The electronic structure of metal oxide/organo metal halide perovskite junctions in perovskite based solar cells

Cross-sections of a hole-conductor-free CH3NH3PbI3 perovskite solar cell were characterized with Kelvin probe force microscopy. A depletion region width of about 45 nm was determined from the measured potential profiles at the interface between CH3NH3PbI3 and nanocrystalline TiO2, whereas a negligible depletion was measured at the CH3NH3PbI3/Al2O3 interface. A complete solar cell can be realized with the CH3NH3PbI3 that functions both as light harvester and hole conductor in combination with a metal oxide. The band diagrams were estimated from the measured potential profile at the interfaces, and are critical findings for a better understanding and further improvement of perovskite based solar cells.

of these solar cells. Perovskite solar cell cross-sections were prepared in a nitrogen glovebox by cleavage. Capillary forces were minimized allowing a small tip-surface distance hence better lateral resolution compared to an AFM setup in ambient. A typical structure of the hole-conductor-free perovskite solar cell is shown in the high resolution scanning electron microscopy (HR-SEM) micrograph (Fig. 1b). The MAPbI 3 perovskite is deposited on top of the metal oxide using the two-step deposition process described earlier 6,20 . Table 1 and figure 1c show the photovoltaic parameters and the current voltage curves of the hole-conductor-free MAPbI 3 (CH 3 NH 3 5 MA) solar cells using nc-TiO 2 or nc-Al 2 O 3 as the metal oxide. Figure 1d presents the incident photon to current efficiency (IPCE) of the corresponding cells. For both cells, the response covers the whole visible range. However, for the nc-TiO 2 /MAPbI 3 cell, the IPCE reaches 75%, while for the nc-Al 2 O 3 /MAPbI 3 cell, 35% IPCE is achieved in good agreement with the current density-voltage (J-V) characteristics measured under a solar simulator. Upon visible-light excitation of the TiO 2 /MAPbI 3 /Au cell, electrons and holes are generated inside the perovskite layer. The electrons are then injected into the conduction band of the TiO 2 and driven towards the FTO, whereas holes are transported through the perovskite layer to the Au contact.
For the Al 2 O 3 /MAPbI 3 /Au configuration, injection of electrons into the nanoporous Al 2 O 3 is energetically not allowed; therefore, the photogenerated charge carriers are transported through the MAPbI 3 film to the appropriate contacts.
The KPFM cross-section images of the two different cells in dark, nc-TiO 2 /MAPbI 3 and the nc-Al 2 O 3 /MAPbI 3 , are presented in Figure 2a and 2b. The high spatial resolution was achieved by maintaining the tip-sample distance below 5 nm at a slow scan rate of 0.1 Hz. The perovskite penetrates through the nc-Al 2 O 3 , sintered to the grounded contact; consequently, the Fermi level is assumed to be aligned and equal in all three layers. This allows a quantitative interpretation of the measured potential for both solar cell structures. Figure 2c and 2d show the CPD statistical distribution for the two interfaces; A CPD variation of 90 mV was determined for both metal oxides and a CPD variation of 65 mV was found for the perovskite layer. These variations are mainly due to the nanostructure and inhomogeneous surface of the measured materials entail a high amount of defect states. The measured average potential difference, DCPD, at the interface between nc-TiO 2 and MAPbI 3 , as well as between the nc-Al 2 O 3 and MAPbI 3 interface is 0.155 V and 0.12 V, respectively. Although the nc-TiO 2 /MAPbI 3 based solar cell has a higher efficiency, the potential difference between the two interfaces in the dark (35 mV) is negligible small. Some of the perovskite grains have a lower CPD and are expected to be less active in the solar cell, as previously observed with EBIC measurements 15 .   The CPD is higher (more positive) at the grain boundaries between adjacent perovskite crystals by ca. 25 mV (see figure 3b), implying a minor effect on the perovskite solar cell performance as this potential barrier is around the thermal voltage at room temperature. Figure 3a shows a schematic band diagram for the case of positive charge accumulation (hole traps) at the grain boundary. It is possible to calculate the width of the barrier potential at the grain boundaries using known barrier models 21,22 provided the grain boundary is of negligible width compared to the grain size. This model was used to analyze grain boundaries in p-type polycrystalline silicon, II-VI semiconductors and chalcopyrite materials.
