Report

A hole-conductor–free, fully printable mesoscopic perovskite solar cell with high stability

See allHide authors and affiliations

Science  18 Jul 2014:
Vol. 345, Issue 6194, pp. 295-298
DOI: 10.1126/science.1254763

Improved perovskite photovoltaic performance

A recent entry in the solar cell race is perovskite cells, named for the structure adopted by salt made from metal halides and organic cations that absorb the light and generate charges. The charges generated have to be transferred to a metal oxide (typically titanium oxide), and some of these charge carriers are lost in the transfer. Mei et al. made this process more efficient by growing a more crystalline perovskite with fewer defects inside porous versions of titanium and zirconium oxide. They added a second organic cation that stuck to the pore walls and directed the growth of the perovskite crystals. The improved solar cells operated for more than 1000 hours under full sunlight.

Science, this issue p. 295

Abstract

We fabricated a perovskite solar cell that uses a double layer of mesoporous TiO2 and ZrO2 as a scaffold infiltrated with perovskite and does not require a hole-conducting layer. The perovskite was produced by drop-casting a solution of PbI2, methylammonium (MA) iodide, and 5-ammoniumvaleric acid (5-AVA) iodide through a porous carbon film. The 5-AVA templating created mixed-cation perovskite (5-AVA)x(MA)1-xPbI3 crystals with lower defect concentration and better pore filling as well as more complete contact with the TiO2 scaffold, resulting in a longer exciton lifetime and a higher quantum yield for photoinduced charge separation as compared to MAPbI3. The cell achieved a certified power conversion efficiency of 12.8% and was stable for >1000 hours in ambient air under full sunlight.

Organic-inorganic metal halide perovskite mesoscopic solar cells (13) have rapidly reached conversion efficiencies over 15% (4, 5), and most use a methylammonium lead halide (MAPbX3, X = halogen) and its mixed-halide crystal analog (614). These perovskites have a large absorption coefficient, high carrier mobility, direct band gap, and high stability. They are composed of Earth-abundant materials and can be deposited by low-temperature solution methods. However, perovskite crystallization from solution produces large morphological variations (14) and incomplete filling of the mesoporous oxide scaffold, which results in an unwanted spread of the photovoltaic performance of the resulting devices. Most devices use gold as a back contact, in conjunction with organic hole conductors acting as electron-blocking layers. Hole conductors such as the widely used arylamine spiro-OMeTAD are not only expensive but can limit the long-term stability of the device. Hole-conductor–free perovskite photovoltaics were initially reported by Etgar et al. (15), and to date power conversion efficiencies (PCEs) of 10.85% (1618) have been achieved, with gold as a counter-electrode.

We fabricated a solar cell using a double layer of mesoporous TiO2 and ZrO2 covered by a porous carbon film. The metal halide perovskite was infiltrated into the porous TiO2/ZrO2 scaffold by drop-casting a solution through the printed carbon layer that contained PbI2 in γ-butyrolactone, together with methylammonium (MA) and 5-aminovaleric acid (5-AVA) cations. 5-AVA replaced part of the MA cations in the cuboctahedral site of MAPbI3, forming the new mixed-cation perovskite (5-AVA)x(MA)1-xPbI3. The orthorhombic perovskite phase formed in the presence of 5-AVA showed greatly increased PCE compared with the neat analog in a hole-conductor–free cell that used a simple mesoscopic TiO2/ZrO2/C triple layer as a scaffold to host the perovskite absorber. The role of 5-AVA cations was to template the formation of perovskite crystals in the mesoporous oxide host and induce preferential growth in the normal direction. The resulting mixed-cation perovskite (5-AVA)x(MA)1-xPbI3 has better surface contact with the TiO2 surface and lower defect concentration than the single-cation form MAPbI3, resulting in improved conversion efficiency. The triple-layer device also exhibits excellent long-term stability.

A schematic cross section of the triple-layer, perovskite-based, fully printable mesoscopic solar cell (Fig. 1A) shows that the mesoporous layers of TiO2 and ZrO2 have thicknesses of ~1 and 2 μm, respectively, that were deposited on a F-doped SnO2 (FTO) –covered glass sheet. The mesoporous layers were infiltrated with perovskite by drop-casting from solution through a 10-μm-thick carbon layer printed on top of the ZrO2. The energy band alignment of the device (Fig. 1B) prevents electrons, injected from the photoexcited perovskite into the TiO2 scaffold, from reaching the back contact because of the 0.6-eV offset between its conduction band and that of ZrO2. The compact TiO2 layer deposited on the FTO conducting glass prevents the valence band holes from reaching the FTO-covered front electrode. The CH3NH3PbI3 perovskite unit cell (Fig. 1C) contains one Pb2+, one organic ammonium cation, and three I anions (19).

