Iodide management in formamidinium-lead-halide–based perovskite layers for efficient solar cells

See allHide authors and affiliations

Science  30 Jun 2017:
Vol. 356, Issue 6345, pp. 1376-1379
DOI: 10.1126/science.aan2301

Healing defects with triiodide ions

Deep-level defects in organic-inorganic perovskites decrease the performance of solar cells through unproductive recombination of charge carriers. Yang et al. show that introducing additional triiodide ions during the formation of layers of formamidinium lead iodide, which also contain small amounts of methylammonium lead bromide, suppresses the formation of deep-level defects. This process boosts the certified efficiency of 1-cm2 solar cells to almost 20%.

Science, this issue p. 1376


The formation of a dense and uniform thin layer on the substrates is crucial for the fabrication of high-performance perovskite solar cells (PSCs) containing formamidinium with multiple cations and mixed halide anions. The concentration of defect states, which reduce a cell’s performance by decreasing the open-circuit voltage and short-circuit current density, needs to be as low as possible. We show that the introduction of additional iodide ions into the organic cation solution, which are used to form the perovskite layers through an intramolecular exchanging process, decreases the concentration of deep-level defects. The defect-engineered thin perovskite layers enable the fabrication of PSCs with a certified power conversion efficiency of 22.1% in small cells and 19.7% in 1-square-centimeter cells.

The main interest in inorganic-organic hybrid perovskite materials stems from the huge success of their application in perovskite solar cells (PSCs), which has led to power conversion efficiency (PCE) of >20% (15). To date, several factors, such as architecture, deposition process, and compositional manipulation of the perovskite materials have been probed and altered in attempts to fabricate efficient PSCs (410). Previously, we proposed a bilayer architecture consisting of perovskite-infiltrated mesoporous TiO2 with thin upper layers to fabricate efficient PSCs (6, 7). Furthermore, we devised a solvent-engineering process, which used a dropwise dripping of an orthogonal solvent during spinning of the perovskite precursor in a mixture of γ-butyl lactone and dimethyl sulfoxide (DMSO) to produce a bilayer architecture with an extremely uniform and dense surface morphology (7). This process was successfully repeated and applied to the fabrication of highly efficient PSCs with PCEs of more than 20% (24). Further, we demonstrated the fabrication of efficient PSCs based on formamidinium lead iodide (FAPbI3) using a two-step process based on intramolecular exchange between organic cations and DMSO molecules (5).

Theoretical calculations have shown that point defects, such as vacancies (VMA, VPb, and VI) and interstitials (MAi, Pbi, and Ii), are the most likely defects in methylammonium (MA) lead iodide (MAPbI3) perovskites because of their low energies of formation (11, 12). Moreover, cation substitutions (MAPb and PbMA) and antisite substitutions (MAI, PbI, IMA, and IPb) can also form. Although most vacancy defects create shallow electronic levels near the band edges (1113), the formation of deep-level defects, such as interstitial and antisite defects, which are responsible for nonradiative recombination centers in perovskite layers, depends on I-poor or I-rich conditions (11).

Recombination centers in polycrystalline perovskite layers can also be generated by grain boundaries (GBs) (1418), but electron beam–induced current studies (19) have shown that GBs are inert toward charge recombination and might even be beneficial for charge separation, as demonstrated by using scanning Kelvin probe force microscopy (20). In situ or posttreatment of GBs with, for example, residual PbI2 (14, 21); 1,2-ethanedithiol (15); pyridine (16); phenyl-C61-butyric acid methyl ester (17); and guanidinium (18) improved performance by passivating undercoordinated Pb and I ions. Nevertheless, the observed increases in performance were limited. Internal passivation methods exploiting the balance of halide ions during the processing stage have not yet been reported for perovskite layers, and approaches that would reduce the number of compositional point defects are necessary to achieve further improvements in the performance of PSCs.

Stewart et al. (22) suggested that a two-step deposition method could suppress the formation of interstitial iodide. In particular, the presence of I-rich tetraiodoplumbates (PbI4 species) in the precursor solution, where PbI2, formamidinium iodide (FAI), and methylammonium iodide (MAI) are dissolved in dimethylformamide (DMF), may be related to the density of defects stemming from interstitial iodine (Ii) and act as charge recombination centers. Nevertheless, a two-step process that used DMSO as the mediator may also result in the formation of [(Pb3I8)n]2n, which is an I-deficient intermediate that is charge-balanced by FA (or MA) (23).

