Low-Temperature Growing Anatase TiO2/SnO2 Multi-dimensional Heterojunctions at MXene Conductive Network for High-Efficient Perovskite Solar Cells

Highlights Nanoscale multi-dimensional heterojunctions in situ grow at the edge of two-dimensional MXene conductive network. A controlled anneal procedure in 150 °C is for preparing anatase TiO2/SnO2 heterojunctions with oxygen vacancy scramble effect. The perovskite solar cells achieve high power conversion efficiency and high moisture-resistance stability. Electronic supplementary material The online version of this article (10.1007/s40820-020-0379-5) contains supplementary material, which is available to authorized users.


Introduction
The organic-inorganic hybrid lead halide perovskite solar cells (PSCs) have shown tremendous promotion of power conversion efficiency (PCE) from 3.8% to a certified 25.2% in less than a decade, with fascinating industrial potentials of low cost and simple solution process [1,2]. As an important functional interlayer in PSC device, the electron transport layer (ETL) plays an important role. In a conventional upright structure in PSCs, the light absorption property of perovskite layer can be affected by the optical transmittance property of ETL by reducing optical energy loss [3]. The charge carriers in device can be regulated by the semiconductive energy-level property of ETL, through selectively transporting electrons and blocking holes from the adjacent perovskite layer [4]. Further, the amount of photon-excited carrier in perovskite layer can be restricted by the ETL, because the crystal size of perovskite can be affected by the hydrophobicity of the below ETL in preparation [5,6]. Therefore, in order to achieve high PCE, the ETL is required to be facially designed and fabricated based on its material and nanostructure to acquire high optical transmittance, matched semiconductive band, suitable material, and morphological features [7][8][9].
The SnO 2 has been demonstrated one of the superior electron transport materials (ETMs) in PSCs by harvesting the highest PCE of 21.6% to this day, because of its better optical and electric properties, energy band with respect to perovskite layer, and simple solution preparation in low-temperature anneal method, which can be potentially applied in flexible devices [10][11][12]. However, it still suffers from some drawbacks, such as poor wetting property, weak transmittance, and low conductivity [12]. The doping in solution preparation is one of simple effective methods to improve ETM, exhibiting a fascinating potential in large-scale industrial application. For the SnO 2 -based ETL, the anneal temperature is generally below 185 °C. On the basis of the solution preparation in low-temperature anneal, the energy band alignment can be engineered by ion doping of Y 3+ , Sb 3+ , Li + , Ti 4+ , Nb 5+ , and ethylene diamine tetraacetic acid in Sn ion or SnO 2 solution [13][14][15][16][17], and the electric conductivity can be facilitated by Li + -, Mg 2+ -, and Sb 3+ -based p-type dopant [15,18]. The ETL/perovskite interface can be improved by passivating the bulk defects and mitigating the interfacial charge recombination with Cl − and Y 3+ dopant [3,13]. Hereof, the Ti-based dopant is attractive, because the TiO 2 , as a most widely used ETM, is demonstrated to form effective bilayered TiO 2 /SnO 2 heterojunction, achieving the highest PCE of 21.4% recently [19,20]. However, the anneal temperature as high as 450-500 °C is generally necessary in these reports for forming effective TiO 2 phase in anatase lattice [21]. On the basis of a low-temperature anneal at 75 °C, the TiCl 4 solution can be translated into needle-like TiO 2 layer in rutile phase when it is heated by covering a soft film, yet the PSC based on this TiO 2 ETL just shows a PCE of 17.09% [22]. In a low-temperature anneal at 120 °C, a TiO 2 layer can be also formed based on aqueous TiCl 4 . Although the resultant TiO 2 /SnO 2 composite can contribute a maximum PCE of 21.27% [23], this PCE is still restricted by the amorphous phase of TiO 2 here [17]. Therefore, it is still deserved to adjust the Ti component on its lattice phase and nanostructure for gaining higher PCE of device.
The 2D materials and their tailored products, such as black phosphorene and 0D graphene quantum dot (QD), have been employed as dopants to modify the SnO 2 ETLs in PSCs on the electrical conductivity, energy-level properties, and reducing the surface defects and surface electron traps [24,25]. As a new group of 2D materials, the MXene is constructed of M N+1 X N T X structure, where the M represents an early transition metal such as Ti, and the X representing C and/or N element, and the T X meaning the terminal functional groups [26]. With its excellent metallic electrical conductivity, hydrophilic surface, and pseudocapacitance properties, the MXene has been explored as electrode materials for solar cells [27][28][29], zinc-ion capacitors [26,30], supercapacitors [31], or sensors [32]. In PSC application, it has also been reported that the PCE can be enhanced on the basis of the improved functional layers with MXene dopant. With Ti 3 C 2 T X MXene-doped perovskite layer, the crystal size of CH 3 NH 3 PbI 3 and charge transfer can be both increased [28]. Also, with Ti 3 C 2 MXene-doped SnO 2 ETL, the charge-transfer paths, electron mobility/extraction can be all improved in device [29]. Further, it is potential to advance the MXenedoped SnO 2 ETL by introducing the TiO 2 in suitable crystal phase for forming effective heterojunction structure.