Fig. 1 The triple-layer perovskite junction.

(A) Schematic drawing showing the cross section of the triple-layer perovskite–based fully printable mesoscopic solar cell. The mesoporous layers of TiO2 and ZrO2 have a thickness of ~1 and 2 μm, respectively, and are deposited on a FTO-covered glass sheet shown in blue and gray. They are infiltrated with perovskite by drop-casting from solution. (B) Energy band diagram of the triple-layer device. Energies are expressed in electron volts, using the electron energy in vacuum as a reference. The energy levels of the conduction band edges of TiO2, ZrO2, and MAPbI3 are at –4.0, –3.4, and –3.9 eV, respectively, whereas the valence band edge of the perovskite is at –5.4 eV and that of the Fermi level of carbon is at –5.0 eV. (C) The crystal structure of MAPbI3 perovskite.

We introduced the 5-AVA cation into the perovskite lattice with a mixture of MA and 5-AVA in the precursor solution that maintained a 1:1 molar ratio of organic ammonium cations and PbI2. The optimal molar ratio of ammonium cations in (5-AVA)x(MA)1-xPbI3 was determined to be between 1:20 and 1:30. The x-ray diffraction (XRD) patterns of neat MAPbI3 and (5-AVA)x(MA)1-xPbI3 infiltrated into the mesoporous ZrO2/TiO2 film are compared in Fig. 2. The newly emerging diffraction peaks arising from the (001) and (111) lattice planes for (5-AVA)x(MA)1-xPbI3 are much stronger than the corresponding ones for MAPbI3. And even at low 5-AVA/MA molar ratios, 5-AVA substantially increases the b and c lattice parameters. The large expansion of the c axis induced by 5-AVA indicates its preferential alignment along this axis through contact with lead and iodide ions, and the c axis becomes the dominant orientation during crystal growth. Mercier (20) used 4-ammonium-butyric acid (4-ABA) as a template for the engineering of (4-ABA)2PbI4 and mixed-cation (4-ABA)2MAPb2I7 perovskite crystals. The ABA molecules formed linear hydrogen-bonded chains that act as supramolecular synthons for layered plumboiodide perovskite structures. We propose that the 5-AVA affects the crystal growth of (5-AVA)x(MA)1-xPbI3 in a similar fashion through the formation of hydrogen bonds between its COOH and NH3+ groups and I ions from the PbI6 octahedra.

Fig. 2 Diffraction data.

XRD patterns of mesoscopic ZrO2/TiO2 film on FTO glass infiltrated with the perovskites (5-AVA)x(MA)1-xPbI3 (red trace) and MAPbI3 (blue trace), as well as a blank ZrO2/TiO2 film (black trace).

We applied transmission electron microscopy (TEM) and scanning electron microscopy (SEM) to unravel differences in the morphology of the two perovskites. High-resolution TEM images (Fig. 3, A to C) reflecting the morphology of neat TiO2 particles, as well as MAPbI3 or (5-AVA)x(MA)1-xPbI3–covered TiO2 particles that were scraped off of mesoporous TiO2 films infiltrated with MAPbI3 or (5-AVA)x(MA)1-xPbI3, respectively. Substantial differences appeared between the morphologies of the samples produced from the two different perovskite precursor solutions. The observed pattern indicates that there is a much denser coverage of the TiO2 particles with (5-AVA)x(MA)1-xPbI3 as compared to MAPbI3. The (5-AVA)x(MA)1-xPbI3 appears to exhibit crystalline features on a much longer length scale than MAPbI3, covering most of the TiO2 surface in a uniform fashion. In contrast, the single-step solution precipitation of MAPbI3 only partially coated the substrate and left large areas where the perovskite absorber was completely absent. The improvement of the perovskite crystal quality and the higher loading of the mesopores of the oxide scaffold in the presence of 5-AVA cations probably resulted from the templating action of the AVA affecting the perovskite crystal nucleation and growth within the mesoscopic oxide scaffold. The COOH groups of 5-AVA anchor a monolayer of the amino acid to the surface of the mesoporous TiO2 and ZrO2 film by coordinative binding to the exposed Ti(IV) or Zr(IV) ions. In the adsorbed state, the terminal -NH3+ groups of 5-AVA face the perovskite solution and hence serve as nucleation sites. This role we attribute to AVA is confirmed by recent work (21) showing that protonated amino acids of the AVA type indeed template the crystal growth of plumbohalide perovskites in mesoscopic TiO2 films, improving the crystalline network and charge-carrier lifetime of the CH3NH3PbI3 inserted into the porous metal oxides.