We report a two-step deposition method for the fabrication of high-quality perovskite films that uses an optimized dripping solution to reduce the degree of iodide deficiency, which is common with the two-step process. The PSCs fabricated using this method showed remarkably improved performance compared with the control cell. The introduction of additional iodide ions into the dripping solution facilitated the fabrication of a PSC with a certified PCE of 22.1% under 1 sun illumination (100 mW cm–2).

To facilitate an α-phase FAPbI3 perovskite as a light harvester that is stable at room temperature, we manipulated the composition of FAPbI3 by introducing a small amount of MAPbBr3 (8). We incorporated MAPbBr3 into FAPbI3 perovskite films using a modified intramolecular exchange process (5). PbI2 and PbBr2 (mole ratio = 95/5) were dissolved in a mixture of DMF and DMSO (80/20, v/v). We first deposited a PbI2(PbBr2)-DMSO film by spin-coating a PbX2 (X = I, Br) solution containing PbI2 and PbBr2 dissolved in a mixture of DMF and DMSO onto a nanocrystalline TiO2 scaffold. In the second step, a solution of FAI and MABr, containing iodide ions (I) dissolved in isopropyl alcohol (IPA), was spin-coated.

To prepare this solution, an iodide solution was first prepared by stirring solid iodine (I2) in IPA at 80°C. Iodine can be ionized to I by oxidation of IPA with the formation of acetone via Eq. 1 (24).Embedded Image(1)A continued exergonic equilibrium gives rise to triiodide ions (Embedded Image), as shown in Eq. 2:

Embedded Image(2)

Analysis of the solution of I2 in IPA by ultraviolet-visible (UV-vis) spectroscopy (fig. S1) (25) revealed that the intensities of the absorption peaks at 291 and 361 nm increased during the course of stirring, indicating that Embedded Image increased as a function of the stirring time through the reactions shown in Eqs. 1 and 2 (26), with all iodine eventually converted into the Embedded Image after stirring for 7 days. Subsequently, FAI(+MABr) was dissolved in an IPA solution containing 1 to 5 mmol of Embedded Image to prepare the dripping solution for perovskite layer formation. After spin-coating, this solution was retained on the surface to induce intramolecular exchange between DMSO and FAI(+MABr), followed by annealing at 150°C. We fabricated PSCs with the following architecture: FTO/thin-barrier TiO2 (~60 nm)/mesoporous-TiO2:perovskite composite layer (~150 nm)/perovskite upper layer (~500 nm)/PTAA (~50 nm)/Au (~100 nm), where FTO is F-doped SnO2 and PTAA is poly(triarylamine) [fig. S2A (25)].

The performance of the PSCs increased with the addition of any Embedded Image to the dripping solution, with the maximum performance or “target device” attained when the dripping solution contained 3 mmol of Embedded Image (Fig. 1A). The current-voltage (I-V) characteristics (Fig. 1B) of the target device were compared with those of the control (no added Embedded Image). The target PSC exhibited an open-current voltage (Voc) of 1.1 V, a short-circuit current (Jsc) of 24.1 mA·cm–2, and a fill factor (FF) of 81.9%, for a PCE of 21.6%. The control device showed an overall PCE of 20.3% with a Voc of 1.07 V, a Jsc of 23.5 mA·cm–2, and a FF of 80.8%. The enhanced Jsc value arose from improved light-harvesting capacity rather than increased charge collection, as the thicknesses and surface morphologies of the target and control PSCs are almost identical (fig. S2B) (25). Moreover, the proportion of the α-FAPbI3 phase is increased by the addition of Embedded Image, as will be evidenced by the UV-vis absorption spectra and the grazing-incidence wide-angle x-ray scattering (GIWAXS) analysis, despite the similar thicknesses. The improvements in both Voc and FF are indicative of improved crystallinity and a smaller number of trapping sites, which are responsible for nonradiative charge recombination of perovskite domains. In particular, the unavoidable introduction of defects into the crystalline lattice during processing increases charge recombination. High crystallinity and preferred orientation, likewise, will influence charge dissociation, transport, and diffusion length considerably (27). The Jsc value obtained from the J-V characteristics is well matched (within 5%) with the external quantum efficiency (EQE) obtained by the integration of the spectral response (Fig. 1C). Nevertheless, the EQE exhibited very interesting features over the spectral range of the device response. Namely, the EQE value is higher at longer wavelengths, most likely from slight changes in the ratio of nonperovskite and perovskite materials between the surface and the interior of the films.

Fig. 1 Device performance analysis and UV-vis absorption spectra of perovskite layers deposited via a two-stage process with and without the addition of triiodide ions into the dripping solution.