By oxidation in a hydrothermal route and subsequent hydrazine hydrate reduction, a Ti 3+ -doped TiO 2 in rutile lattice can be composited with 2D Ti 3 C 2 MXene sheets, using MXene itself as Ti source [33]. By partial oxidation hydrothermal route, the 2D anatase TiO 2 can be in situ grown on Ti 3 C 2 for the interface with minimized defects [34]. Through in situ hydrolysis and heat treatment, the 3D anatase TiO 2 nanoparticles can be decorated on Ti 3 C 2 MXene [35]. Based on these hydrothermal methods, the nanoparticle size of anatase TiO 2 is about 30 nm, and the TiO 2 is assembled on the surface of Ti 3 C 2 T X MXene nanosheets [36][37][38][39]. Then, Schottky junction can be formed at the interface of TiO 2 /MXene for promoting the carrier separation by hole trapping effect [33,34]. Besides, with a method of high-energy ball mill, the anatase TiO 2 can be also generated on carbon nanosheets, and the raw 2D Ti 2 CT x MXene serves as Ti and carbon source [40]. On the other hand, the Ti-based oxide can be also grown at MXene in air environment, because of its unstable property. By annealing in 250 °C in air, the anatase TiO 2 can be formed at MXene [41]. Among these methods, the anneal is relatively economical. The resultant MXene-based anatase TiO 2 is potential to promote the PCE of SnO 2 -based PSCs by forming TiO 2 /SnO 2 heterojunction and conductive MXene pathways, because it is reported that the stack of MXene sheets can be prevented by the SnO 2 QDs as "spacer", and the highly conductive Ti 3 C 2 T X MXene can provide efficient pathways for fast transport of electrons [42]. However, the annealing temperature for MXene-based anatase TiO 2 is higher than that for SnO 2 in lowtemperature fabrication. Therefore, it is desired to improve this low-temperature anneal method for economical application.
Herein, the TiO 2 QDs in anatase phase are composited with SnO 2 ETM by adopting MXene in SnO 2 solution with a controlled anneal method in 150 °C low temperature, implemented in air and N 2 atmospheres. The nanoscale TiO 2 /SnO 2 heterojunctions are formed, and the connectivity of ETL is enhanced as the 2D MXene acting conductive bonded bridges, contributing a multi-dimensional conductive network (MDCN) structure for ETM. In addition, the crystal size of upper perovskite layer is enlarged and the ETL/perovskite interface is improved. With the optimal MDCN ETL, a champion PCE of 19.14% of PSC is achieved. The PSC also exhibits negligible hysteresis and higher moisture-resistance stability even after 45 days at room temperature, compared to the controlled devices.

MXene Preparation
The MXene nanosheet solution was prepared by the following manufacturing procedures [31]. After the deionized water was added to a hydrochloric acid solution to prepare 10 mL 9 M hydrochloric acid, 0.5 g LiF powder was slowly added to this hydrochloric acid solution and then was stirred for 10 min. Then, 0.5 g Ti 3 AlC 2 was slowly added into the above-mixed solution and then stirred at 35 °C for 24 h. Next, the residue was washed through centrifugation with deionized water at 3500 rpm for several times until pH > 6. Then, the product was sonicated under ice bath in an Ar atmosphere for 30 min. Finally, the product was centrifuged for 30 min at 3500 rpm to collect the MXene nanosheets. The concentration of resultant MXene was adjusted as 9 mg mL −1 .

MDCN Precursor Preparation
A 2.67 wt% SnO 2 solution was obtained by diluting SnO 2 colloid solution with deionized water. Then, the SnO 2 -MXene mix precursor was prepared by adding 0.02 wt% MXene into this SnO 2 solution. The mix precursor was ultrasonically treated over 30 min at room temperature before use so as to acquire more tiny-MXene flakes.

Device Fabrication
The fabrication process for perovskite layer with MDCN ETL is sketched in Fig. 1. The fluorine-doped SnO 2 (FTO) glasses as substrates were ultrasonically cleaned with detergent, deionized water, acetone, isopropanol and ethanol for 20 min in sequence. After the substrates were treated with UV-O 3 for 25 min, the SnO 2 -MXene mix precursor was spin-coated on the substrates at 4000 rpm for 25 s and then was controlled annealed at a low temperature of 150 °C. This controlled annealing process is operated in air for 5 min and then in N 2 atmosphere for 25 min in a glove box, resulting the MDCN ETL. For comparison, the annealing process is just operated in air for entire 30 min, resulting in the pristine SnO 2 ETL.
