Efficient Hybrid Solar Cells Based on Meso-Superstructured Organometal Halide Perovskites

See allHide authors and affiliations

Science  02 Nov 2012:
Vol. 338, Issue 6107, pp. 643-647
DOI: 10.1126/science.1228604


The energy costs associated with separating tightly bound excitons (photoinduced electron-hole pairs) and extracting free charges from highly disordered low-mobility networks represent fundamental losses for many low-cost photovoltaic technologies. We report a low-cost, solution-processable solar cell, based on a highly crystalline perovskite absorber with intense visible to near-infrared absorptivity, that has a power conversion efficiency of 10.9% in a single-junction device under simulated full sunlight. This “meso-superstructured solar cell” exhibits exceptionally few fundamental energy losses; it can generate open-circuit photovoltages of more than 1.1 volts, despite the relatively narrow absorber band gap of 1.55 electron volts. The functionality arises from the use of mesoporous alumina as an inert scaffold that structures the absorber and forces electrons to reside in and be transported through the perovskite.

An efficient solar cell must absorb over a broad spectral range, from visible to near-infrared (near-IR) wavelengths (350 to ~950 nm), and convert the incident light effectively into charges. The charges must be collected at a high voltage with suitable current in order to do useful work (18). A simple measure of solar cell effectiveness at generating voltage is the difference in energy between the optical band gap of the absorber and the open-circuit voltage (Voc) generated by the solar cell under simulated air mass (AM) 1.5 solar illumination of 100 mW cm−2 (9). For instance, gallium arsenide (GaAs) solar cells exhibit Voc of 1.11 V and an optical band gap of 1.4 eV, giving a difference of ~0.29 eV (2). For dye-sensitized and organic solar cells, this difference is usually on the order of 0.7 to 0.8 eV (2, 9). For organic solar cells, such losses are predominantly caused by their low dielectric constants. Tightly bound excitons form, which require a heterojunction with an electron acceptor with a large energy offset to enable ionization and charge separation (10, 11). Likewise, dye-sensitized solar cells (DSSCs) have losses, both from electron transfer from the dye (or absorber) into the TiO2, which requires a certain “driving force,” and from dye regeneration from the electrolyte, which requires an overpotential. Efforts have been made to reduce such losses in DSSCs by moving from a multielectron iodide–tri-iodide redox couple to one-electron outer-sphere redox couples, such as cobalt complexes or a solid-state hole conductor (1, 4, 12, 13).

Inorganic semiconductor–sensitized solar cells have recently become a focus of interest (14, 15). An extremely thin absorber (ETA) layer, 2 to 10 nm in thickness, is coated upon the internal surface of a mesoporous TiO2 electrode and then contacted with an electrolyte or solid-state hole conductor. These devices have achieved power conversion efficiencies of up to 6.3% (15). However, the ETA concept suffers from rather low Voc; the problem may lie in the electronically disordered, low-mobility n-type TiO2 (16). Perovskites are relatively underexplored alternatives (Fig. 1A) that provide a framework for binding organic and inorganic components into a molecular composite. With careful consideration of the interaction between organic and inorganic elements and suitable control of the size-tunable crystal cell (17), rudimentary wet chemistry can be used to create new and interesting materials. Era, Mitzi, and co-workers have shown that layered perovskites based on organometal halides demonstrate excellent performance as light-emitting diodes (18, 19) and transistors with mobilities comparable to amorphous silicon (20). Organometal halide perovskites have been used as sensitizers in liquid electrolyte–based photoelectrochemical cells with conversion efficiencies from 3.5 to 6.5% (21, 22). Recently, a CsSnI3 perovskite was shown to function efficiently as a hole conductor in solid-state DSSCs, delivering up to 8.5% power conversion efficiency (23, 24).

