Enhanced mobility CsPbI₃ quantum dot arrays for record-efficiency, high-voltage photovoltaic cells

Perovskite quantum dot coupling tuned by A-site cation halide treatment yields a record quantum dot solar cell efficiency.

Abstract

We developed lead halide perovskite quantum dot (QD) films with tuned surface chemistry based on A-site cation halide salt (AX) treatments. QD perovskites offer colloidal synthesis and processing using industrially friendly solvents, which decouples grain growth from film deposition, and at present produce larger open-circuit voltages (V_OC’s) than thin-film perovskites. CsPbI₃ QDs, with a tunable bandgap between 1.75 and 2.13 eV, are an ideal top cell candidate for all-perovskite multijunction solar cells because of their demonstrated small V_OC deficit. We show that charge carrier mobility within perovskite QD films is dictated by the chemical conditions at the QD-QD junctions. The AX treatments provide a method for tuning the coupling between perovskite QDs, which is exploited for improved charge transport for fabricating high-quality QD films and devices. The AX treatments presented here double the film mobility, enabling increased photocurrent, and lead to a record certified QD solar cell efficiency of 13.43%.

INTRODUCTION

Colloidal quantum dot (QD) materials offer unique properties over those afforded by conventional thin films for optoelectronic and other applications. While it is well known that colloidal QD systems offer immense tunability in the material bandgap, energetic position of the electronic states, and surface chemistry, more unique features are being discovered (1–4). For photovoltaics (PVs), colloidal QDs exhibit efficient multiple exciton generation, which led to the first PV and photoelectrochemical cells with external quantum efficiency (EQE) exceeding unity within the solar spectrum (5, 6). Additionally, it has recently been shown that colloidal CsPbI₃ QD materials stabilize the cubic perovskite crystal phase (4, 7), whereas thin-film CsPbI₃ materials relax to an orthorhombic phase at ambient temperature (8).

CsPbI₃ with $\mathit{P}\mathit{m}\overline{3}m$ cubic symmetry exhibits the lowest bandgap (E_g = 1.73 eV) (9) among the all-inorganic Pb-halide perovskite materials for PV. However, at room temperature, the orthorhombic phase with Pnma space group symmetry (E_g = 2.82 eV) is thermodynamically preferred (10). To overcome this phase instability, the addition of bromide (CsPbI_3−xBr_x) has been shown to reduce the phase transition temperature from over 300° to ~110°C (10); however, for PV applications, this reduction in temperature is not low enough, and alloying comes at the expense of increasing the bandgap to ~1.9 eV for CsPbI₂Br (11, 12). In an alternative approach, we previously demonstrated that, by leveraging the surface energy of QDs, the cubic phase can be stabilized at room temperature and below, leading to a PV device with efficiency >10% (4). Although we observed a high open-circuit voltage (V_OC) (~85% of the maximum voltage from the Shockley-Queisser analysis for the given bandgap), the short-circuit current density (J_SC) was limited by transport. This is a common trade-off in QD solar cells (QDSCs) whereby increasing light absorption with thick absorber layers reduces charge extraction efficiency in transport-limited QD films (13). Here, we show, for the first time, an AX posttreatment [where A = formamidinium (FA⁺), methylammonium (MA⁺), or cesium (Cs⁺) and X = I⁻ or Br⁻] that tunes and greatly improves the electronic coupling between QDs, which enhances carrier mobility. We characterize the structural, optical, and electrical properties of the resultant CsPbI₃ QD films and determine that the fabrication process enables the films to retain nanocrystalline character to preserve quantum confinement and that the AX salt species coats the QDs in the array rather than alloying or inducing grain growth in the films (these AX-posttreated CsPbI₃ QD films are thus subsequently referred to as AX-coated).