The perovskite layer with (FAPbI 3 ) 0.97 (MAPbBr 3 ) 0.03 was fabricated on the ETL by a two-step deposition strategy as described in our previous work [21,43]. In belief, 1.3 M PbI 2 in the mixture solvent of DMF/DMSO (9.5:0.5, by volume) was spin-coated onto the ETL at 1500 rpm for 30 s and then annealed at 70 °C for 1 min in N 2 atmosphere in a glove box. After the PbI 2 layer has been cooled down, 50 μL mixture solution of FAI/MABr/MACl (1:0.15:0.25, molar ratio, in 1 mL IPA) was poured on the PbI 2 layer during the spincoating process, and then the substrate was continuously spun at 1400 rpm for 30 s. Then the resultant substrates were transferred into ambient air in 40-60% humidity and were heated on a hot plate at 150 °C for 16 min. After the perovskite layer was cooled to the room temperature, the perovskite crystals with large grain size was generated, which will be discussed in the following sections. Then, the precursor for hole transporting layer (HTL) was coated on the perovskite film at 4000 rpm for 20 s, resulting a HTL. Here, the HTL precursor contained 85.8 mg spiro-OMeTAD, 19.5 μL Li-TFSI (520 mg in 1 mL acetonitrile), 35.5 μL tBP and 8 μL FK209 (400 mg in 1 mL acetonitrile) in 1 mL chlorobenzene. Finally, an Au counter was fabricated on the HTL as top electrode by thermal evaporation in high vacuum.  diffraction (XRD, SmartLab, 9 kW), X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi), and ultraviolet photoelectron spectroscopy (UPS, ESCALAB 250Xi), respectively. The external quantum efficiency (EQE) was measured by an Enli Tech EQE measurement system (Taiwan, China) with a whole PSC device, and light intensity was calibrated by using a standard single crystal Si photovoltaic cell. The samples for SEM, XPS, UV-Vis absorption, and transmission spectra are fabricated on FTO substrates with the dip-coating method described in experimental part. The steady-state photoluminescence (PL, QM400-TM) is measured at excited laser at 470 nm, and the time-

Results and Discussion
The morphology and crystallinity of synthesized pristine MXene materials are verified by SEM, TEM, AFM, and XRD analysis. As shown in SEM image in Fig. S1a, the MXenes show a stacked lamellar structure in cross-sectional view. In the TEM images in Fig. S1b, c, it is clear that the dispersive lamellar structures of MXenes are 2D sheet-like. Then, a representative MXene sheet is measured with AFM. As shown in Fig. S1d and its inset, it is observed that the thickness of 2D MXene sheet is about 2.2 nm, which is close to the 1.5 nm thickness of monolayer MXene in our previous work [31]. As shown in XRD measure in Fig. S2, the main phases at 7.3°, 14.7°, 27.7°, and 60.6° apiece correspond to the (002), (004), (008), and (110) facets, indicating the Ti and Al crystalline phases in MXene [44,45]. Here, the weak peaks at 36.0° and 41.6° correspond to the impurity of TiC, as a residue in MAX phase precursor [32]. These reveal that the synthesized MXene is in lamellar morphology and typical crystallinity.
With SEM measure, the surface morphology of MDCN film is researched, as shown in Fig. 2a. Note that small flakes sparsely cling on the surface, as indicated in the dotted box. Its chiffon-like structure is clearly shown in the enlarged image in the inset. With the targeted region shown in Fig. 2b, the component of MDCN sample is analyzed in EDS. The elemental layered image and the elemental mapping images for Sn, O, C, and Ti are shown in Fig. 2c-g, respectively. These uniform elemental distributions can be attributed to the SnO 2 particles effectively spacing the MXene sheets [42] and indicate that the uniform Ti and C doping from MXene precursor.