The electrostatic screening length, L S (within Ls charges are electrostatically screened; beyond this length charges are not affected by other charges) is extracted by fitting the measured potential profile across the GB ( Fig. 3b) with an exponential function and is approximately 6.5 nm, which is much smaller than the average grain size of the perovskite where e 5 e 0 e p (e p < 20 23 is the dielectric constant of the perovskite and e 0 is the permittivity), k b is the Boltzmann constant, T is the temperature and q is the elementary charge. By using extracted L S in equation 1, the dopant density (N d ) was estimated to be 7 3 10 17 cm 23 , which is in a good agreement with reported values observed by Guerrero et al. 24 The CPD profiles of the metal oxide/MAPbI 3 interfaces, shown in figures 4a-d, provide estimation for the junction depletion width. Figure 4c and 4d show the CPD and the electric field, E 5 dCPD/dx, across the interface for the nc-TiO 2 /MAPbI 3 and the nc-Al 2 O 3 / MAPbI 3 cell structure, respectively. Accordingly, the depletion width for the nc-TiO 2 /MAPbI 3 heterojunction is <45 nm with a maximum electric field of E < 9?10 4 V/cm. In the case of the nc-Al 2 O 3 /MAPbI 3 interface, there is no electron injection from the MAPbI 3 into the nc-Al 2 O 3 , since it functions rather as a scaffold. The electric field profile for the nc-Al 2 O 3 /MAPbI 3 interface is narrower, showing a depletion (space charge region) width of around W p < 10 nm in the perovskite side of the junction (figure 4d). This corresponds to the depletion region width at the perovskite side at the nc-TiO 2 /MAPbI 3 interface.      Figures 5a and 5b show the suggested energy band diagrams for the nc-TiO 2 /MAPbI 3 and the nc-Al 2 O 3 /MAPbI 3 interfaces using the measured work functions in dark. As opposed to the nc-Al 2 O 3 , the nc-TiO 2 has an active role in the electron transport. For the nc-TiO 2 / MAPbI 3 interface (Fig. 5a) electron transport takes place mainly through the nc-TiO 2 since the electron injection is favorable. The measured space charge region at this interface assists in the charge separation and inhibits recombination which assist to the PV performance.
For the nc-Al 2 O 3 /MAPbI 3 interface, an oxide-semiconductor band structure is proposed with a relatively small band bending at the perovskite side. Since the 80 nm thin Al 2 O 3 layer is nanoporous, the perovskite penetrates through the Al 2 O 3 enabling electron transport to the contacts. We note that experiments with an Al 2 O 3 layer thickness .80 nm drastically reduced the solar cell performance.
The long diffusion length assists in the charges transport to the corresponding contacts. Table 2 summarizes the parameters obtained in this study. The carrier density of the perovskite was calculated using the electronic barrier model; the depletion region width and the electric field were estimated from the cross section KPFM measurements.

Conclusions
Cross-sections of hole-conductor-free perovskite solar cells were measured using KPFM. The measured potential differences between the nanoporous metal oxide and the perovskite in dark are consistent for both nc-TiO 2 (DCPD 5 0.155 V) and nc-Al 2 O 3 (DCPD 5 0.12 V) based cells and were used to estimate the band diagrams. Moreover, the measured width (w p 5 10 nm) of the depletion region at the perovskite side is the same. However, while the measured band bending at the nc-Al 2 O 3 /MAPbI 3 side is negligible small, we measured a 45 nm depletion region width at the TiO 2 side of the nc-TiO 2 /MAPbI 3 interface. This depletion region contributes to the separation of photogenerated charge carriers and thus a higher performance of the TiO 2 based cells. The measured potential difference of 25 mV at the perovskite grain boundaries reveals hole accumulation at the GBs and plays a minor role for the solar cell performance. The results shed additional light on the electronic structure of these highly efficient perovskite based solar cells.

Experimental
Material synthesis. TiO 2 paste DSL 90-T composed of 20 nm particles was purchased from DYESOL. The TiO 2 paste was diluted with ethanol in ratio of 154 by weight and deposited by spin coating at 2000 r.p.m. for 10 sec. The TiO 2 film was annealed at 500uC for 30 min. The nc-Al 2 O 3 20 wt.% in isopropanol (,50 nm particle size) was purchased from Sigma-Aldrich. The nc-Al 2 O 3 was diluted in isopropanol in ratio of 1512 by weight. The deposition and the annealing conditions were the same as for TiO 2 . CH 3 NH 3 I was synthesized as described previously 25 , by reacting 30 mL of methylamine (40% in methanol, TCI) and 32.3 mL of hydroiodic acid (57 wt% in water, Aldrich) in a 250 mL round bottom flask at 0uC for 2 h with stirring. The precipitate was recovered by putting the solution on a rotavap and carefully removing the solvents at 50uC. The yellowish raw product of methylammonium iodide (CH 3 NH 3 I) was washed with ethanol by stirring the mixture for 30 min. Then the mixture was filtered and washed three times with diethylether. After filtration, the solid was collected and dried at 70uC in a vacuum oven for 24 h.