(A) Relation between the relative power conversion efficiency and the concentration of iodide ions added into the dripping solution. (B) J-V characteristics determined under simulated AM 1.5 G illumination for the control and target devices. (C) External quantum efficiency and integrated Jsc for the control and target devices. (D) UV-vis absorption spectra for the control and target perovskite layers (inset is magnified absorption).

The UV-vis absorption spectra obtained for perovskite thin films deposited using PbI2(PbBr2)-DMSO, followed by FAI(+MABr) in IPA solution with and without external I3, showed very similar absorption profiles but a slightly higher absorption over the entire region below 780 nm in comparison with the control layer (Fig. 1D). This outcome suggests that the proportion of α-FAPbI3 phases increased in the target layer. The fraction of MAPbBr3 in FAPbI3 was estimated to be ~5 mol% from the blue-shift of the onset absorption (5). We also conducted GIWAXS analysis using a two-dimensional (2D) detector to probe the orientation and crystallinity of internal layers for the FAPbI3 layers deposited with and without Embedded Image with the x-ray incident angle at 0.05° (near surface) or 0.3° (full depth) to the sample (figs. S3 and S4) (25). Signals corresponding to PbI2 disappeared at the lower incident angle (0.05°) and decreased substantially at the higher angle (0.3°) in the target layer relative to the control, but those for both the α- and δ-FAPbI3 increased in intensity in the target layer. The difference between the surface and full-depth diffractions of PbI2 reflects the reaction between PbI2-DMSO and FAI(+MABr), which is supplied externally. The formation of FAPbI3 likely occurs via intermediates such as PbI2-DMSO-FAI(+MABr) and PbI2-DMSO phases through the two-step process. The intermediate consists of an I-deficient (organic cations)–Pb3I8 phase (23) or an I-deficient PbI2-DMSO complex forms because PbI2 itself is often I-deficient (28). Thus, Embedded Image in the dripping solution would both increase the proportion of the crystalline α-FAPbI3 phase and reduce the concentration of defects in the entire perovskite layer.

To gain deeper insight into the nature of these defects, we performed deep-level transient spectroscopy (DLTS) (29) and steady-state and time-resolved photoluminescence (PL). The temperature scans for the control layer reveal three defect levels, denoted A1, A2, and A3 (Fig. 2A). The activation energy (Ea) and capture cross-section (EC) values determined from the Arrhenius plots in fig. S5 (25) were EC of –0.82 eV and 2.47 × 10−12 cm2 in A1, EC of –0.78 eV and 2.39 × 10−12 cm2 in A2, and EC of –0.46 eV and 8.65 × 10−16 cm2 in A3, respectively. In the target layer, the A1 signal disappeared readily and the defect concentration for A2 decreased from 5.28 × 1014 cm–3 to 8.81 × 1013 cm–3 through interaction with Embedded Image. From the disappearance of the A1 peak, we surmise that defect sites located in the A1 level can be easily healed by the presence of additional Embedded Image when the perovskite phases are formed. Dominant deep-level defects (A1 and A2) and semishallow defects (A3) with energy levels of 0.78 to 0.82 eV and 0.46 eV below the conduction band, respectively, have not yet been experimentally identified but may be associated with defect states originating from interstitial Pb (Pbi) and antisite defects (MAI, PbI, IMA, and IPb) (12, 13). Indeed, Buin et al. (12) theoretically predicted that these deep-level trap concentrations would be affected by changes in the concentration of iodide anions and calculated the energy levels of IPb, PbI, and Pb defects.

Fig. 2 Deep-level transient spectroscopy (DLTS), and time-resolved photoluminescence (TRPL) measurements.

(A) DLTS spectra of the control and target layers measured in between 150 K and 330 K. (B) TRPL decay curves on the control and target perovskite layers emitted at λ = 825 nm with the biexponential fitting.

The target layer exhibited an increase in PL intensity of ~225% relative to the control layer for excitation at 620 nm (fig. S6) (25). The improvement in the PL intensity can be correlated to the difference (ΔV) in the Voc measured from the J-V characteristics, as described in the supplementary materials, suggesting that Embedded Image decreases the numbers of deep-level traps within bulk perovskite layers. The carrier dynamics derived from the transient PL behavior provides information of charge recombination via nonradiative recombination related to the defect concentration. The PL decay (Fig. 2B) measured by time-correlated single-photon counting was fitted to a biexponential equation: Y = A1exp(−t1) + A2exp(−t2) (30). We attributed quenching of the short PL lifetime to defect-induced nonradiative recombination and of the long PL lifetime to radiative recombination (31, 32). The target layer displays a PL lifetime of ~138 ns and a long carrier lifetime of ~1105 ns, whereas the control layer exhibits PL lifetimes of ~72 ns and ~228 ns, respectively. The longer lifetime of the PL transition in the target layer may be attributed to the decrease in the concentration of defects and the increase in crystallinity. In addition, the PL quantum efficiency (PLQE) as a function of excitation power for the radiative and nonradiative recombination is higher in the target layer at all light intensities [fig. S7 (25)] (3335).