To investigate the influence from controlled anneal method in Air&N 2 atmospheres on the crystallinity of MDCN, the test samples of MDCN before and after In order to investigate the valence state transform of material in MDCN compared to pristine SnO 2 counterpart, the XPS measures for individual element are employed. The survey, Ti 2p, Sn 3d, and O1s spectra are shown in Fig. 3a-d, respectively, where the peak positions have been calibrated by the surface carbon signal at 284.6 eV as an internal standard. In survey spectra in Fig. 3a, it is observed that the Sn, O, and C patterns coexist in pristine SnO 2 sample. In contrast, extra Ti 2p and F 1s peaks are noticed for MDCN sample. In addition, the atomic contents of Sn, O, C, Ti, and F have also been obtained as 17.44%, 48.09%, 29.79%, 2.99%, and 1.68%, respectively, by this XPS measure. These indicate that the Ti element in small quantity have been successfully introduced by MXene addition, conforming to the above EDS result. Here, the extra F element can be attributed to the residual LiF corrosion. However, the F elements have not been detected in the above EDS analysis, which may be attributed to its relative tiny amount. Then, the effect from the controlled anneal method on the F content in the MDCN sample is investigated. The MDCN samples before controlled anneal are fabricated with the thickness as same as that for MDCN samples after controlled anneal (i.e., aforementioned MDCN samples). The survey spectra of MDCN before controlled anneal are shown in Fig. S4a, and the extracted F 1s spectra of MDCN samples before and after controlled anneal are exhibited in Fig. S4b to F-C, F-Al, and F-Ti bonds, separately. Compared to the MDCN sample before controlled anneal, it is observed that the position and intensity of these three disintegrated peaks for the MDCN sample after controlled anneal is unchanged, indicating the negligible change of F element by this controlled anneal method. Further, in Ti 2p spectra in Fig. 3b, the MDCN sample shows the coexistence of TiC, Ti 2+ , Ti 3+ , and Ti 4+ bonds, referring to literatures [35,46,47]. In order to investigate the Ti change in the MDCN sample, the XPS spectra of pure MXene are measured, with the survey spectrum shown in Fig. S5a and high-resolution Ti 2p spectrum shown in Fig. S5b. Note that the pure MXene also shows the same Ti-based bonds as those for MDCN sample. The atomic contents of these different Ti-based bonds in these Ti 2p spectra for the two samples are compared in Table S1. It is observed that the Ti 4+ content in MDCN is obviously larger than that in pure MXene, revealing the Ti elements in other bonds, such as Ti 2+ , Ti 3+ bonds, in pure MXene have been converted into the Ti 4+ bond in MDCN. These further demonstrate the existence of TiO 2 crystal in MDCN, confirming to the above XRD results.
In Sn 3d spectra in Fig. 3c, the same peaks at 486.6 and 495.1 eV can be observed in SnO 2 and MDCN samples, which can be attributed to Sn 3d 5/2 and Sn 3d 3/2 , respectively, confirming the successful formation of SnO 2 phase [45]. In O 1s spectra in Fig. 3d, the pristine SnO 2 sample shows three peaks with binding energies of 530.2, 530.8, and 532.0 eV, which can be attributable to Sn-O, OH − radical, and adsorbed oxygen or adsorbed H 2 O, separately [46,48,49]. In contrast, an extra peak is observed for MDCN sample at 529.7 eV, which can be originated from the Ti-O bond [49]. This Ti-O bond can be attributed to the Ti-OH functional group or the Ti-based oxides. Considering the abundant Ti-based bonds of Ti 2+ , Ti 3+ , and Ti 4+ in the above Ti 2p spectrum for MDCN sample, it can be deducted the existence of Ti-based oxides with the Ti-O bond. Therefore, this coexistence of Sn-based oxides with less Ti-based oxides in MDCN sample is beneficial to form a semiconductor oxide heterojunction.
The lattice configurations of MDCN material are further characterized by TEM and HRTEM measures, as shown in Fig. 4. In Fig. 4a, the sample shows a 3D stacked structure. The characteristic regions in left dotted and right dotted boxes in Fig. 4a are enlarged and shown in Fig. 4b, c, respectively. In Fig. 4b, the crystal particles are as small as about 5 nm, inferring the QD phase, and the lattice fringes spacing at d 1 = 0.35 nm and d 2 = 0.24 nm correspond to (101) and (103) planes of anatase TiO 2 , respectively, indicating the 0D anatase TiO 2 QDs [50]. Therefore, it can be demonstrated that the aforementioned Ti-based oxides in XPS measure is the anatase TiO 2 . In Fig. 4c, abundant nanoscale particles are observed. With its enlarged HRTEM analysis in Fig. 4d, the lattice image spacing of d 1 = 0.33 and d 2 = 0.26 nm corresponds to the (110) and (101) planes of rutile SnO 2 phases [51], indicating that those stacked 3D structures in Fig. 4a are constructed by SnO 2 crystals. Therefore, it is evident the intertwining of tiny anatase 0D TiO 2 QDs and 3D SnO 2 particles, constructing TiO 2 /SnO 2 heterojunction structure in the MDCN sample.
