectra were calibrated to be 100 mW cm⁻² by a standard Si cell. The IPCE spectra were obtained from a home‐made setup consists of a stabilized Xenon lamp, beamsplitter, monochromator, photodiode connected to a power meter and a Keithley 2400 SourceMeter to record the photocurrent. Device working‐life plot was measured by tracing the photocurrent of different unencapsulated solar cells under their maximal power point voltage. All the cells are continuously working under a Xenon lamp (AM 1.5, without any optical filter) in ambient atmosphere, with a continuously supplying MMP bias for mimicking real working condition, which is different from prior “aging test” that tests periodically “working/rest” device. Device shelf‐stability was determined by measuring the J–V characteristics of different unencapsulated solar cells under one‐sun illumination every day. All the cells are stored in ambient atmosphere with Realtime relative humidity (RH) ranging from 20–87%. TPC and TPV measurements were conducted by a custom built‐automated setup, where a white light bias was generated from a diode array to simulate the illuminating working condition for the sample device, a pulsed laser was used as the perturbation source, and a digital oscilloscope was used to collect the data.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Hu X., Liu C., Zhang Z., Jiang X.‐F., Garcia J., Sheehan C., Shui L., Priya S., Zhou G., Zhang S., Wang K., 22% Efficiency Inverted Perovskite Photovoltaic Cell Using Cation‐Doped Brookite TiO₂ Top Buffer. Adv. Sci. 2020, 7, 2001285 10.1002/advs.202001285