RESULTS

PV device architecture and certified efficiency

Metal halide salts, such as CdCl₂ and PbI₂ (13–16), and molecular halides (17) have been used to improve QD coupling, passivate surfaces, tune device energetics, and improve stability of the more mature Pb-chalcogenide QDSC technology (14, 18). Banking on some potential similarities in the surface chemistry between CsPbI₃ and PbSe or PbS, we explore a variety of AX salts to improve electronic coupling in CsPbI₃ QDs. Figure 1 details one of the better performing devices made in our laboratory using formamidinium iodide (FAI) as the AX treatment. A schematic of the device architecture used (Fig. 1A) is shown mapped to a scanning electron microscopy (SEM) cross-sectional image (Fig. 1B) with false coloring applied to delineate the various layers used in the stack. Figure 1 (C to E) shows measurements performed by the National Renewable Energy Laboratory (NREL) PV Performance Characterization Team including a current density–voltage (J-V) scan (Fig. 1C) for determining standard PV performance characteristics such as V_OC, J_SC, and the maximum power point. The device was held at a voltage of 0.95 V, and a measurement of the output current versus time was recorded and is shown in Fig. 1D. Because of the hysteretic nature of perovskite solar cells (19, 20), this stabilized current output at a set voltage is used to determine the certified AM1.5G efficiency after the spectral mismatch (calculated, in part, from the EQE shown in Fig. 1E), and device active area were determined. The stabilized current at 0.95 V matches well to the reverse J-V scan shown in Fig. 1C, and the results indicate an overall efficiency of 13.43%, which is currently the highest efficiency colloidal QDSC reported to date (21).

Fig. 1. High-efficiency FAI-coated CsPbI₃ QDSCs.

(A) Schematic cross section of a solar cell. FTO, fluorine-doped tin oxide. (B) SEM image of a device cross section. (C to E) NREL-certified (C) J-V characteristics from forward bias to reverse bias, (D) stabilized current at a constant voltage of 0.95 V, and (E) EQE.

Characterization of AX-coated CsPbI₃ QDSCs and films

A variety of AX salts were investigated as potential treatments for modifying the QD array. To fabricate the CsPbI₃ QD films for this work, CsPbI₃ QDs were synthesized as described previously (4, 7) and deposited in a layer-by-layer fashion by spin coating from octane. To partially remove the native ligands and allow for further layers to be deposited without redispersing the existing layers, we immersed the film into a saturated lead(II) nitrate [Pb(NO₃)₂] solution in methyl acetate (MeOAc) after each QD layer (4). After building up a sufficiently thick CsPbI₃ QD film (three to four deposition cycles, leading to 200- to 400-nm-thick films), the film was immersed in a saturated AX salt solution in ethyl acetate (EtOAc) for ~10 s. Figure 2A shows a flow diagram depicting this process for film assembly.

Fig. 2. Effect of AX salts on CsPbI₃ QD films and PV performance.

(A) Schematic of the film deposition process and AX salt posttreatment. (B) J-V characteristics of CsPbI₃ QD devices treated with FAI (pink), methylammonium iodide (MAI) (green), formamidinium bromide (FABr) (yellow), methylammonium bromide (MABr) (gray), cesium iodide (CsI) (dark blue), and neat EtOAc control (blue). (C and D) ToF-SIMS depth profile of CsPbI₃ QD films (C) without and (D) with an FAI posttreatment. Intensity is normalized to total counts at each data point. (E) Ratios of Cs/Pb, I/Pb, and FA/Pb calculated from the integrated ToF-SIMS signal intensities throughout the film. int., integrated; arb., arbitrary.

A series of solar cells with varied composition of the AX salt posttreatment, as well as a control device where only neat EtOAc is used, were fabricated and investigated to determine the role of the various A and X components as well as the structural makeup of the resulting films. Figure 2B compares J-V scans for devices with the various AX salt treatments. In comparison to the control, all the AX salt posttreatments markedly increase the performance of the CsPbI₃ QDSCs, demonstrating the generality and effectiveness of this surface treatment scheme. Among the different AX salt posttreatments explored in this series, the highest power conversion efficiency (PCE) of 13.4% was achieved from the FAI treatment with a J_SC of 14.37 mA/cm², which was remarkably better than the neat EtOAc control with a PCE and J_SC of 8.5% and 9.22 mA/cm², respectively. As seen in Fig. 2B, the posttreatments have little impact on the V_OC or fill factor (FF), and the PCE improvement is almost entirely attributed to the increased J_SC (see Table 1). This suggests that the different AX salt treatments affect the QD coupling to varying degrees and may be used to tune the electronic properties of the QD films.