On the basis of the above characterizations, the growth mechanism for MDCN is deduced with atomic models sketched in Fig. 5, where the 2D structure of MXene refers literatures [52]. When the MXene is added into the SnO 2 precursor, the 2D MXene are dispersed due to strong van der Waals interaction between adjacent nanosheets. The OH − functional group is generated surrounding 2D MXene sheets in the solution, and the SnO 2 is dispersed around the MXene sheets, as shown in Fig. 5a. In the first anneal procedure in air atmosphere, the rutile SnO 2 crystal enlarges, and the anatase TiO 2 crystal is generated by oxidation effect at the edge of MXene sheet, and then the MXene sheet is torn apart to form defect vacancies at the damaged sizes, as shown in Fig. 5b. In the next anneal procedure in N 2 atmosphere, the SnO 2 and anatase TiO 2 crystals are keeping on grow. However, the generation of new TiO 2 crystal is restricted without outside O element source in N 2 atmosphere. This requisite O element source can only come from the as-formed TiO 2 crystal and OH − functional groups. Therefore, in order to affiliate more stable phase in SnO 2 surroundings, these O elements in as-formed TiO 2 crystal and OH − functional groups transfer into new-formed anatase TiO 2 crystal, through jumping at the defect sites in MXene sheet, as shown in Fig. 5c, forming the incomplete crystals for all anatase TiO 2 because of general lack of O elements. In the subsequent anneal in N 2 , these incomplete crystals tend to assemble with adjacent SnO 2 crystals to form more stable phase, and then nanoscale TiO 2 /SnO 2 heterojunctions are formed. Here, the two anneal procedures in N 2 atmosphere shown in Fig. 5b, c are belonged to the beginning and subsequent procedures of N 2 anneal for 25 min in a glove box described in the above experimental part. Briefly, the nanoscale anatase TiO 2 QDs in 0D, derived from the edge region in 2D MXene sheets, are grown and rooted on surrounding SnO 2 particles in 3D, forming nanoscale TiO 2 / SnO 2 heterojunctions. Herein, the 2D MXene sheets act as conductive bridges bonding these TiO 2 /SnO 2 heterojunctions at edge region of sheet, constructing the final MDCN structure, as shown in Fig. 5d.
The wetting property of ETLs is characterized by comparing the contact angles formed by deionized water droplet. As shown in Fig. S6, the contact angle of MDCN ETL is measured only 20°, which is much smaller than that on SnO 2 ETL (40°), indicating a less hydrophobic surface. This high hydrophilicity offers higher spreading property with polar solvents (such as DMF) on MDCN layer, and then is beneficial to form homogeneous perovskite films. Further, the 3D surface profiles of SnO 2 and MDCN layer on FTO substrates are measured by AFM, as shown in Fig. S7a, b, respectively.  Fig. S8a-f, respectively, and corresponding statistic grain size distributions are drawn in Fig. S8g Fig. 6a, where the functional layers, such as FTO front contact, MDCN ETL, perovskite absorption layer, spiro-OMeTAD HTL, and Au counter electrode, are clearly signed. The SEM image in cross-sectional view for a real device with MDCN-0.02 ETL is shown in Fig. 6b, where the thicknesses of ETL, perovskite layer, HTL, and Au layer are measured as 20, 540, 195, and 80 nm, respectively.
In order to inspect the influence from anneal method of MDCN ETL on the upper perovskite layer, the surface morphology and grain size of resulted perovskite layers are investigated. The samples are also fabricated with a perovskite/ETL/FTO architecture just with the same MDCN-0.02  Fig. 6c-e, respectively, and the corresponding statistic grain size distributions are diagramed in Fig. 6f-h, respectively, and the average grain sizes are obtained as 556, 658, and 397 nm, respectively. Here, Fig. 6d, g is the same as Fig. S8c, i, respectively, for the completely same samples. Note that the MDCN-Air&N 2 still achieves the maximum average grain size, which can be attributed to the rough surface of below ETL, as discussed in Fig. S8. For the MDCN-Air sample, the smaller average grain size can be attributed to the excessive TiO 2 on the surface because of sufficient oxidation in air atmosphere, which will passivate and smooth the ETL surface, and then shrinks its hydrophilicity. In addition, for the MDCN-N 2 sample, the smaller average grain size can be attributed to the lacking TiO 2 on the surface, on which the 2D MXene sheet can smooth the surface, and then shrinks its hydrophilicity.  Fig. S9. It is observed that the MDCN-0.02 ETL shows a better transmittance feature. Further, with different anneal methods, the absorption and transmission spectra of ETLs are further investigated, as shown in Fig. 7a, b, respectively, for the SnO 2 , MDCN-Air, MDCN-Air&N 2 and MDCN-N 2 ETLs. Here, these absorption or transmission spectra are slightly wavy. It is because that a higher power of incident light is set in test for distinguishing the slight differences of samples, by which the irregular background signals are also magnified. However, in the same test conditions, it is not affected to analyze the intensity difference of test samples. In Fig. 7a, in the UV light region ranging from 200 to 380 nm, the light absorption intensities of MDCN-Air&N 2 and MDCN-Air are approximate, and they are larger than those of SnO 2 sample, confirming the inactive UV photocatalysis of SnO 2 material [13]. In the visible