Fig. 1

(A) Left: Three-dimensional schematic representation of perovskite structure ABX3 (A = CH3NH3, B = Pb, and X = Cl, I). Right: Two-dimensional schematic illustrating the perovskite unit cell. (B) Ultraviolet to visible (UV-Vis) absorbance spectra of the photoactive layer in the solar cell (mesoporous oxide; perovskite absorber; spiro-OMeTAD) sealed between two sheets of glass in nitrogen and exposed to simulated AM1.5 sunlight at 100 mW cm−2 irradiance for up to 1000 hours. No additional UV filtration was used for the solar irradiance. Inset: Extracted optical density at 500 nm as a function of time. (C) Left: Schematic representation of full device structure, where the mesoporous oxide is either Al2O3 or anatase TiO2. Right: Cross-sectional SEM image of a full device incorporating mesoporous Al2O3. Scale bar, 500 nm.

We report on a solution-processable solar cell that overcomes the fundamental losses of organic absorbers and disordered metal oxides. We followed the ETA approach and used a perovskite absorber, mesoporous TiO2 as the transparent n-type component, and 2,2′,7,7′-tetrakis-(N,N-di-p-methoxyphenylamine)9,9′-spirobifluorene (spiro-OMeTAD) as the transparent p-type hole conductor. These devices exhibited power conversion efficiencies near 8%. Remarkably, we also found that replacement of the mesoporous n-type TiO2 with insulating Al2O3 improved the power conversion efficiency. The Al2O3 is an insulator with a wide band gap (7 to 9 eV) and purely acts as a “scaffold” upon which the perovskite is coated. We observed that electron transport through the perovskite layer was much faster than through the n-type TiO2. In addition, we observed an increase in Voc (moving from the TiO2 to the insulating Al2O3 scaffold) of a few hundred millivolts and a power conversion efficiency of 10.9% under simulated AM1.5 full solar illumination.

The specific perovskite we used is of mixed-halide form: methylammonium lead iodide chloride (CH3NH3PbI2Cl), which was processed from a precursor solution in N,N-dimethylformamide via spin-coating in ambient conditions. X-ray diffraction analysis for CH3NH3PbI2Cl prepared on glass (fig. S1) (25) showed diffraction peaks at 14.20°, 28.58°, and 43.27°, which we assigned as the (110), (220), and (330) planes, respectively, of a tetragonal perovskite structure with lattice parameters a = 8.825 Å, b = 8.835 Å, c = 11.24 Å, similar to the CH3NH3PbI3 previously reported (21). The extremely narrow diffraction peaks suggest that the films have long-range crystalline domains (>200 nm, peak width limited by instrument broadening) and are highly oriented with the a axis (21, 26). In contrast to the methylammonium trihalogen plumbates previously reported in solar cells (i.e., CH3NH3PbI3) (21, 22), this iodide-chloride mixed-halide perovskite was remarkably stable to processing in air. The absorption spectra (Fig. 1B) demonstrated good light-harvesting capabilities over the visible to near-IR spectrum and was also stable to prolonged light exposure, as demonstrated by 1000 hours of constant illumination under simulated full sunlight. The absorbance of the film at 500 nm remained around 1.8 throughout the entire measurement period (absorbance of 1.8 corresponds to 98.4% absorption) (Fig. 1B, inset). Note that the scale is optical density, where absorbance of ~0.5 at 700 nm corresponds to ~70% attenuation in a single pass; in the solar cell, there are two passes of light leading to ~91% absorption at this wavelength.

The solar cells were fabricated on semitransparent fluorine-doped tin oxide (FTO)–coated glass coated with a compact layer of TiO2 that acted as an anode. The porous oxide films were fabricated from sol-gel–processed sintered nanoparticles. The perovskite precursor solution was infiltrated into the porous oxide mesostructure via spin-coating and was dried at 100°C, which enabled the perovskite to form via self-assembly of the constituent ions. Dark coloration was observed only after this drying step.

With respect to the perovskite coating process, there has been extensive work done on investigating how solution-cast materials infiltrate into mesoporous oxides (2732). If the concentration of the solution is low enough and the solubility of the cast material high enough, the material will completely penetrate the pores as the solvent evaporates. Typically, the material forms a “wetting” layer upon the internal surface of the mesoporous film that uniformly coats the pore walls throughout the thickness of the electrode (2831). The degree of “pore filling” can be controlled by varying the solution concentration (2932). If the concentration of the casting solution is high, then maximum pore filling occurs, and any “excess” material forms a “capping layer” on top of the filled mesoporous oxide.