Table 1. Effect of AX salts on PV parameters extracted from J-V scans.

AX salt posttreatment	VOC (V)	JSC (mA cm−2)	FF	PCE (%)
FAI (EtOAc)	1.20	14.37	0.78	13.4
FABr (EtOAc)	1.22	12.70	0.81	12.6
MAI (EtOAc)	1.20	13.39	0.79	12.6
MABr (EtOAc)	1.21	11.27	0.82	11.2
CsI (EtOAc)	1.20	10.64	0.81	10.3
Neat EtOAc	1.17	9.22	0.78	8.5

By comparing the champion device treated with FAI to those treated with neat EtOAc (fig. S1), we measure a ~60% improvement in J_SC, improving the PCE obtained from a J-V scan from 8.5 to 13.8% and the stabilized power output (SPO) from 6.1 to 13.5%. This photocurrent increase occurs because of a broadband improvement in the overall EQE rather than a change in the EQE onset, which would indicate a change in the bandgap (fig. S1). Moreover, the absorbance spectra of the films using neat EtOAc and FAI treatments show similar onsets at ~700 nm, consistent with the EQE spectra, but also show similar overall light absorption (fig. S2), ruling out the possibility that the improvement seen in J_SC arises from improved light absorption. We find a similar broadband improvement in EQE for all of the investigated AX salt posttreatments (fig. S3), suggesting that this AX posttreatment scheme is a general way to improve the charge carrier collection efficiency in the CsPbI₃ QD films. We investigated the case of FAI in detail; however, other AX salt posttreatments after optimization could yield even higher performance. In addition to an improvement in the PCE, we also observe a noticeable reduction in hysteresis between the forward (that is, J_SC to V_OC) and reverse (that is, V_OC to J_SC) J-V scans, as well as better agreement between the SPO and reverse J-V scan after the FAI posttreatment (fig. S1). Histograms of the PCE, V_OC, J_SC, and FF of 78 devices fabricated with this FAI posttreatment are provided in fig. S4 to demonstrate the reproducibility of these high-efficiency (>12%), high-voltage (>1.15 V) QDSCs.

We confirm the presence of the FAI salt in the film after the posttreatment via time-of-flight secondary ion mass spectrometry (ToF-SIMS) depth profiling, as seen in Fig. 2 (C to E) [see Materials and Methods for description of mass/charge ratio (m/z) values assigned to ions used to track each component]. Comparing the FAI-coated film (Fig. 2D) to the control film (Fig. 2C), the Cs, Pb, and I signals remained approximately the same, whereas there is a significant increase in the FA signal intensity. To directly compare the Cs, I, and FA, we calculate ratios of the signal intensities integrated over the thickness of the film (Fig. 2E). The FAI-coated film has a significant increase in the FA/Pb ratio (as expected, the signal at m/z = 45 is at the instrument response limit before treatment) and a slight increase in the I/Pb ratio, whereas the Cs/Pb ratio is unchanged. Furthermore, the ToF-SIMS trace of the FA component indicates a higher concentration of FA⁺ at the sample surface, which is corroborated by x-ray photoemission spectroscopy (XPS) (fig. S5), confirming the presence of FA⁺ on the surface. The slight increase in the I/Pb ratio is confirmed for the sample surface by XPS as well. We also identify the presence of FA⁺ in the film by transmission mode Fourier transform infrared (FTIR) spectroscopy (fig. S6) from the emergence of a peak at 1712 cm⁻¹ following FAI treatment, which is characteristic of the C═N stretch in FA⁺.

It is clear that FA⁺ is present in the treated films; therefore, four scenarios regarding the interaction between FAI and the CsPbI₃ QD film are hypothesized: (i) The FAI interactions could be limited to the QD surface, either binding to the QD or infilling the void space between QDs; (ii) a partial cation exchange could result in an alloyed Cs_xFA_1−xPbI₃ QD core; (iii) a perovskite FAPbI₃ matrix or shell could form around the CsPbI₃ QDs if excess of Pb²⁺ and I^- are present; and (iv) the FAI salt could be inducing grain growth in the CsPbI₃ QD film, similar to the observed Ostwald ripening of MAPbI₃ thin films treated with MABr (22). In scenario (iv), there would no longer be individual QDs present in the treated films, and the films would resemble bulk CsPbI₃ or alloyed Cs_xFA_1−xPbI₃ films.