light region ranging from 380 to 780 nm wavelength, the light absorption and transmission properties of these samples are different, which can be attributed to the introduced MXene material and the resultant film nanostructure. When the MXene is introduced into the SnO 2 materials, on the one hand, the MXenes as barriers in MDCN can increase the light absorption and then decrease the light transmission. On the other hand, the relatively rough surface of MDCN sample in comparison to SnO 2 sample, as discussed in the above AFM measured in Fig. S7, can reduce the light reflection and then enlarge the light transmission, based on light scattering effect in its suede-like rough surface. Then, it can be concluded that the MXene dopants offer a harmonious optical influence composed with two interinhibitive effects of light-barrier and anti-reflection. Here, compared to SnO 2 sample, the MDCN-N 2 sample shows the enhanced absorption and declined transmission. This can be attributed to the dominant light-barrier effect of MXene dopant. Then, compared to this MDCN-N 2 sample, the MDCN-Air sample shows the declined absorption and declined transmission. This can be attributed to the extra introduced excessive TiO 2 additions, which concentrate near the surface of SnO 2 layer because of oxidation from surrounding air, as discussed in aforementioned section. Generally, the TiO 2 as extra lightbarrier will further enhance absorption and decline transmission. However, the declined absorption here can be ascribed to the higher reflection on surface, because of the excessive TiO 2 additions can passivate the antireflection effect of MXene. In addition, compared to the MDCN-Air sample or even the SnO 2 sample, the MDCN-Air&N 2 sample shows the declined absorption and enhanced transmission, which can be ascribed to the dominant antireflection effect of MXene dopants. Therefore, integrating the above results, Absorbance (a.u.)  Table S2. In general, the V OC , J SC , FF, and PCE values all increase and then decrease with increasing MXene content. Here, the MDCN-0.02 ETL-based PSC achieves the highest average PCE of 18.44%. In contrast, the SnO 2 -based counterpart only exhibits an average PCE of 16.42%. These devices are further investigated by EIS measure, with Nyquist plots shown in Fig. S10 b with equivalent circuit in inset. The series resistance (R s ) mainly reflects electrical contacts, wires, and sheet resistance of the electrodes [43,53]. The charge-transfer resistance (R tr ) means the charge-transfer property at the interface, and R rec means the recombination of carriers [43,54]. Fitted with ZView software, it can be observed that the R s values are similar for these samples. With increasing MXene content, the R tr value decreases and then increases, revealing that the charge-transfer speed at perovskite/ETL interface increases and then decreases, and the minimum R tr value of MDCN-0.02-based sample means the fastest charge transfer, which can be used to explain the almost same change regularity of FF for corresponding devices, as listed in Table S2.
The J − V performance of PSCs based on MDCN-0.02 ETL with different anneal methods is further investigated. The typical J − V curves are shown in Fig. 7c with photovoltaic parameters listed in Table 1. The J − V curve and photovoltaic parameters for the MDCN-Air&N 2 sample are the same as those of MDCN-0.02-based sample described in Fig. S10a and Table S2, because of the completely same sample. It is observed that the device based on MDCN-Air&N 2 ETL still shows the highest PCE than others. As shown in Fig. S11, the histogram of photovoltaic parameters for SnO 2 and MDCN-Air&N 2 -based PSCs further exhibits the repeatability of high performance of MDCN-Air&N 2based devices. The performance is further analyzed by EIS measure, and their Nyquist plots are shown in Fig. 7d with equivalent circuit in inset. Here, the plot for MDCN-Air&N 2 sample is the same to MDCN-0.02-based sample described in Fig. S10b because of the completely same sample. Among these samples, their R s values are similar, and hence, the main contribution for PCE can be attributed to their different R tr values. In addition, it is observed that the FF plays an important role on the PCE, therefore, the difference in FF can be explained by the R tr regularity. Here, the chargetransfer speed in device, reversely denoted by the R tr value, can come from two types of interfaces. One type is the perovskite/MXene interface between perovskite layer and ETL, and another type is the nanoscale interface of TiO 2 /SnO 2 heterojunction inside the ETL. Compared to SnO 2 -based sample, the smaller R tr value of MDCN-N 2 -based sample can be attributed to the faster charge transfer at perovskite/ MXene interface, because TiO 2 /SnO 2 heterojunctions cannot be formed without air atmosphere. Compared to MDCN-N 2 -based sample, the smaller R tr value of MDCN-Air-based sample can be attributed to the faster charge transfer through TiO 2 /SnO 2 heterojunction, because the sufficient heterojunctions generated by oxidation in air atmosphere near the ETL surface, which offer more contribution between perovskite layer and ETL. Further, compared to MDCN-Air-based sample, the smaller R tr value of MDCN-Air&N 2 -based sample can be attributed to the fastest charge transfer through these two types of interfaces, because the optimal MDCN structure