For the optimum perovskite precursor concentrations we used, there was no appearance of a capping layer, which implies that the perovskite was predominantly formed within the mesoporous film. We verified that the perovskite was within and uniformly distributed throughout the mesoporous oxide films by performing cross-sectional scanning electron microscopy (SEM) with elemental mapping via energy-dispersive x-ray (EDX) analysis (fig. S2) (25). To complete the photoactive layer, the perovskite-coated porous electrode was further filled with the hole transporter, spiro-OMeTAD, via spin-coating; as shown in Fig. 1C, the spiro-OMeTAD forms a capping layer that ensures selective collection of holes at the silver electrode.

In Fig. 2A, the incident photon-to-electron conversion efficiency (IPCE) action spectrum is shown for the devices that use mesoporous TiO2 and Al2O3, exhibiting spectral sensitivity spanning from the visible to the near-IR (400 to 800 nm) with a peak IPCE of >80% for both oxides. The slight difference in shape arises from the slightly different perovskite concentrations in the optimized devices. In Fig. 2B, we show current density–voltage (J-V) curves measured under simulated AM1.5 illumination of 100 mW cm−2. The sensitized TiO2 solar cell exhibited a short-circuit photocurrent (Jsc) = 17.8 mA cm−2, Voc = 0.80 V, and a fill factor of 0.53, yielding an overall power conversion efficiency (η) of 7.6%. We present two different J-V curves for the Al2O3-based device. The most efficient device exhibited Jsc = 17.8 mA cm−2, Voc = 0.98 V, and a fill factor of 0.63, yielding η = 10.9%. The third curve (dashed trace) shows a device with Jsc = 15.4 mA cm−2 and Voc = 1.13 V but a low fill factor of 0.45, yielding η = 7.8%. [See (25) for histograms of device performance parameters for the TiO2- and Al2O3-based devices (fig. S3)].

Fig. 2

(A) IPCE action spectrum of an Al2O3-based and perovskite-sensitized TiO2 solar cell, with device structure as follows: FTO / compact TiO2 / mesoporous Al2O3 (red trace with crosses) or mesoporous TiO2 (black trace with circles) / CH3NH3PbI2Cl / spiro-OMeTAD / Ag. (B) Current density–voltage characteristics under simulated AM1.5 100 mW cm−2 illumination for Al2O3-based cells, one cell exhibiting high efficiency (red solid trace with crosses) and one exhibiting VOC > 1.1 V (red dashed line with crosses); for a perovskite-sensitized TiO2 solar cell (black trace with circles); and for a planar-junction diode with structure FTO / compact TiO2 / CH3NH3PbI2Cl / spiro-OMeTAD / Ag (purple trace with squares).

The general trend is that the Al2O3 cells generated open-circuit voltages that were >200 mV higher than those generated by the sensitized TiO2 solar cells, with comparable short-circuit currents and slightly lower fill factors. From the solar cell measurements on alumina-based devices, it was apparent that the perovskite layer could function as both absorber and n-type component, transporting electronic charge out of the device. We further illustrate the “semiconducting” nature of the perovskite by the construction of a planar-junction diode with the structure FTO / compact TiO2 / CH3NH3PbI2Cl / spiro-OMeTAD / Ag. The perovskite film was ~150 nm thick in this configuration, and the solar cell exhibited Jsc = 7.13 mA cm−2, Voc = 0.64 V, a fill factor of 0.4, and η = 1.8%.