Because the EQE onset (figs. S1 and S3) and absorption (fig. S2) are effectively unchanged after FAI treatment, drastic changes in the chemical composition of the film are unlikely, as we would expect a greater red shift in the case of a Cs_xFA_1−xPbI₃ alloying and an absorption onset at ~840 nm in the case of a FAPbI₃ matrix (6). X-ray diffraction (XRD) patterns before and after the FAI treatment (fig. S7) are absent of any detectable peak shifts or new peaks, which would indicate alloying or formation of a new perovskite phase, respectively. Therefore, scenarios (ii) and (iii) are ruled out. Additionally, we find a decrease in the C–H modes associated with the as-synthesized oleate and oleylammonium ligands near 3000 cm⁻¹ in the FTIR, suggesting that additional native ligands have been removed following the FAI treatment. Removal of native ligands reduces interparticle spacing and increases QD coupling [scenario (i)], which is well known to induce a bathochromic shift in QD samples (23).

To assess whether the increased charge carrier collection efficiency is the result of CsPbI₃ QD grain growth induced by the FAI treatment [scenario (iv)], we characterized the photoluminescence (PL) from the films of three different sizes of QDs, which are readily synthesized by varying the reaction temperature (with lower reaction temperatures producing smaller QDs with increasing bandgap) (4, 7). After performing the FAI treatment, there is still a significant blue shift in the spectral position of the PL peak with decreasing QD size (Fig. 3A) compared to bulk CsPbI₃, which verifies that most of the quantum confinement is retained following the FAI treatment. High-resolution atomic force microscopy (AFM) (Fig. 3, B and C) and SEM (fig. S8) images of the QD films before and after FAI treatment readily resolve individual QDs in the films and confirm that the QD film morphology is maintained after FAI treatment. Note that the grains shown in the AFM images in Fig. 3 (B and C) appear larger than those observed in SEM images (fig. S8) because of the tip-sample convolution. Thus, we conclude that the CsPbI₃ QD structure remains intact, despite the addition of FA to the film. Given that the QD films perform well with <30-nm grains is encouraging because it departs from the growing notion that large grains are absolutely essential for high-performance perovskite devices. The QD films described here decouple the grain growth from the film formation, which could be highly beneficial for rapid manufacturing.

Fig. 3. Confinement in FAI-coated films.

(A) Normalized (norm.) PL of CsPbI₃ QD films fabricated with three different sizes (as indicated by the injection temperature during synthesis) of QDs with (pink) and without (blue) FAI treatment. (B and C) AFM images of CsPbI₃ QD films (B) without and (C) with an FAI posttreatment.

The combined evidence of absorption, EQE, PL, ToF-SIMS, FTIR, XRD, XPS, and multimodal microscopy are consistent with the addition of the FAI to the CsPbI₃ QD film without significant changes in the composition of the QD core. We therefore refer to the treated CsPbI₃ QD films as AX-coated films. Given the fast anion exchange reported in the CsPbX₃ QD systems and some limited reports on the A-site cation exchange, it is notable that the AX treatment is limited to QD surface interactions but is able to penetrate through the depth of the QD film. A more detailed study of the interaction of the FAI treatment on the CsPbI₃ QD surface is forthcoming.

Because it is clear that quantum confinement is preserved in the FAI-coated films, we hypothesize that the slight red shift seen in the absorbance and PL following FAI treatment arises from the improved electronic coupling between individual QDs. Increasing the wave function overlap between neighboring QDs relaxes confinement in the QDs and improves charge carrier transport, which explains the increased photocurrent following AX coating. We therefore probed the mobility of FAI-coated CsPbI₃ QD films using time-resolved terahertz spectroscopy (TRTS), which is a contactless pump-probe optical method capable of characterizing both conductivity and carrier dynamics in semiconductor films. In addition to the CsPbI₃ QD films, we also probed conventional MAPbI₃ thin films as well as PbSe and PbS QD films to compare CsPbI₃ QDs with more heavily studied materials (24–27). The PbS QD sample was composed of 3-nm QDs used commonly in literature for high-efficiency QDSCs, whereas larger 6-nm PbSe QDs were chosen for their higher reported mobility (25).