On the basis of these results, we repeated the fabrication procedure to further enhance the PSC performance, using the tuned dripping solution in a two-step process. The J-V curves presented in Fig. 3A were determined for one of the most efficient cells, measured with a 50-ms scanning delay in both reverse- and forward-scan modes under standard air-mass 1.5 global (AM 1.5 G) illumination. The J-V curves show no appreciable hysteresis between the two modes. The Jsc, Voc, and FF values obtained from the J-V curve in the reverse-scan mode were 25.0 mA·cm−2, 1.11 V, and 0.817, respectively, yielding a PCE of 22.6% under standard AM 1.5 G conditions. The corresponding values obtained from the J-V curve in the forward-scan mode were 25.0 mA·cm−2, 1.10 V, and 0.805, respectively, yielding a similar overall efficiency of 22.2%. For certification purposes, these representative devices were sent to the independent laboratory, which revealed an overall efficiency of 22.1% for the best-performing cell (fig. S8) (25), the highest efficiency ever reported for a PSC. The efficiency of this cell was maintained at >93% of the initial value after ~13 months of storage at ambient conditions (fig. S9) (25).

Fig. 3 Current density–voltage (J-V) curves and photovoltaic parameters of the best-performing small and large PSCs and device reproducibility.

(A) J-V curves of a small PSC (0.095 cm2) in forward- and reverse-scan modes and the corresponding photovoltaic parameters. (B) Histogram of the average power conversion efficiency determined for 80 PSC devices. (C) J-V curve for a large PSC (1.0 cm2) plotted as the average of the reverse- and forward-scan modes and the corresponding photovoltaic parameters.

We prepared 80 cells independently under the same experimental conditions. The histogram of the average PCEs (Fig. 3B) shows that ~90% of the cells had PCEs exceeding 20.0% under 1-sun illumination with an average efficiency of 21.25 ± 1.08%. We also fabricated cells with active areas of ~1 cm2 to check the uniformity of the perovskite layers produced using the above-described methodology. The J-V curve determined as an average of the reverse and the forward scans with corresponding Jsc, Voc, and FF values of 24.2 mA cm–2, 1.14 mV, and 72.7%, respectively, achieving an overall PCE of 20.0% under standard AM 1.5 G conditions (Fig. 3C); the certified performance was 19.7% under AM 1.5 G full-sun conditions (fig. S10) (25). The slight decrease in the performance of the large cell relative to that of smaller cells can be attributed to the large sheet resistance of the FTO substrate. This study demonstrates that careful control of the growth conditions of perovskite layers with management of deficient halide anions is essential for realizing high-efficiency thin-film PSCs based on lead-halide-perovskite absorbers.


Materials and Methods

Figs. S1 to S10

References (3638)


  1. Materials and methods are available as supplementary materials.
  2. Acknowledgments: This work was supported by the Global Frontier R&D (Research and Development) Program of the Center for Multiscale Energy System (NRF-2011-0031565 and NRF-2016M3A6A7945503) and the Climate Change Program (NRF-2015M1A2A2056542) through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT (Information and Communication Technology) and Future Planning. This work was also supported by a brand project (1.170003.01) of the Ulsan National Institute of Science and Technology (UNIST) and internal project (no. KK1702-A01) of the Korea Research Institute of Chemical Technology (KRICT). The GIWAXS experiment at the PLS-II 6D UNIST–Pohang Accelerator Laboratory beamline was supported in part by the Ministry of Education, Science, and Technology; Pohang University of Science and Technology; and UNIST–Central Research Facilities Center. We thank Tae Joo Shin for the indexing of 2D GIWAXS patterns. N.J.J. acknowledges financial support from the Basic Science Research Program (NRF-2015R1A6A3A04058164). All results are reported in the main text and supplementary materials. W.S.Y., J.H.N., J.S., and S.I.S. are inventors on patent application (KR 10-2016-0097290) submitted by both UNIST and KRICT that covers the preparation of perovskite films.

Stay Connected to Science


Navigate This Article