Fig. 3 Microscopy images.

(A to C) TEM images of (A) TiO2 and mixtures of (B) MAPbI3 and (C) (5-AVA)x(MA)1-xPbI3, with TiO2 particles scraped off of a mesoporous film infiltrated with perovskites. (D and E) SEM images of the cross section of (D) MAPbI3– and (E) (5-AVA)x(MA)1-xPbI3–based perovskite mesoscopic solar cells.

We performed photoluminescence (PL) decay measurements in order to extract quantitative information on the yield of light-induced charge separation. Excitons generated by light excitation of MAPbI3 dissociate into free charge carriers within 1 ps (22). We infiltrated mesoporous TiO2 and ZrO2 films with MAPbI3 or (5-AVA)x(MA)1-xPbI3 and measured the PL decay; ZrO2 serves as a reference because its conduction band is not accessible for electron injection. The PL decay of the MAPbI3 perovskite contained in ZrO2 films exhibits a time constant of τe = 8.14 ns, whereas for the (5-AVA)x(MA)1-xPbI3, τe is 23.7 ns [fig. S1 (23)]. The longer charge-carrier lifetime observed with (5-AVA)x(MA)1-xPbI3 indicates a much lower defect concentration. With the TiO2 scaffold, τe for MAPbI3 and (5-AVA)x(MA)1-xPbI3 were similar, 1.71 and 1.36 ns, respectively; and from the branching ratios, led to calculated quantum efficiencies of 94 and 80%, respectively.

Cross-sectional SEM views of the (5-AVA)x(MA)1-xPbI3– and MAPbI3–based mesoscopic solar cells, which confirm the different extent of pore filling by the two perovskites are shown in Fig. 3, D and E. Thus, for (5-AVA)x(MA)1-xPbI3, most of the mesopores of the ZrO2/TiO2 double layer are fully loaded with the perovskite, whereas for the device with MAPbI3, the pore filling is less complete, with a substantial amount of voids remaining within the mesoporous ZrO2/TiO2 double-layer film. Optical studies also confirmed the higher perovskite loading in the presence of 5-AVA [figs. S2 and S3 (23)].

We measured the photocurrent density as a function of forward bias voltage (J-V curves) of the (5-AVA)x(MA)1-xPbI3– and the MAPbI3– based mesoscopic solar cells under standard reporting conditions [air mass 1.5 (AM1.5) global solar light at 100 mW cm−2 and room temperature]. Figure S4 (23) presents the temporal evolution of the device performance for (5-AVA)x(MA)1-xPbI3 –based mesoscopic solar cells during the initial phase of illumination. The four key photovoltaic parameters; i.e., the open-circuit voltage (Voc), short-circuit photocurrent (Jsc), fill factor (FF), and PCE, increased to a stable value within the first 3 min and then exhibited excellent stability during exposure to full AM 1.5 simulated sunlight over 1008 hours. The final PCE increased slightly during this period, which is more remarkable given that the test was performed with an unsealed device, the perovskite being protected by the 10-μm-thick carbon layer back contact acting as a water-retaining layer.