Enhanced mobility in FAI-coated CsPbI₃ QDs

The TRTS measurements were performed under identical conditions for a 3-nm PbS QD film, a 6-nm PbSe QD film, a CsPbI₃ QD film, an FAI-coated CsPbI₃ QD film, and a conventional, nonquantized MAPbI₃ film. For thin films, conductivity is directly proportional to the change in terahertz transmission (28). Figure 4A shows the time-resolved change in terahertz transmission (ΔE/E) for each of the films probed. Mobility and lifetime values were extracted from the traces in Fig. 4B, as described in Materials and Methods and summarized in Fig. 4 (C and D, respectively). We find that the mobility for the control CsPbI₃ QD film is 0.23 cm² V⁻¹ s⁻¹, already significantly higher than that of the lead chalcogenide QD films (mobility of PbS and PbSe QD films were measured to be 0.042 and 0.090 cm² V⁻¹ s⁻¹, respectively). When the CsPbI₃ QD films are treated with FAI, the mobility improves from 0.23 to 0.50 cm² V⁻¹ s⁻¹. The mobility reaches a factor of 5 lower than the MAPbI₃ thin film (2.3 cm² V⁻¹ s⁻¹ in this study, similar to other literature reports) (27). We also compare the carrier lifetimes that are extracted from the terahertz measurements. The lifetimes of the control CsPbI₃ and FAI-coated CsPbI₃ QD films were found to be 1.8 and 1.6 ns, respectively (neglecting the early sub–50-ps dynamics), largely similar to the 2.2-ns lifetime of the MAPbI₃ film. The dynamics for the PbS and PbSe QD films are quite distinct. At the earliest times, the response decays rapidly; however, this is likely not recombination of carriers but rather a reduction in carrier mobility as carriers find lower-energy states with lower carrier mobility. After this initial decay, the mobility is reduced by another order of magnitude, and we find the mobility to be ~0.003 cm² V⁻¹ s⁻¹, which falls in-line with other widely reported measurements of the carrier mobility in these QD films (24–26).

Fig. 4. Improved mobility with FAI coating.

(A) Time-domain terahertz pulses through each of the films described in the text recorded at a pump delay time of ~1ps. (B) Time-resolved terahertz photoconductivity measurements of the films of CsPbI₃ control sample (blue) compared to the FAI-coated CsPbI₃ QD films (pink) along with a traditional MAPbI₃ (black) thin film and films of 6-nm PbSe QDs (orange) and 3-nm PbS QDs (green). (C and D) Comparison of the extracted (C) mobility and (D) terahertz lifetimes for each of the films.

Note that the carrier dynamics extracted with terahertz spectroscopy for both CsPbI₃ QD films are more similar to MAPbI₃ than PbS and PbSe carrier dynamics. This finding strongly suggests that the carrier transport mechanism is quite distinct in perovskite QD films compared to lead chalcogenide QD films. More work is under way exploring observed differences to understand their microscopic origins. For PV applications, mobility-lifetime product is of particular interest because of its relation to charge carrier diffusion length. The mobility-lifetime product was larger for the FAI-coated CsPbI₃ QD film than the untreated CsPbI₃ QD film (fig. S9) and justifies the enhanced charge extraction. Time-resolved PL (tr-PL) measurements (fig. S10) show an increase in PL lifetime for FAI-coated films on a longer time scale than what is measured using terahertz spectroscopy, whereas the decay dynamics within the terahertz window are similar.