If we take the optical band gap of CH3NH3PbI2Cl to be 1.55 eV from the IPCE onset at 800 nm (33) and the open-circuit voltage to be 1.1 V, this represents a difference in energy of only 0.45 eV, competitive with the best thin-film technologies (2). To understand why we observed such an increase in voltage over the TiO2 cells, we need to consider the operational mode of the two concepts (Fig. 3A). For sensitized TiO2 devices, we would expect that after light absorption in the perovskite, electrons would be transferred to the TiO2 (with subsequent electron transport to the FTO electrode through the TiO2) and holes would be transferred to the spiro-OMeTAD (with subsequent transport to the silver electrode). For Al2O3-based cells, the electrons must remain in the perovskite phase (34) until they are collected at the planar TiO2-coated FTO electrode, and must hence be transported throughout the film thickness in the perovskite. Hole transfer from the photoexcited perovskite to the spiro-OMeTAD should occur in much the same way as in the sensitized device. Al2O3 did not act as an n-type oxide in DSSCs (fig. S4) (25).

Fig. 3

(A) Schematic illustrating the charge transfer and charge transport in a perovskite-sensitized TiO2 solar cell (left) and a noninjecting Al2O3-based solar cell (right); a representation of the energy landscape is shown below, with electrons shown as solid circles and holes as open circles. (B) Photoinduced absorbance (PIA) spectra of the mesoporous TiO2 films (black circles) and Al2O3 films (red crosses) coated with perovskite with (solid lines) and without (dashed lines) spiro-OMeTAD hole transporter, under 496.5 nm excitation at 23 Hz repetition rate. (C) Charge transport lifetime determined by small-perturbation transient photocurrent decay measurement of perovskite-sensitized TiO2 cells (black circles) and Al2O3 cells (red crosses), both with lines to aid the eye. Inset shows normalized photocurrent transients for Al2O3 cells (red trace with crosses every 7th point) and TiO2 cells (black trace with circles every 7th point), set to generate 5 mA cm−2 photocurrent from the background light bias.

To examine the charge generation in these devices, we performed photoinduced absorption (PIA) spectroscopy on the oxide films coated with the perovskite, both with and without the addition of spiro-OMeTAD. For the mesoporous TiO2 film coated with perovskite, the PIA spectrum revealed features in the near-IR assigned to the free electrons in the titania (35), confirming effective sensitization of the titania by the perovskite. In contrast, films made of Al2O3 coated with perovskite exhibited no PIA signal, confirming the insulating role of alumina. After addition of spiro-OMeTAD, we could efficiently monitor the oxidized species of spiro-OMeTAD created after photoexcitation of the perovskite. They had absorption features at 525 and 750 nm, as well as a broad band around 1200 nm, assigned to the hole located on the triarylamine moieties (28, 36), which dominated the spectra in both the TiO2- and Al2O3-based samples. These results indicate that hole transfer is highly effective from the photoexcited perovskite to spiro-OMeTAD, and specifically that a hole conductor is required to enable long-lived charge species within the perovskite coated on the Al2O3. We note that the PIA signal depended both on the concentration and lifetime of the species monitored; hence, from this measurement alone, quantification of the relative charge-generated yield is not possible.

To probe the effectiveness of the perovskite layer at transporting electronic charge out of the device, we performed small-perturbation transient photocurrent decay measurements (37). The solar cells were exposed to simulated sunlight and “flashed” with a small red light pulse; in such experiments, the decay rate of the transient photocurrent signal is approximately proportional to the rate of charge transport out of the photoactive layer (37). As shown in Fig. 3C, we observed that charge collection in the Al2O3-based devices was faster than in the TiO2-based sensitized devices by a factor of >10, indicating faster electron diffusion through the perovskite phase than through the n-type TiO2.