A hysteresis effect often appears during measurement of the J-V curves of perovskite solar cells. For a mesoscopic triple-layer solar cell with perovskite (5-AVA)x(MA)1-xPbI3 under standard 1 sun irradiation with the scanning rate of 3 mV s−1, there was no substantial effect [fig. S5 (23)]. However, a hysteresis effect appeared at higher scan rates unless the cell was subjected to light soaking for a few minutes, which eliminated the hysteresis effects even at scan rates as high as 250 mV s−1. The perovskite systems show mixed ionic and electronic conduction (24), and under illumination an electric field is set up, which causes a slow drift of ions through the perovskite device to screen the space charge that is generated by light. Figure 4A presents J-V curves collected after 3 min of light soaking with simulated sunlight. In keeping with our previous work (19), the optimized MAPbI3-based heterojunction perovskite solar cell produced a Jsc of 13.9 mA cm−2, a Voc of 855 mV, and a FF of 0.61, yielding a PCE of 7.2%. The (5-AVA)x(MA)1-xPbI3–based cells showed greatly improved performance, with Jsc, Voc, and FF reaching values of 21.1 mA cm−2, 843 mV, and 0.65, respectively, corresponding to a PCE of 11.6%. We sent one of our cells to an accredited photovoltaic calibration laboratory for certification, which measured a Jsc value of 22.8 mA cm−2 [fig. S6 (23)]. The Voc of the device was 858 mV and the FF was 0.66, corresponding to a PCE of 12.84%. Our certified photocurrent density is superior to values reported so far for any perovskite-based photovoltaic, for which the highest certified Jsc is 21.3 mA cm−2 (4). Also in Fig. 4A, the slope of the J-V curve at open circuit voltage is steeper for (5-AVA)x(MA)1-xPbI3 perovskite than for MAPbI3, suggesting a lower series resistance and higher electric conductivity for the former than for the latter perovskite. This shows the better connectivity of the AVA template perovskite crystals formed in the porous oxide network.

Fig. 4 Device performance.

(A) J-V curves under simulated AM 1.5 solar irradiation at an intensity of 100 mW cm−2 measured at room temperature. (B) IPCE curves taken with monochromatic light without applied white-light bias for (5-AVA)x(MA)1-xPbI3 (red curve) or MAPbI3 (blue curve). The integrated photocurrents calculated from the overlap integral of the IPCE spectra with the AM 1.5 solar emission are also shown. The integrated photocurrent of the (5-AVA)x(MA)(1-x)PbI3–based photovoltaic is 17.8 mA cm−2, which agrees closely with the photocurrent density of 18.4 mA cm−2 measured at the beginning of testing, which rose to 21.1 mA cm−2 after 3 min of light soaking.

In the absence of ZrO2, the performance of the perovskite photovoltaic is poor, with Jsc = 12.51 mA cm−2, Voc = 592 mV, FF = 0.56, and PCE = 4.18% [fig. S7 (23)]. This confirms that ZrO2 is blocking the flow of photogenerated electrons to the back contact, preventing recombination with the holes from the perovskite at the back contact.

Figure S8 (23) reports statistical data concerning the numerical distribution of the key photovoltaic parameters; i.e., Voc, Jsc, FF, and PCE for a random selection of 20 photovoltaic devices, which were fully printed on a 10 cm × 10 cm FTO glass as shown by the photograph in fig. S9 (23). (5-AVA)x(MA)1-xPbI3–based mesoscopic solar cells exhibited an average PCE value of 10.3%, Jsc of 21.68 mA cm−2, Voc of 740 mV, and FF of 0.64 together with a small standard deviation, indicating that the (5-AVA)x(MA)1-xPbI3 strongly augments the reproducibility of the perovskite-based cells. Figure 4B shows the incident photon-to-electric current conversion efficiency (IPCE), defined as the number of photogenerated charge carriers contributing to the photocurrent per incident photon. These experiments used monochromatic light without applying a white-light bias. (5-AVA)x(MA)1-xPbI3 achieved much higher IPCEs than MAPbI3 over the whole spectral range between 300 and 800 nm, matching the difference in photocurrents obtained for the two devices.

Supplementary Materials

www.sciencemag.org/content/345/6194/295/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S9

References

References and Notes

  1. See the supplementary materials on Science Online.
  2. Acknowledgment: The authors acknowledge financial support from the Ministry of Science and Technology of China (863 project, grant no. SS2013AA50303), the National Natural Science Foundation of China (grant no. 61106056), the Science and Technology Department of Hubei Province (grant no. 2013BAA090), and the Fundamental Research Funds for the Central Universities (grant no. HUSTNY022). We thank the Analytical and Testing Center of Huazhong University of Science and Technology for field emission SEM, TEM, and XRD testing. M.G. thanks the European Research Council for financial support under advanced research grant ARG 247404, “Mesolight.”
View Abstract

Navigate This Article