DISCUSSION

Although QDs have the advantages of widely tunable bandgap and energetic position and, in the case of CsPbI₃, offer the added benefit of phase stability, QDSCs have traditionally suffered from both low V_OC relative to their bandgap and poor carrier collection efficiency. We demonstrate that both of these disadvantages of QDSCs can be, in large part, overcome by the coupling of CsPbI₃ QD films. The AX treatment strategy developed in this work provides a general method for tuning the electronic properties of the CsPbI₃ QD films. More specifically, the FAI coating yields a doubling of the already-high mobility of CsPbI₃ QD films and results in a certified record PCE of 13.43%, positioning QDSCs above the best reported PCE for dye-sensitized solar cells, organic PVs, and CZTSSe PV technologies. Additionally, the high V_OC, exceeding 80% of the Shockley-Queisser limit, and tunability of the bandgap make these devices well suited for use in high-efficiency tandem solar cell architectures.

MATERIALS AND METHODS

Materials

All chemicals were purchased from Sigma-Aldrich and used as received unless otherwise specified. Cesium carbonate (Cs₂CO₃; 99.9%), PbI₂ (99.9985%; Alfa Aesar), oleic acid (OA; technical grade, 90%), oleylamine (OAm; technical grade, 70%), 1-octadecene (ODE; technical grade, 90%), hexane (reagent grade, ≥95%), octane (anhydrous, ≥99%), MeOAc (anhydrous, 99.5%), Pb(NO₃)₂ (99.999%), EtOAc (anhydrous, 99.8%), CsI (99.999%), FAI (Dyesol), FABr (Dyesol), MAI (Dyesol), MABr (Dyesol), ethanol (EtOH; 200 proof, ≥99.5%), titanium ethoxide (≥97%), hydrochloric acid (HCl; 37% in water), 2,2′,7,7′-tetrakis(N,N-di-p-methoxyphenylamine)-9,9′-spirobifluorene (spiro-OMeTAD; ≥99.5%; Lumtec), chlorobenzene (anhydrous, 99.8%), 4-tert-butylpyridine (4-TBP; 96%), bis(trifluoromethane)sulfonimide lithium salt (Li-TFSI), and acetonitrile (anhydrous, 99.8%).

QD synthesis and purification

CsPbI₃ QDs were synthesized (7) and purified (4) using previously reported methods with slight modifications. Cs oleate in ODE solution (0.125 M) was synthesized by first degassing 0.407 g of Cs₂CO₃, 1.25 ml of OA, and 20 ml of ODE in a 100-ml three-neck flask under vacuum at 120°C for 30 min under stirring. After purging the flask with N₂, the flask was heated to 150°C until the reaction was complete and yielded a clear solution. The Cs-oleate in ODE solution was cooled and stored in N₂ until needed for a QD synthesis. To synthesize CsPbI₃ QDs, 0.5 g of PbI₂ and 25 ml of ODE were degassed in a 100-ml three-neck flask under vacuum at 120°C for 30 min under stirring. Preheated OA and OAm (130°C, 2.5 ml each) were added to the reaction flask and degassed briefly until the PbI₂ had completely dissolved. The reaction flask was purged with N₂ and then heated to the desired reaction temperature (typically 185°C) to achieve the desired nanocrystal size. Once the reaction temperature was reached, 2 ml of preheated Cs-oleate in ODE (130°C, 0.125 M) was injected into the reaction flask. The reaction was quenched in an ice bath after 5 to 10 s. To isolate the QDs, 35 ml of MeOAc was added to 15 ml of the QD reaction liquor and then centrifuged at 7500 rpm for 5 min. The supernatant was discarded, and the QD pellet was redispersed in ~5 ml of hexane. A minimal amount of MeOAc (~5 to 8 ml) was added to the QDs until the dispersion appeared cloudy, and then it was immediately centrifuged at 7500 rpm for 5 min. The resultant QD pellet was dispersed in 15 ml of hexane and then stored in the dark at 4°C for at least 24 hours to precipitate excess Cs-oleate and Pb-oleate. Before use, these solid precipitates were removed from the QD solution via centrifugation at 7500 rpm for 5 min.