generated in these regions. In addition, it is observed that the FF or PCE parameter generally shows the negative relationship with the R tr value, indicating that the main influence on FF or PCE is the charge-transfer speed at interfaces. However, the FF or PCE parameter of MDCN-N 2based sample is not suitable for this regularity. Compared to SnO 2 -based sample, the FF or PCE parameter of MDCN-N 2based sample is smaller, although it owns the faster charge transfer as discussed above. This can be attributed to the defects of MXene inside the ETL, which capture and then recombine the traveling charges, and hence weaken the contribution from the faster charge-transfer speed. The steady-state PL spectra for SnO 2 and MDCN-Air&N 2 samples are measured and shown in Fig. 8a. The test samples are prepared with a device architecture of perovskite/ ETL/FTO with the same experimental procedures described above. Note that these two samples both show strong peaks around 790 nm, and the peak of MDCN-Air&N 2 is weaker than that of SnO 2 sample, suggesting stronger electron quenching and transport ability of MDCN-Air&N 2 ETL [21].
The charge recombination kinetics of ETL is further investigated with the TRPL decay measurement, with a test device architecture as same as that for steady-state PL. The PL decay time, Y(t), is fitted using a bi-exponential equation (Eq. 1) [43]: The average PL decay time (τ ave ) is calculated by the τ i and A i values of fitted curve data with Eq. 2: The fast decay lifetime (τ 1 ) is attributed to the trap-assisted recombination by quenching of photo-generated free carriers transported from the perovskite layer to the ETL, and the slow decay lifetime (τ 2 ) originates from the free carrier recombination before charge collection [55,56]. With the SnO 2 and MDCN-Air&N 2 ETLs, the TRPL spectra are shown in Fig. 8b and detailed fitting parameters are listed in Table 2. Compared to SnO 2 sample, the MDCN-Air&N 2 sample shows the similar τ 1 value, and shortened τ 2 and τ ave values. The shortened τ 2 and τ ave values demonstrate more efficient electron extraction at the perovskite/MDCN-Air&N 2 interface [55,57]. This can be attributed to reduction of defect density and formation of good interface at perovskite/MDCN-Air&N 2 interface. These results are still in agreement with the above steady-state PL results. Further, the TPV and TPC measures are conducted to study the charge recombination and extraction properties of devices. The TPV is used to measure charge recombination under the V OC condition, and the decaying progress of V OC means that the recombination occurring of charges. As plotted in Fig. 8c in TPV curve, the V OC for MDCN-Air&N 2 -based device shows the slower decay, compared that for SnO 2 -based device, suggesting that the reduced charge recombination occurring within the device based on MDCN-Air&N 2 ETL [58]. As plotted in Fig. 8d in TPC curve, the short-circuit current for MDCN-Air&N 2based device shows the faster decay, compared to that for SnO 2 -based device, implying the enhanced charge transfer, reduced interfacial defects, and better crystal of perovskite layer in MDCN-Air&N 2 -based device [58].
The electron mobility (μ) is an important parameter to characterize semiconductive properties [59][60][61]. It can be obtained by SCLC measure with Eq. 3: where J is the current density, ε 0 the vacuum permittivity of 8.85 × 10 −12 F m −1 , ε r the relative dielectric constant of material (13 for SnO 2 ) [62], V the applied bias, and d the ETL thickness. Then, the SCLC measures of SnO 2 and MDCN-Air&N 2 -based electron only devices are carried out, with the preparation according to the literature [57], and the resultant curves are shown in Fig. S12. It can be calculated that the μ value of SnO 2 is 8.29 × 10 −5 cm 2 V −1 s −1 , conforming to literature reports [57]. In contrast, the MDCN-Air&N 2 layer achieves the higher μ value of 4.06 × 10 −4 cm 2 V −1 s −1 , indicating the lower defects in materials. The energy level of MDCN-Air&N 2 and SnO 2 is measured with UPS. The test samples are fabricated by spincoating the precursors on FTO substrates. As shown in the middle plot in Fig. 8e, the left and right intersection regions relative to baseline in dotted boxes are enlarged, respectively. Then, the cutoff binding energy (E CUT ) or Fermi edge (E D ) can be determined by the left or right intersection of baseline with the tangent line, in the left or right enlarged plots. Then, the energy levels at valence band maximum (E VBM ) and conduction band minimum (E CBM ) can be obtained with E VBM = E F − E D (eV) and E CBM = E VBM + E G (eV). Here, the Fermi level (E F ) can be obtained by E F = E CUT − E IP , where the E IP = 21.2 eV is the energy of incident photon in this test, and the E G value is obtained from the UV-Vis absorption spectrum. As shown in Fig. S13, the absorption onset (λ onset ) values for SnO 2 and MDCN-Air&N 2 samples are determined and signed by their absorption edges, based on the UV-Vis absorption spectra in Fig. 7a. It is observed that the λ onset values for both samples are the same as 365 nm, deducing their same E G values of 3.48 eV, based on the E G = 1240/λ onset . On the basis of the above calculations, the energy-level parameters of λ onset , E G , E CBM , and E VBM values are concluded and listed in Table 3. It is observed that the E CBM and E VBM values both downshift 0.21 eV.