Device fabrication. The substrate of the device was a SNO 2 5F (FTO) conducting glass (15 V?cm 21 ), Pilkington). A blocking layer was deposited on the FTO glass using a solution of titanium diisopropoxidebis(acetylacetonate) (TiDIP, 75% in isopropanol, Aldrich) in ethanol. The TiDIP solution was spin coated and then annealed at 450uC for 35 min. The TiO 2 solution or the Al 2 O 3 solution were spin coated and annealed at 500uC for 30 min subsequent to TiCl 4 treatment for 30 min at 70uC and annealing at 500uC for 30 min.
The synthesis of the CH 3 NH 3 PbI 3 on the TiO 2 surface was carried out by a twostep deposition technique.
First, PbI 2 was dissolved in DMF and dropped onto the TiO 2 film and spin coated, followed by annealing at 70uC for 30 min. In the second step, the cell was dipped into methylammonium solution. Following the dipping step, the samples were annealed at 70uC for another 30 min. Finally, the back contact was deposited by evaporating 50 nm of gold under pressure of 5*10 26 Torr. The active area was 0.09 cm 2 .
The KPFM cross sections samples were prepared on a crystalline silicon substrate. The Si wafer was etched from native oxide by hydrofluoric acid (48 wt% in water, Aldrich) for 15 min and cleaned by oxygen plasma for 1 min.
Photovoltaic characterization. Photovoltaic measurements were made on a New Port system, composed of an Oriel I-V test station using an Oriel Sol3A simulator. The solar simulator is class AAA for spectral performance, uniformity of irradiance, and temporal stability. The solar simulator is equipped with a 450 W xenon lamp. The output power is adjusted to match AM1.5 global sunlight (100 mWcm 22 ). The spectral match classifications are IEC60904-9 2007, JIC C 8912, and ASTM E927-05. I-V curves were obtained by applying an external bias to the cell and measuring the generated photocurrent with a Keithley model 2400 digital source meter. The voltage step and delay time of photocurrent were 10 mV and 40 ms, respectively. Oriel IQE-200 was used to determine the monochromatic incident photon-to-electric current conversion efficiency. Under full computer control, light from a 150 W xenon arc lamp was focused through a monochromator in the 300-1800 nm wavelength range onto the photovoltaic cell under test. The monochromator was incremented through the visible spectrum to generate the IPCE (l) as defined by IPCE (l) 5 12,400 (Jsc/l Q), where l is the wavelength, Jsc is the short-circuit photocurrent density (mA cm 22 ), and Q is the incident radiative flux (mWcm 22 ). Photovoltaic performance was measured by using a metal mask with an aperture area of 0.09 cm 2 .
High Resolution Scanning Electron Microscopy (HR-SEM). SIRIONHR SEM was performed with FEI (Field Emission Instruments), The Netherlands. The measurement conditions were 5 kV at a magnification of 39,000.
Kelvin probe force microscopy (KPFM). Various locations of the cleaved crosssections were measured by KPFM and consistent CPD profiles with same variations were obtained for each solar cell type. The measured CPD was constant over a day range. Topography and CPD were measured with a slow scan rate (0.1 Hz) and in the attractive force regime (force-distance curve) in order to avoid tip-sample crashes. Amplitude modulation KPFM was carried out with a commercial AFM (Dimension Edge, Bruker Inc.) inside a nitrogen glove box with less than 1 ppm H 2 O at room temperature. The CPD was measured simultaneously with the topographic signal at an effective tip sample distance of 5 nm during scanning. The topographic height was obtained by maintaining the amplitude of the first cantilever resonance (f 1st < 75 kHz) at a predefined amplitude set point of approximately 10 nm. The CPD was determined by compensating the ac component of the electrostatic force at angular frequency v with an applied dc voltage (5jCPDj) in a feedback control loop. To separate topographic from CPD signal, increase the sensitivity, and minimize probesample convolution effects, the AC electrostatic force component was generated at the second resonance 26 , f 2nd < 450 kHz, of the cantilever by applying an ac voltage of about 500 mV. Highly conductive cantilevers with Pt/Ir coating (PPP EFM, Nanosensors) were used for KPFM.