Film fabrication

Coupled CsPbI₃ QD films were fabricated using previously reported methods (4). Saturated Pb(NO₃)₂ in MeOAc solution and AX salt (where AX = FAI, FABr, MAI, MABr, or CsI) in EtOAc solution were prepared by sonicating 20 mg of Pb(NO₃)₂ (or AX salt) and 20 ml of MeOAc (or EtOAc) for 10 min. Because all the salts were only slightly soluble in their corresponding solvent, excess salt was removed via centrifugation at 3500 rpm for 5 min. Each layer of CsPbI₃ QDs was spin-cast from a concentrated QD solution in octane (~75 mg/ml) at spin speeds of 1000 rpm for 20 s and 2000 rpm for 5 s. The film was then briefly dipped (~1 s) into the Pb(NO₃)₂ in MeOAc solution, rinsed in a solution of neat MeOAc, and then immediately dried in a stream of dry air. This process of spin-coating QDs and dipping in Pb(NO₃)₂ in MeOAc solution was repeated three to five times to achieve a total film thickness of 100 to 500 nm. Once the desired film thickness was achieved, the films were posttreated by soaking the film in the AX salt in EtOAc solution for 10 s before rinsing in MeOAc and then immediately drying under a stream of dry air. All film fabrications were carried out under dry ambient conditions (relative humidity, ~16 to 20%).

Film characterization

Ultraviolet-visible (UV-vis) absorption spectra were recorded using a Shimadzu UV-3600 UV-vis–near-infrared spectrophotometer. Absorptance was calculated from the reflectance and transmittance spectra acquired with an integrating sphere. Steady-state PL was measured using a Horiba Jobin Yvon FluoroMAX-4 spectrophotometer. Powder XRD data were recorded using a Bruker D8 Discover x-ray diffractometer with a Hi-Star 2D area detector using Cu Kα radiation (1.54 Å). FTIR spectra were acquired using a Nicolet 6700 FTIR spectrometer on QD films on Si substrates.

Time-of-flight secondary ion mass spectrometry

SIMS is a powerful analytical technique for determining elemental and isotopic distributions in solids, as well as the structure and composition of organic materials (29–31). An ION-TOF ToF-SIMS 5 ToF-SIMS spectrometer was used to depth-profile the perovskite QD films. Analysis was completed using a three-lens 30-kV BiMn primary ion gun, and the Bi³⁺ primary ion beam (operated in bunched mode; 20 ns pulse width, analysis current 0.7 pA) was scanned over a 25 × 25 micron area. Depth profiling was accomplished with a 3-kV oxygen ion sputter beam (10.8 nA of sputter current) raster of 150 × 150 micron area. All spectra during profiling were collected at or below a primary ion dose density of 1 × 10¹² ions/cm² to remain under the static SIMS limit. The data were plotted with the intensity for each signal at each data point normalized to the total ion counts measured at the data point, which diminishes artifacts from a changing ion yield in different layers as well as artifacts due to minor fluctuations in the primary ion beam current. We assigned the m/z signals at 45, 127, 206, and 266 to FA⁺, I⁺, ²⁰⁶Pb⁺, and Cs₂⁺, respectively, and simply referred to these signals as the FA, I, Pb, and Cs in the manuscript, respectively. The Cs₂⁺ and ²⁰⁶Pb⁺ signals were tracked because they were the most intense species for Cs⁺ and Pb²⁺, respectively, which did not lead to detector saturation. In contrast, Cs⁺, Pb²⁺, and other abundant lead isotopes saturated the detector under the measurement conditions used.

Atomic force microscopy

QD films were prepared on FTO-coated glass as described above. Samples were mounted using double-sided adhesive strips to metal disks exposing the QD layer. Imaging was performed with multimode scanning probe microscope equipped with a NanoScope IV controller (Bruker). Height images were obtained in soft-tapping mode using etched silicon probes (TESP, Bruker) with an autotuned resonance frequency range of 250 to 300 kHz at a scan rate of 2 Hz. Images were analyzed with NanoScope Analysis v1.2 software.