Because there is no n-type oxide in the Al2O3-based cells, the devices are not “sensitized” solar cells, but rather two-component hybrid solar cells. As designed, the Al2O3 is simply acting as a mesoscale “scaffold” upon which the device is structured; we term this concept a “meso-superstructured solar cell” (MSSC). The above measurements demonstrate that long-lived charge carriers can be generated via hole transfer from the perovskite to spiro-OMeTAD and that the perovskite layer is faster at transporting electronic charge than the mesoporous TiO2. However, they do not explain the increase in Voc values. The Voc is generated by the build-up of electrons in the n-type material and holes in the p-type material, resulting in splitting of the quasi Fermi levels for both electrons and holes. For mesoporous TiO2, there exist sites in the tail of the density of states that extend into the band gap (38). These fill with electrons under illumination; the result is that the quasi–Fermi level for electrons (EFn*) is farther from the conduction band, for any given charge density, than would be the case if these states did not exist (i.e., in a highly crystalline semiconductor). The increased charge-storing capacity of materials with a high density of sub–band gap states is termed “chemical capacitance” (38). There is, in essence, no chemical capacitance of the Al2O3, and for the MSSCs all the electronic charge resides in the perovskite, moving the EFn* in this material nearer to the conduction band for the same charge density. The higher voltage indicates that there are fewer surface and sub–band gap states in the perovskite films than in the mesoporous TiO2. Hence, the increased voltage is caused by a substantial reduction of the chemical capacitance of the solar cell. We used a compact layer of TiO2 as the electron-selective anode, but the chemical capacitance of this extremely thin (50 to 100 nm) TiO2 layer was very low because of the low volume and surface area (i.e., flat). In addition, the compact layer deposited via spray pyrolysis has a donor density of ~1018 cm−3 (39), and the sub–band gap sites responsible for the chemical capacitance may be full.

A central question is whether the MSSC is excitonic or a distributed p-n junction. The perovskites tend to form layered structures, with continuous two-dimensional metal halide planes perpendicular to the z axis and the lower dielectric organic components (methyl amine) between these planes. The possible quasi–two-dimensional confinement of the excitons can result in an increased exciton binding energy, which can be up to a few hundred millielectron volts (40). The reasonably high photocurrents from the planar-junction solar cells (Fig. 2B) could be explained by either moderately delocalized and highly mobile excitons being quenched at the perovskite–spiro-OMeTAD interface, or the generation of free charges in the bulk of the perovskite films with reasonably good electron and hole migration out of the devices.

The key limitation in performance of the MSSC at present is a balance between series and shunt resistance. The perovskite absorber is reasonably conductive, measured to be on the order of 10−3 S cm−3; thus, short-circuiting of the device occurs if contact exists between the silver electrode and the perovskite absorber. A thick capping layer of p-type spiro-OMeTAD readily resolves this issue, however; spiro-OMeTAD is less conductive (~10−5 S cm−1), so a thicker capping layer results in high series resistance. Thus, we are presented with a compromise.

Our work represents an evolution of the solid-state sensitized solar cell with low fundamental losses. The application of a mesostructured insulating scaffold upon which extremely thin films of n-type and p-type semiconductors are assembled, termed the meso-superstructured solar cell (MSSC), has proven to be extraordinarily effective with an n-type perovskite, delivering more than 10.9% power conversion efficiency under full solar illumination. Further advances in overall power conversion efficiency are expected by extending the absorption onset toward 940 nm, through the implementation of new perovskites or broadening this concept to other solution-processable semiconductors. Enhancing the light absorption near the band edge through carefully engineered mesostructures or better photon management would lead to increased photocurrent. Reduced series resistance through the use of higher-mobility hole transporters, or better control over the capping layer thickness, would improve the fill factor. Finally, extending this system to multijunction devices (without the requirement for lattice matching, as in conventional multijunction solar cells) would further enhance performance.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgements: Supported by the European Research Council (HYPER project no. 279881), the Strategic International Research Cooperative Program of the UK Engineering and Physical Sciences Research Council, and the Japan Science and Technology Agency. T.M. thanks the funding program for World-Leading Innovative R&D on Science and Technology (FIRST Program), Japan, for hybrid solar cell research. We thank the New Energy and Industrial Technology Development Organization for support. M.M.L. is grateful for support from the Simms Bursary granted by Merton College, Oxford. We thank S. K. Pathak for assistance with x-ray diffraction measurements and analysis, and A. Abrusci, J. Ball, P. Docampo, A. Hey, T. Leijtens, N. Noel, and A. Kojima for valuable discussions. The University of Oxford has filed three patents related to this work.

Stay Connected to Science

Navigate This Article