The schematic energy-level diagram for each layer in working PSC is plotted in Fig. 8f, based on the above energy-level parameters, in which the SnO 2 and MDCN-Air&N 2 are alternative ETLs signed in a dotted box. The energy levels of FTO, (CH 3 NH 3 )PbI 3 , Spiro-OMeTAD refer to our previous work [21], and that of Au is from the literature [63]. In comparison with the SnO 2 ETL, the downshifted E VBM of MDCN-Air&N 2 ETL leads to the higher ability for suppressing the hole recombination toward the FTO layer and then contributes to the higher PCE, although the downshifted E CBM means the stronger ability for extracting electron from perovskite layer. Nevertheless, the downshifted E CBM also means more recombination probability for the electrons at the conduction band of ETL, relative to the holes at the valence band of perovskite layer, which will weaken the improvement of PCE of devices.
The device performances are further demonstrated by investigating their champion PCE. As shown in J − V curves in Fig. 9a, for MDCN-Air&N 2 -based sample, the champion PCE is 19.14%, with J SC of 24.16 mA cm −2 , V OC of 1.07 V and FF of 74.05%. In contrast, the champion PCE for SnO 2 -based sample is only 16.83%, with J SC of 21.88 mA cm −2 , V OC of 1.07 V, and FF of 71.90%. The EQE of MDCN-Air&N 2 -based device is measured. As shown in Fig. 9b, the spectrum expresses a wide photo-response from 350 to 800 nm with a peak of ≈ 92% at ≈ 510 nm. The integrated current density from the EQE is obtained as 22.68 mA cm −2 , which is almost in agreement with the measure from the solar simulator in this research.
In order to correctly evaluate device performance, the hysteresis characteristics of typical cells with SnO 2 and MDCN-Air&N 2 ETLs are measured with a single J − V loop, as shown in Fig. 9c, where the symbol of (R) or (F) indicates the reverse or forward scan method, respectively, with the V OC , J SC , FF, and PCE listed in Table 4. The hysteresis index HI = (PCE(R) − PCE(F))/PCE(R) of SnO 2 and MDCN-Air&N 2 device is calculated as 0.10 and 0.03. The far less HI of MDCN-Air&N 2 -based device further reveals the electron extraction is effectively accelerated with the MDCN structure.
Moreover, the moisture-resistance ability of PSC is evaluated by testing the devices without encapsulation in 30-40% humidity air at room temperature. As shown in Fig. 9d, the device with MDCN-Air&N 2 ETL exhibits high stability retaining almost 85% of initial PCE, even after 45 days. In contrast, the device with SnO 2 ETL degrades below 75% of its initial performance in the same test environment. This represents the superior moisture-resistance ability of the

Conclusion
With a controlled anneal method implemented in air and then in N 2 atmospheres, a film in MDCN structure is successfully designed and fabricated, by means of adopting the Ti 3 C 2 T X MXene into SnO 2 precursor. In the resulted MDCN, the 0D anatase TiO 2 quantum dots, surrounding on 2D conducting Ti 3 C 2 T X sheets, are in situ rooted on 3D SnO 2 nanoparticles, constructing nanoscale TiO 2 /SnO 2 heterojunctions, with an oxygen vacancy scramble effect. With the optimized MDCN-0.02 ETL, the optical transmittance of ETL is enhanced, the interface impedance is decreased, and the crystallinity of upper perovskite layer is enlarged, achieving more amount carrier and the rapider carrier transfer. The champion PCE of resulted PSC achieves 19.14%, yet that of counterpart is just 16.83%. It can also maintain almost 85% of its initial performance for more than 45 days in 30-40% humidity air, and the counterpart comparatively degrades below 75% of its initial performance.