PV device fabrication

A ~50-nm TiO₂ layer was deposited via a sol-gel method onto prepatterned FTO on glass substrates (Thin Film Devices Inc.). Sol-gel TiO₂ was prepared by mixing 5 ml of EtOH, two drops of HCl, 125 μl of deionized water, and 375 μl of titanium ethoxide, resulting in a clear solution. The headspace of the vial was filled with nitrogen, and the solution was stirred for 48 hours and then kept in the freezer until use. The sol-gel was spin-cast at 3000 rpm for 20 s and annealed at 115° and 450°C for 30 min each. The CsPbI₃ QD photoactive layer was deposited using the procedure described above, resulting in a total thickness of 100 to 400 nm. The spiro-OMeTAD hole transporting material was spin-cast at 4000 rpm for 30 s from a solution with a nominal concentration of 72.3 mg of spiro-OMeTAD, 1 ml of chlorobenzene, 28.8 μl of 4-TBP, and 17.5 μl of Li-TFSI stock solution (520 mg/ml in acetonitrile). All the spin-coating processes were performed under ambient conditions. MoO_x was deposited at a rate of 0.2 to 1.0 Å/s at a base pressure lower than 2 × 10⁻⁷ torr for a total thickness of 15 nm. Al electrodes were evaporated at a rate ranging from 0.5 to 2 Å/s for a total thickness of 200 nm. A 70-nm layer of MgF₂ was evaporated on the glass side as an antireflective coating for the best-performing, certified PV device.

Device characterization

Devices were tested in a N₂-filled glovebox with a Newport Oriel Sol3A solar simulator with a xenon lamp source. Light intensity was calibrated with a KG5 filtered Si reference diode to minimize the spectral mismatch and to set the intensity of the lamp to 100 mW/cm² AM1.5 conditions. Devices were illuminated through a metal aperture (0.058 cm²). SPO was measured by holding the device at a constant voltage corresponding to the voltage at the maximum power point of the J-V scan. EQE measurements were taken using a Newport Oriel IQE200.

Certification

Devices certified by the NREL PV Performance Characterization Team were loaded into an air-free holder in a N₂-filled glovebox. The device active area was illuminated through a metal aperture with an area of 0.058 cm², which was measured by the PV Performance Characterization Team. As an accredited PV testing facility, the PV Performance Characterization Team received and tested the packaged solar cell in accordance with their independent testing procedures.

Terahertz spectroscopy

The films for terahertz measurements were deposited directly onto 1 × 2–cm quartz substrates following similar spin-coating procedures described above. PbS and PbSe QD samples were synthesized and fabricated as previously published using PbI₂ ligand exchange treatment (13, 32). MAPbI₃ films were prepared, as described by Ahn et al. (33), by spin coating from a N,N′-dimethylformamide/dimethyl sulfoxide solution containing stoichiometric MAI/PbI₂. The change in terahertz transmission was recorded as a function of delay from an optical pump pulse (wavelength, 516 nm) with a fluence of 5 × 10¹⁵ photons/cm² per pulse. For thin films, the mobility and carrier lifetimes are directly proportional to the change in terahertz transmission through the following relationship: (μ_e ⋅ f_e(τ) + μ_h ⋅ f_h(τ)) = [ΔE(τ)/E] ⋅ [(1 + n_s)cε₀/(e ⋅ J_abs)] (28). Here, ΔE/E is the measured terahertz response (see Fig. 4A) at a pump delay of τ, n_s is the refractive index of the substrate (n_s = 2.19 here), c is the speed of light, ε₀ is the permittivity of free space, e is the charge of an electron, and J_abs is the absorbed photon fluence and is related to the film optical density by J_abs = (1 − R) ⋅ (1 − 10^−OD) ⋅ J_in, where R is the reflection coefficient, and J_in is the measured photon fluence at the sample position. At τ = 0, both the normalized electron and hole carrier decay are 1 (f_e(0) = f_h(0) = 1), and the signal in Fig. 4B is the sum of the electron and hole mobilities: (μ_e + μ_h) = μ_s.

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/10/eaao4204/DC1

Supplementary Materials and Methods

fig. S1. Performance of FAI-coated and control devices.

fig. S2. Light absorption following FAI posttreatment.

fig. S3. Comparison of EQE with AX posttreatment.

fig. S4. Reproducibility of FAI-coated CsPbI₃ QD device performance.

fig. S5. XPS spectroscopy.

fig. S6. FTIR spectra of CsPbI₃ QDs.

fig. S7. Crystal structure of CsPbI₃ QDs.

fig. S8. CsPbI₃ QD film morphology.

fig. S9. Comparison of terahertz μ_s × τ product.

fig. S10. PL lifetime of CsPbI₃.

Reference (34)